From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-2.2 required=3.0 tests=HEADER_FROM_DIFFERENT_DOMAINS, MAILING_LIST_MULTI,SPF_PASS,URIBL_BLOCKED,USER_AGENT_MUTT autolearn=ham autolearn_force=no version=3.4.0 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id 1EAF9C6778D for ; Wed, 12 Sep 2018 17:14:49 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.kernel.org (Postfix) with ESMTP id CB5042087F for ; Wed, 12 Sep 2018 17:14:48 +0000 (UTC) DMARC-Filter: OpenDMARC Filter v1.3.2 mail.kernel.org CB5042087F Authentication-Results: mail.kernel.org; dmarc=fail (p=none dis=none) header.from=intel.com Authentication-Results: mail.kernel.org; spf=none smtp.mailfrom=linux-kernel-owner@vger.kernel.org Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1727760AbeILWUP (ORCPT ); Wed, 12 Sep 2018 18:20:15 -0400 Received: from mga02.intel.com ([134.134.136.20]:12102 "EHLO mga02.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1726428AbeILWUO (ORCPT ); Wed, 12 Sep 2018 18:20:14 -0400 X-Amp-Result: UNSCANNABLE X-Amp-File-Uploaded: False Received: from orsmga002.jf.intel.com ([10.7.209.21]) by orsmga101.jf.intel.com with ESMTP/TLS/DHE-RSA-AES256-GCM-SHA384; 12 Sep 2018 10:14:45 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.53,365,1531810800"; d="scan'208";a="91077724" Received: from sjchrist-coffee.jf.intel.com (HELO linux.intel.com) ([10.54.74.20]) by orsmga002.jf.intel.com with ESMTP; 12 Sep 2018 10:14:34 -0700 Date: Wed, 12 Sep 2018 10:14:34 -0700 From: Sean Christopherson To: Will Deacon Cc: linux-kernel@vger.kernel.org, linux-mm@kvack.org, cpandya@codeaurora.org, toshi.kani@hpe.com, tglx@linutronix.de, mhocko@suse.com, akpm@linux-foundation.org Subject: Re: [PATCH 4/5] lib/ioremap: Ensure phys_addr actually corresponds to a physical address Message-ID: <20180912171434.GA31712@linux.intel.com> References: <1536747974-25875-1-git-send-email-will.deacon@arm.com> <1536747974-25875-5-git-send-email-will.deacon@arm.com> <20180912150939.GA30274@linux.intel.com> <20180912163914.GA16071@arm.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20180912163914.GA16071@arm.com> User-Agent: Mutt/1.5.24 (2015-08-30) Sender: linux-kernel-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Wed, Sep 12, 2018 at 05:39:14PM +0100, Will Deacon wrote: > Hi Sean, > > Thanks for looking at the patch. > > On Wed, Sep 12, 2018 at 08:09:39AM -0700, Sean Christopherson wrote: > > On Wed, Sep 12, 2018 at 11:26:13AM +0100, Will Deacon wrote: > > > The current ioremap() code uses a phys_addr variable at each level of > > > page table, which is confusingly offset by subtracting the base virtual > > > address being mapped so that adding the current virtual address back on > > > when iterating through the page table entries gives back the corresponding > > > physical address. > > > > > > This is fairly confusing and results in all users of phys_addr having to > > > add the current virtual address back on. Instead, this patch just updates > > > phys_addr when iterating over the page table entries, ensuring that it's > > > always up-to-date and doesn't require explicit offsetting. > > > > > > Cc: Chintan Pandya > > > Cc: Toshi Kani > > > Cc: Thomas Gleixner > > > Cc: Michal Hocko > > > Cc: Andrew Morton > > > Signed-off-by: Will Deacon > > > --- > > > lib/ioremap.c | 28 ++++++++++++---------------- > > > 1 file changed, 12 insertions(+), 16 deletions(-) > > > > > > diff --git a/lib/ioremap.c b/lib/ioremap.c > > > index 6c72764af19c..fc834a59c90c 100644 > > > --- a/lib/ioremap.c > > > +++ b/lib/ioremap.c > > > @@ -101,19 +101,18 @@ static inline int ioremap_pmd_range(pud_t *pud, unsigned long addr, > > > pmd_t *pmd; > > > unsigned long next; > > > > > > - phys_addr -= addr; > > > pmd = pmd_alloc(&init_mm, pud, addr); > > > if (!pmd) > > > return -ENOMEM; > > > do { > > > next = pmd_addr_end(addr, end); > > > > > > - if (ioremap_try_huge_pmd(pmd, addr, next, phys_addr + addr, prot)) > > > + if (ioremap_try_huge_pmd(pmd, addr, next, phys_addr, prot)) > > > continue; > > > > > > - if (ioremap_pte_range(pmd, addr, next, phys_addr + addr, prot)) > > > + if (ioremap_pte_range(pmd, addr, next, phys_addr, prot)) > > > return -ENOMEM; > > > - } while (pmd++, addr = next, addr != end); > > > + } while (pmd++, addr = next, phys_addr += PMD_SIZE, addr != end); > > > > I think bumping phys_addr by PXX_SIZE is wrong if phys_addr and addr > > start unaligned with respect to PXX_SIZE. The addresses must be > > PAGE_ALIGNED, which lets ioremap_pte_range() do a simple calculation, > > but that doesn't hold true for the upper levels, i.e. phys_addr needs > > to be adjusted using an algorithm similar to pxx_addr_end(). > > > > Using a 2mb page as an example (lower 32 bits only): > > > > pxx_size = 0x00020000 > > pxx_mask = 0xfffe0000 > > addr = 0x1000 > > end = 0x00040000 > > phys_addr = 0x1000 > > > > Loop 1: > > addr = 0x1000 > > phys = 0x1000 > > > > Loop 2: > > addr = 0x20000 > > phys = 0x21000 > > Yes, I think you're completely right, however I also don't think this > can happen with the current code (and I've failed to trigger it in my > testing). The virtual addresses allocated for VM_IOREMAP allocations > are aligned to the order of the allocation, which means that the virtual > address at the start of the mapping is aligned such that when we hit the > end of a pXd, we know we've mapped the previous PXD_SIZE bytes. > > Having said that, this is clearly a change from the current code and I > haven't audited architectures other than arm64 (where IOREMAP_MAX_ORDER > corresponds to the maximum size of our huge mappings), so it would be > much better not to introduce this funny behaviour in a patch that aims > to reduce confusion in the first place! > > Fixing this using the pxx_addr_end() macros is a bit strange, since we > don't have a physical end variable (nor do we need one), so perhaps > something like changing the while condition to be: > > do { > ... > } while (pmd++, phys_addr += (next - addr), addr = next, addr != end); > > would do the trick. What do you reckon? LGTM. I like that there isn't a separate calculation for phys_addr's offset.