From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-qk0-f198.google.com (mail-qk0-f198.google.com [209.85.220.198]) by kanga.kvack.org (Postfix) with ESMTP id 8E5A62803E9 for ; Wed, 23 Aug 2017 14:12:31 -0400 (EDT) Received: by mail-qk0-f198.google.com with SMTP id n27so3932137qki.2 for ; Wed, 23 Aug 2017 11:12:31 -0700 (PDT) Received: from mx1.redhat.com (mx1.redhat.com. [209.132.183.28]) by mx.google.com with ESMTPS id w41si1956839qtk.104.2017.08.23.11.12.30 for (version=TLS1_2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128/128); Wed, 23 Aug 2017 11:12:30 -0700 (PDT) From: Andrea Arcangeli Subject: [PATCH 1/1] userfaultfd: non-cooperative: closing the uffd without triggering SIGBUS Date: Wed, 23 Aug 2017 20:12:27 +0200 Message-Id: <20170823181227.19926-1-aarcange@redhat.com> Sender: owner-linux-mm@kvack.org List-ID: To: Andrew Morton , linux-mm@kvack.org Cc: Mike Rapoport , Mike Kravetz , Pavel Emelyanov This is an enhancement to avoid a non cooperative userfaultfd manager having to unregister all regions before it can close the uffd after all userfaultfd activity completed. The UFFDIO_UNREGISTER would serialize against the handle_userfault by taking the mmap_sem for writing, but we can simply repeat the page fault if we detect the uffd was closed and so the regular page fault paths should takeover. Acked-by: Mike Rapoport Signed-off-by: Andrea Arcangeli --- fs/userfaultfd.c | 20 +++++++++++++++++++- 1 file changed, 19 insertions(+), 1 deletion(-) diff --git a/fs/userfaultfd.c b/fs/userfaultfd.c index 272c21d8d532..186831c80a75 100644 --- a/fs/userfaultfd.c +++ b/fs/userfaultfd.c @@ -394,8 +394,26 @@ int handle_userfault(struct vm_fault *vmf, unsigned long reason) * in __get_user_pages if userfaultfd_release waits on the * caller of handle_userfault to release the mmap_sem. */ - if (unlikely(ACCESS_ONCE(ctx->released))) + if (unlikely(ACCESS_ONCE(ctx->released))) { + /* + * Don't return VM_FAULT_SIGBUS in this case, so a non + * cooperative manager can close the uffd after the + * last UFFDIO_COPY, without risking to trigger an + * involuntary SIGBUS if the process was starting the + * userfaultfd while the userfaultfd was still armed + * (but after the last UFFDIO_COPY). If the uffd + * wasn't already closed when the userfault reached + * this point, that would normally be solved by + * userfaultfd_must_wait returning 'false'. + * + * If we were to return VM_FAULT_SIGBUS here, the non + * cooperative manager would be instead forced to + * always call UFFDIO_UNREGISTER before it can safely + * close the uffd. + */ + ret = VM_FAULT_NOPAGE; goto out; + } /* * Check that we can return VM_FAULT_RETRY. -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@kvack.org. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: email@kvack.org