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 Received: from vger.kernel.org (vger.kernel.org [23.128.96.18]) by smtp.lore.kernel.org (Postfix) with ESMTP id A1830C4332F for ; Mon, 21 Nov 2022 12:28:57 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S230367AbiKUM24 (ORCPT ); Mon, 21 Nov 2022 07:28:56 -0500 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:46546 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S230101AbiKUM2g (ORCPT ); Mon, 21 Nov 2022 07:28:36 -0500 Received: from foss.arm.com (foss.arm.com [217.140.110.172]) by lindbergh.monkeyblade.net (Postfix) with ESMTP id A57965FDF for ; Mon, 21 Nov 2022 04:27:55 -0800 (PST) Received: from usa-sjc-imap-foss1.foss.arm.com (unknown [10.121.207.14]) by usa-sjc-mx-foss1.foss.arm.com (Postfix) with ESMTP id B8ACB23A; Mon, 21 Nov 2022 04:28:01 -0800 (PST) Received: from FVFF77S0Q05N.cambridge.arm.com (FVFF77S0Q05N.cambridge.arm.com [10.1.25.132]) by usa-sjc-imap-foss1.foss.arm.com (Postfix) with ESMTPSA id 299953F7D8; Mon, 21 Nov 2022 04:27:54 -0800 (PST) Date: Mon, 21 Nov 2022 12:27:51 +0000 From: Mark Rutland To: Anshuman Khandual Cc: Nathan Chancellor , linux-arm-kernel@lists.infradead.org, Catalin Marinas , Will Deacon , Andrew Morton , linux-kernel@vger.kernel.org, llvm@lists.linux.dev Subject: Re: [PATCH] arm64/mm: Drop redundant BUG_ON(!pgtable_alloc) Message-ID: References: <20221118053102.500216-1-anshuman.khandual@arm.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Mon, Nov 21, 2022 at 11:00:42AM +0530, Anshuman Khandual wrote: > Hello Nathan, > > Thanks for the report. > > On 11/20/22 21:46, Nathan Chancellor wrote: > > Hi Anshuman, > > I just bisected a boot failure in our QEMU-based continuous integration > > setup to this change as commit 9ed2b4616d4e ("arm64/mm: Drop redundant > > BUG_ON(!pgtable_alloc)") in the arm64 tree. There is no output so the > > panic clearly happens early at boot. If I move back to the previous > > commit and add a WARN_ON() like so: > > > > diff --git a/arch/arm64/mm/mmu.c b/arch/arm64/mm/mmu.c > > index d386033a074c..9280a92ff920 100644 > > --- a/arch/arm64/mm/mmu.c > > +++ b/arch/arm64/mm/mmu.c > > @@ -383,6 +383,7 @@ static void __create_pgd_mapping_locked(pgd_t *pgdir, phys_addr_t phys, > > phys &= PAGE_MASK; > > addr = virt & PAGE_MASK; > > end = PAGE_ALIGN(virt + size); > > + WARN_ON(!pgtable_alloc); > > > > do { > > next = pgd_addr_end(addr, end); > > > > I do see some stacktraces. I have attached the boot log from QEMU. > > > > If there is any additional information I can provide or patches I can > > test, I am more than happy to do so. > > There are couple of instances, where __create_pgd_mapping() function gets called > without a valid pgtable alloc function (NULL is passed on instead), as it is not > expected to allocate page table pages, during the mapping process. The following > change after this patch should solve the reported problem. > > diff --git a/arch/arm64/mm/mmu.c b/arch/arm64/mm/mmu.c > index 9ea8e9039992..a00563122fcb 100644 > --- a/arch/arm64/mm/mmu.c > +++ b/arch/arm64/mm/mmu.c > @@ -42,6 +42,7 @@ > #define NO_BLOCK_MAPPINGS BIT(0) > #define NO_CONT_MAPPINGS BIT(1) > #define NO_EXEC_MAPPINGS BIT(2) /* assumes FEAT_HPDS is not used */ > +#define NO_ALLOC_MAPPINGS BIT(3) /* does not allocate page table pages */ > > int idmap_t0sz __ro_after_init; > > @@ -380,7 +381,7 @@ static void __create_pgd_mapping_locked(pgd_t *pgdir, phys_addr_t phys, > phys &= PAGE_MASK; > addr = virt & PAGE_MASK; > end = PAGE_ALIGN(virt + size); > - BUG_ON(!pgtable_alloc); > + BUG_ON(!(flags & NO_ALLOC_MAPPINGS) && !pgtable_alloc); > > do { > next = pgd_addr_end(addr, end); > @@ -453,7 +454,7 @@ static void __init create_mapping_noalloc(phys_addr_t phys, unsigned long virt, > return; > } > __create_pgd_mapping(init_mm.pgd, phys, virt, size, prot, NULL, > - NO_CONT_MAPPINGS); > + NO_CONT_MAPPINGS | NO_ALLOC_MAPPINGS); > } > > void __init create_pgd_mapping(struct mm_struct *mm, phys_addr_t phys, > @@ -481,7 +482,7 @@ static void update_mapping_prot(phys_addr_t phys, unsigned long virt, > } > > __create_pgd_mapping(init_mm.pgd, phys, virt, size, prot, NULL, > - NO_CONT_MAPPINGS); > + NO_CONT_MAPPINGS | NO_ALLOC_MAPPINGS); > > /* flush the TLBs after updating live kernel mappings */ > flush_tlb_kernel_range(virt, virt + size); This is now more complicated than what we had originally, and it doesn't catch the case where the caller sets NO_ALLOC_MAPPINGS but the callee ends up needing to perform an allocation, which the old code would have caught. This is clearly more subtle than we thought initially; for now could we please just drop the patch? Thanks, Mark.