Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754990AbbK3PSv (ORCPT ); Mon, 30 Nov 2015 10:18:51 -0500 Received: from mx2.suse.de ([195.135.220.15]:42723 "EHLO mx2.suse.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754901AbbK3PSs (ORCPT ); Mon, 30 Nov 2015 10:18:48 -0500 Subject: Re: mm: BUG in __munlock_pagevec To: Sasha Levin , "Kirill A. Shutemov" References: <565C5C38.3040705@oracle.com> Cc: "linux-mm@kvack.org" , LKML From: Vlastimil Babka Message-ID: <565C68D3.7000001@suse.cz> Date: Mon, 30 Nov 2015 16:18:43 +0100 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:38.0) Gecko/20100101 Thunderbird/38.3.0 MIME-Version: 1.0 In-Reply-To: <565C5C38.3040705@oracle.com> Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 2466 Lines: 57 On 11/30/2015 03:24 PM, Sasha Levin wrote: > Hi all, > > I've hit the following while fuzzing with trinity on the latest -next kernel: > > > [ 850.305385] page:ffffea001a5a0f00 count:0 mapcount:1 mapping:dead000000000400 index:0x1ffffffffff > [ 850.306773] flags: 0x2fffff80000000() > [ 850.307175] page dumped because: VM_BUG_ON_PAGE(1 && PageTail(page)) > [ 850.308027] page_owner info is not active (free page?) > [ 850.308925] ------------[ cut here ]------------ > [ 850.309614] kernel BUG at include/linux/page-flags.h:326! > [ 850.310333] invalid opcode: 0000 [#1] PREEMPT SMP KASAN > [ 850.311176] Modules linked in: > [ 850.311650] CPU: 5 PID: 7051 Comm: trinity-c129 Not tainted 4.4.0-rc2-next-20151127-sasha-00012-gf0498ca-dirty #2661 > [ 850.313115] task: ffff8806eaf08000 ti: ffff8806b1170000 task.ti: ffff8806b1170000 > [ 850.314085] RIP: __munlock_pagevec (include/linux/page-flags.h:326 mm/mlock.c:296) That's TestClearPageMlocked(page) which has PF_NO_TAIL. The page dump suggests the page was freed between the check triggering, and the page being dumped. But being on munlock's pagevec should pin the page. So a pin/unpin mismatch somewhere, together with a race? Moreover, a PageTail(page) shouldn't even get on the pagevec, munlock_vma_pages_range() skips tail pages. So another race that made the page a Tail after it was added to pagevec? Or maybe __munlock_pagevec_fill() encountered a tail page, and since it assumes that it can't happen, there's no check. Maybe a VM_BUG_ON_PAGE() there would catch this earlier? Could be related to "thp: allow mlocked THP again". Ah, __munlock_pagevec_fill() does a get_page(), which would increase page->count on the compound head, which could also explain the mismatch. ------8<------ diff --git a/mm/mlock.c b/mm/mlock.c index af421d8bd6da..156d2840aa62 100644 --- a/mm/mlock.c +++ b/mm/mlock.c @@ -393,7 +393,9 @@ static unsigned long __munlock_pagevec_fill(struct pagevec *pvec, if (!page || page_zone_id(page) != zoneid) break; + VM_BUG_ON_PAGE(PageTail(page), page); get_page(page); + /* * Increase the address that will be returned *before* the * eventual break due to pvec becoming full by adding the page -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/