From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S935983AbcIWMmy (ORCPT ); Fri, 23 Sep 2016 08:42:54 -0400 Received: from mail-pf0-f194.google.com ([209.85.192.194]:34931 "EHLO mail-pf0-f194.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S932390AbcIWMmt (ORCPT ); Fri, 23 Sep 2016 08:42:49 -0400 Date: Fri, 23 Sep 2016 14:42:45 +0200 From: Michal Hocko To: zijun_hu Cc: zijun_hu@htc.com, Andrew Morton , linux-mm@kvack.org, linux-kernel@vger.kernel.org, tj@kernel.org, mingo@kernel.org, rientjes@google.com, iamjoonsoo.kim@lge.com, mgorman@techsingularity.net Subject: Re: [PATCH 1/1] lib/ioremap.c: avoid endless loop under ioremapping page unaligned ranges Message-ID: <20160923124244.GN4478@dhcp22.suse.cz> References: <57E20A69.5010206@zoho.com> <20160922124735.GB11204@dhcp22.suse.cz> <35661a34-c3e0-0ec2-b58f-ee59bef4e4d4@zoho.com> <20160923084551.GG4478@dhcp22.suse.cz> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: User-Agent: Mutt/1.6.0 (2016-04-01) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Fri 23-09-16 20:29:20, zijun_hu wrote: > On 2016/9/23 16:45, Michal Hocko wrote: > > On Thu 22-09-16 23:13:17, zijun_hu wrote: > >> On 2016/9/22 20:47, Michal Hocko wrote: > >>> On Wed 21-09-16 12:19:53, zijun_hu wrote: > >>>> From: zijun_hu > >>>> > >>>> endless loop maybe happen if either of parameter addr and end is not > >>>> page aligned for kernel API function ioremap_page_range() > >>> > >>> Does this happen in practise or this you found it by reading the code? > >>> > >> i found it by reading the code, this is a kernel API function and there > >> are no enough hint for parameter requirements, so any parameters > >> combination maybe be used by user, moreover, it seems appropriate for > >> many bad parameter combination, for example, provided PMD_SIZE=2M and > >> PAGE_SIZE=4K, 0x00 is used for aligned very well address > >> a user maybe want to map virtual range[0x1ff800, 0x200800) to physical address > >> 0x300800, it will cause endless loop > > > > Well, we are relying on the kernel to do the sane thing otherwise we > > would be screwed anyway. If this can be triggered by a userspace then it > > would be a different story. Just look at how we are doing mmap, we > > sanitize the page alignment at the high level and the lower level > > functions just assume sane values. > > > ioremap_page_range() is exported by EXPORT_SYMBOL_GPL() as a kernel interface > so perhaps it is called by not only any kernel module authors but also other > kernel parts > > if the bad range is used by a careless kernel user really, it seems a better > choice to alert the warning message or panic the kernel than hanging the system > due to endless loop, it can help them locate problem usefully I absolutely do not want to panic my system just because a crapy module or whatnot doesn't provide an aligned address. Warning and a fixup sounds much more sane to me. [...] > >> no, it don't work for many special case > >> for example, provided PMD_SIZE=2M > >> mapping [0x1f8800, 0x208800) virtual range will be split to two ranges > >> [0x1f8800, 0x200000) and [0x200000,0x208800) and map them separately > >> the first range will cause dead loop > > > > I am not sure I see your point. How can we deadlock if _both_ addresses > > get aligned to the page boundary and how does PMD_SIZE make any > > difference. > > > i will take a example to illustrate my considerations > provided PUD_SIZE == 1G, PMD_SIZE == 2M, PAGE_SIZE == 4K > it is used by arm64 normally > > we want to map virtual range [0xffffffff_ffc08800, 0xffffffff_fffff800) by > ioremap_page_range(),ioremap_pmd_range() is called to map the range > finally, ioremap_pmd_range() will call > ioremap_pte_range(pmd, 0xffffffff_ffc08800, 0xffffffff_fffe0000) and > ioremap_pte_range(pmd, 0xffffffff_fffe0000, 0xffffffff fffff800) separately but those ranges are not aligned and it ioremap_page_range fix them up to _be_ aligned then there is no problem, right? So either I am missing something or we are talking past each other. -- Michal Hocko SUSE Labs