From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from lindbergh.monkeyblade.net ([23.128.96.19]:34748 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1727032AbgIJSNe (ORCPT ); Thu, 10 Sep 2020 14:13:34 -0400 Received: from mail-qt1-x844.google.com (mail-qt1-x844.google.com [IPv6:2607:f8b0:4864:20::844]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 845C5C061757 for ; Thu, 10 Sep 2020 11:13:21 -0700 (PDT) Received: by mail-qt1-x844.google.com with SMTP id h6so5600427qtd.6 for ; Thu, 10 Sep 2020 11:13:21 -0700 (PDT) Date: Thu, 10 Sep 2020 15:13:19 -0300 From: Jason Gunthorpe Subject: Re: [RFC PATCH v2 1/3] mm/gup: fix gup_fast with dynamic page table folding Message-ID: <20200910181319.GO87483@ziepe.ca> References: <20200907180058.64880-1-gerald.schaefer@linux.ibm.com> <20200907180058.64880-2-gerald.schaefer@linux.ibm.com> <0dbc6ec8-45ea-0853-4856-2bc1e661a5a5@intel.com> <20200909142904.00b72921@thinkpad> <20200909192534.442f8984@thinkpad> <20200909180324.GI87483@ziepe.ca> <20200910093925.GB29166@oc3871087118.ibm.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: Sender: linux-s390-owner@vger.kernel.org List-ID: To: Linus Torvalds Cc: Alexander Gordeev , Gerald Schaefer , Dave Hansen , John Hubbard , LKML , linux-mm , linux-arch , Andrew Morton , Russell King , Mike Rapoport , Catalin Marinas , Will Deacon , Michael Ellerman , Benjamin Herrenschmidt , Paul Mackerras , Jeff Dike , Richard Weinberger , Dave Hansen , Andy Lutomirski , Peter Zijlstra , Thomas Gleixner , Ingo Molnar , Borislav Petkov , Arnd Bergmann , Andrey Ryabinin , linux-x86 , linux-arm , linux-power , linux-sparc , linux-um , linux-s390 , Vasily Gorbik , Heiko Carstens , Christian Borntraeger , Claudio Imbrenda On Thu, Sep 10, 2020 at 10:35:38AM -0700, Linus Torvalds wrote: > On Thu, Sep 10, 2020 at 2:40 AM Alexander Gordeev > wrote: > > > > It is only gup_fast case that exposes the issue. It hits because > > pointers to stack copies are passed to gup_pXd_range iterators, not > > pointers to real page tables itself. > > Can we possibly change fast-gup to not do the stack copies? > > I'd actually rather do something like that, than the "addr_end" thing. > As you say, none of the other page table walking code does what the > GUP code does, and I don't think it's required. As I understand it, the requirement is because fast-gup walks without the page table spinlock, or mmap_sem held so it must READ_ONCE the *pXX. It then checks that it is a valid page table pointer, then calls pXX_offset(). The arch implementation of pXX_offset() derefs again the passed pXX pointer. So it defeats the READ_ONCE and the 2nd load could observe something that is no longer a page table pointer and crash. Passing it the address of the stack value is a way to force pXX_offset() to use the READ_ONCE result which has already been tested to be a page table pointer. Other page walking code that holds the mmap_sem tends to use pmd_trans_unstable() which solves this problem by injecting a barrier. The load hidden in pte_offset() after a pmd_trans_unstable() can't be re-ordered and will only see a page table entry under the mmap_sem. However, I think that logic would have been much clearer following the GUP model of READ_ONCE vs extra reads and a hidden barrier. At least it took me a long time to work it out :( I also think there are real bugs here where places are reading *pXX multiple times without locking the page table. One was found recently in the wild in the huge tlb code IIRC. The mm/pagewalk.c has these missing READ_ONCE bugs too. So.. To change away from the stack option I think we'd have to pass the READ_ONCE value to pXX_offset() as an extra argument instead of it derefing the pointer internally. Jason