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=-0.8 required=3.0 tests=HEADER_FROM_DIFFERENT_DOMAINS, MAILING_LIST_MULTI,SPF_PASS 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 AF1F5C3279B for ; Fri, 6 Jul 2018 08:58:27 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.kernel.org (Postfix) with ESMTP id 7143423EF1 for ; Fri, 6 Jul 2018 08:58:27 +0000 (UTC) DMARC-Filter: OpenDMARC Filter v1.3.2 mail.kernel.org 7143423EF1 Authentication-Results: mail.kernel.org; dmarc=none (p=none dis=none) header.from=arm.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 S1753684AbeGFI6W (ORCPT ); Fri, 6 Jul 2018 04:58:22 -0400 Received: from usa-sjc-mx-foss1.foss.arm.com ([217.140.101.70]:32836 "EHLO foss.arm.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753075AbeGFI6U (ORCPT ); Fri, 6 Jul 2018 04:58:20 -0400 Received: from usa-sjc-imap-foss1.foss.arm.com (unknown [10.72.51.249]) by usa-sjc-mx-foss1.foss.arm.com (Postfix) with ESMTP id 6A441ED1; Fri, 6 Jul 2018 01:58:20 -0700 (PDT) Received: from [10.1.206.34] (melchizedek.cambridge.arm.com [10.1.206.34]) by usa-sjc-imap-foss1.foss.arm.com (Postfix) with ESMTPSA id BD5F63F2EA; Fri, 6 Jul 2018 01:58:17 -0700 (PDT) Subject: Re: [PATCH v3 3/5] arm64/mm: Create initial page tables in init_pg_dir To: Jun Yao , linux-arm-kernel@lists.infradead.org Cc: catalin.marinas@arm.com, will.deacon@arm.com, suzuki.poulose@arm.com, linux-kernel@vger.kernel.org References: <20180702111659.25570-1-yaojun8558363@gmail.com> <20180702111659.25570-4-yaojun8558363@gmail.com> From: James Morse Message-ID: <3977afaa-63ae-e390-7660-4d25146140b8@arm.com> Date: Fri, 6 Jul 2018 09:58:16 +0100 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:52.0) Gecko/20100101 Thunderbird/52.8.0 MIME-Version: 1.0 In-Reply-To: <20180702111659.25570-4-yaojun8558363@gmail.com> Content-Type: text/plain; charset=utf-8 Content-Language: en-US Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Hi Jun, On 02/07/18 12:16, Jun Yao wrote: > Create initial page tables in init_pg_dir and then create final > page tables in swapper_pg_dir directly. This is what the patch does, but doesn't preserve the why for the git-log. Could you expand it to describe why we're doing this. The series so far fails to boot from me. This is because the kaslr code tries to read the kaslr-seed from the DT, via the fixmap. To do this, it needs the fixmap tables installed, which early_fixmap_init() does. early_fixmap_init() calls pgd_offset_k(), which assumes init_mm.pgd is in use, so we hit a BUG_ON() when trying to setup the fixmap because we added the tables to the wrong page tables. If you enable 'CONFIG_RANDOMIZE_BASE', even without EFI you should see the same thing happen. I think we should move this hunk: > diff --git a/arch/arm64/kernel/setup.c b/arch/arm64/kernel/setup.c > index 30ad2f085d1f..b065c08d4008 100644 > --- a/arch/arm64/kernel/setup.c > +++ b/arch/arm64/kernel/setup.c > @@ -249,6 +249,7 @@ void __init setup_arch(char **cmdline_p) > init_mm.end_code = (unsigned long) _etext; > init_mm.end_data = (unsigned long) _edata; > init_mm.brk = (unsigned long) _end; > + init_mm.pgd = init_pg_dir; > > *cmdline_p = boot_command_line; > into early_fixmap_init(), which is the first thing setup_arch() calls. Something like [0] fixes it. This looks to be the only thing that goes near init_mm.pgd before paging_init(). > diff --git a/arch/arm64/include/asm/pgtable.h b/arch/arm64/include/asm/pgtable.h > index 7c4c8f318ba9..3b408f21fe2e 100644 > --- a/arch/arm64/include/asm/pgtable.h > +++ b/arch/arm64/include/asm/pgtable.h > @@ -718,6 +718,8 @@ static inline pmd_t pmdp_establish(struct vm_area_struct *vma, > } > #endif > > +extern pgd_t init_pg_dir[PTRS_PER_PGD] __section(.init.data); > +extern pgd_t init_pg_end[] __section(.init.data); normally we'd spell this '__initdata', but if we move them out of the '.init.data' section we shouldn't, otherwise the extern definition doesn't match where the symbol appears. (Looks like I was wrong to think that tools like sparse pick this up, that's just the __iomem/__user stuff.) > diff --git a/arch/arm64/mm/mmu.c b/arch/arm64/mm/mmu.c > index 2dbb2c9f1ec1..a7ab0010ff80 100644 > --- a/arch/arm64/mm/mmu.c > +++ b/arch/arm64/mm/mmu.c > @@ -628,26 +628,10 @@ static void __init map_kernel(pgd_t *pgdp) > */ > void __init paging_init(void) > { > - phys_addr_t pgd_phys = early_pgtable_alloc(); > - pgd_t *pgdp = pgd_set_fixmap(pgd_phys); > - > - map_kernel(pgdp); > - map_mem(pgdp); > - > - /* > - * We want to reuse the original swapper_pg_dir so we don't have to > - * communicate the new address to non-coherent secondaries in > - * secondary_entry, and so cpu_switch_mm can generate the address with > - * adrp+add rather than a load from some global variable. > - * > - * To do this we need to go via a temporary pgd. > - */ > - cpu_replace_ttbr1(__va(pgd_phys)); > - memcpy(swapper_pg_dir, pgdp, PGD_SIZE); > - cpu_replace_ttbr1(lm_alias(swapper_pg_dir)); > - > - pgd_clear_fixmap(); > - memblock_free(pgd_phys, PAGE_SIZE); > + map_kernel(swapper_pg_dir); > + map_mem(swapper_pg_dir); > + cpu_replace_ttbr1(swapper_pg_dir); The lm_alias() here is important: cpu_replace_ttbr1() calls virt_to_phys() on its argument, virt_to_phys() is only intended for addresses in the linear-map region of the VA space, as it works by doing some arithmetic with the address. swapper_pg_dir is the name of the kernel symbol, so its address will be in the kernel-text region of the VA space. Today virt_to_phys() catches this happening and fixes it, CONFIG_DEBUG_VIRTUAL will give you a warning, at some point virt_to_phys()'s safety net will go-away. The original call that did this was wrapped in lm_alias(), which gives you the linear-map address of a symbol in the kernel-text. Thanks, James [0] make sure fixmap tables go in the init page tables diff --git a/arch/arm64/mm/mmu.c b/arch/arm64/mm/mmu.c index 117d080639b3..e097c78a66f8 100644 --- a/arch/arm64/mm/mmu.c +++ b/arch/arm64/mm/mmu.c @@ -770,6 +771,13 @@ void __init early_fixmap_init(void) pmd_t *pmdp; unsigned long addr = FIXADDR_START; + /* + * During early setup we use init_pg_dir, update init_mm so its + * implicit use by pgd_offset_k() gets the live page tables. + * swapper_pg_dir is restored by paging_init(). + */ + init_mm.pgd = init_pg_dir; + pgdp = pgd_offset_k(addr); pgd = READ_ONCE(*pgdp); if (CONFIG_PGTABLE_LEVELS > 3 &&