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 kanga.kvack.org (kanga.kvack.org [205.233.56.17]) (using TLSv1 with cipher DHE-RSA-AES256-SHA (256/256 bits)) (No client certificate requested) by smtp.lore.kernel.org (Postfix) with ESMTPS id C0562CD4F54 for ; Wed, 27 May 2026 12:24:30 +0000 (UTC) Received: by kanga.kvack.org (Postfix) id A4E046B00AD; Wed, 27 May 2026 08:24:29 -0400 (EDT) Received: by kanga.kvack.org (Postfix, from userid 40) id 9FE6D6B00AE; Wed, 27 May 2026 08:24:29 -0400 (EDT) X-Delivered-To: int-list-linux-mm@kvack.org Received: by kanga.kvack.org (Postfix, from userid 63042) id 8EDB96B00AF; Wed, 27 May 2026 08:24:29 -0400 (EDT) X-Delivered-To: linux-mm@kvack.org Received: from relay.hostedemail.com (smtprelay0017.hostedemail.com [216.40.44.17]) by kanga.kvack.org (Postfix) with ESMTP id 7B8916B00AD for ; Wed, 27 May 2026 08:24:29 -0400 (EDT) Received: from smtpin22.hostedemail.com (lb01a-stub [10.200.18.249]) by unirelay01.hostedemail.com (Postfix) with ESMTP id 2EDF91C1988 for ; Wed, 27 May 2026 12:24:29 +0000 (UTC) X-FDA: 84813117858.22.3143F0C Received: from foss.arm.com (foss.arm.com [217.140.110.172]) by imf06.hostedemail.com (Postfix) with ESMTP id 08BBC18000F for ; Wed, 27 May 2026 12:24:26 +0000 (UTC) Authentication-Results: imf06.hostedemail.com; dkim=pass header.d=arm.com header.s=foss header.b=lBe3ZX3V; spf=pass (imf06.hostedemail.com: domain of dev.jain@arm.com designates 217.140.110.172 as permitted sender) smtp.mailfrom=dev.jain@arm.com; dmarc=pass (policy=none) header.from=arm.com ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=hostedemail.com; s=arc-20220608; t=1779884667; h=from:from:sender:reply-to:subject:subject:date:date: message-id:message-id:to:to:cc:cc:mime-version:mime-version: content-type:content-type: content-transfer-encoding:content-transfer-encoding: in-reply-to:in-reply-to:references:references:dkim-signature; bh=cGeamoObf1qbjghhGJkIscnw/Fsp6/UIlrdZOPEQVao=; b=MM0qDd3ZgJYzfaQQv/N9a4SVDUt5eHhJbXfgWOVpkFKAQ6HfGt084xj/nlkeFFEMphg3Us U002iIU2OGF6Y4DdS51VPmM/c0Ba7DEcke/7AC8+8rKJi7My3wcM8eeL3HF/DB5FasdF4l 5rh9S7p2C1SFVz55cjdaOvFX/R8L91M= ARC-Seal: i=1; s=arc-20220608; d=hostedemail.com; t=1779884667; a=rsa-sha256; cv=none; b=GT3cbXYnWRjuPkfK3JkD3726OVwY4UHiXdiK+S7WWNtLSWbFCa+hY1qZ9VyYFJaek9lwAW Qb+2iPa4laUQtM7HzUxyJ1VlpaayVkw6UKTyh58qp/t6vnCTivmuO+bLkkJTFztsxXv8+I kMYSc1u7PMV/PIcsmh2gCvy589+mR4k= ARC-Authentication-Results: i=1; imf06.hostedemail.com; dkim=pass header.d=arm.com header.s=foss header.b=lBe3ZX3V; spf=pass (imf06.hostedemail.com: domain of dev.jain@arm.com designates 217.140.110.172 as permitted sender) smtp.mailfrom=dev.jain@arm.com; dmarc=pass (policy=none) header.from=arm.com 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 9B6942C3D; Wed, 27 May 2026 05:24:20 -0700 (PDT) Received: from [10.164.19.7] (unknown [10.164.19.7]) by usa-sjc-imap-foss1.foss.arm.com (Postfix) with ESMTPSA id A70A43F905; Wed, 27 May 2026 05:24:08 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=simple/simple; d=arm.com; s=foss; t=1779884665; bh=UotxsZ+zZr/XhydpYBQL8pv0WmoT4CF/hvzZHlKS1wI=; h=Date:Subject:To:Cc:References:From:In-Reply-To:From; b=lBe3ZX3VofaulxbGxDNMYHhGSvjPT3XdKH4EXZZetS4djxu0JFRfKn/pdasiSVkNW 2Nu5MPl/CJbYh+REJ+3ixRKC6FiHTlZ3FU5o3LG8N2CLRJTSRzn8/IxQ9AL2d/C7tr YIEo8Qurl5gJrZCjR8uegmh+i9jLeQziPgLEAD3w= Message-ID: <38ee9d6e-4331-48b0-82e4-2e6ae0aee705@arm.com> Date: Wed, 27 May 2026 17:54:05 +0530 MIME-Version: 1.0 User-Agent: Mozilla Thunderbird Subject: Re: [PATCH mm-unstable RFC v4 4/7] mm/huge_memory: refactor copy_huge_pmd() To: Yin Tirui , Andrew Morton , Matthew Wilcox , David Hildenbrand , Lorenzo Stoakes , Juergen Gross , Jonathan Cameron , Will Deacon Cc: Catalin Marinas , Peter Xu , Luiz Capitulino , Thomas Gleixner , Ingo Molnar , Borislav Petkov , Dave Hansen , "H . Peter Anvin" , Andy Lutomirski , Peter Zijlstra , Madhavan Srinivasan , Michael Ellerman , Nicholas Piggin , Christophe Leroy , "Liam R . Howlett" , Zi Yan , Baolin Wang , Nico Pache , Ryan Roberts , Barry Song , Lance Yang , Vlastimil Babka , Mike Rapoport , Suren Baghdasaryan , Michal Hocko , Anshuman Khandual , Rohan McLure , Kevin Brodsky , Alistair Popple , Andrew Donnellan , Pasha Tatashin , Baoquan He , Thomas Huth , Coiby Xu , Dan Williams , Yu-cheng Yu , Lu Baolu , Conor Dooley , Rik van Riel , wangkefeng.wang@huawei.com, chenjun102@huawei.com, linux-mm@kvack.org, linux-kernel@vger.kernel.org, x86@kernel.org, linux-arm-kernel@lists.infradead.org, linuxppc-dev@lists.ozlabs.org, linux-pm@vger.kernel.org References: <20260526145003.88445-1-yintirui@huawei.com> <20260526145003.88445-5-yintirui@huawei.com> Content-Language: en-US From: Dev Jain In-Reply-To: <20260526145003.88445-5-yintirui@huawei.com> Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 7bit X-Rspamd-Server: rspam11 X-Stat-Signature: 8hykjty8eie3qn3knnz69jmc4s5q3mwo X-Rspamd-Queue-Id: 08BBC18000F X-Rspam-User: X-HE-Tag: 1779884666-906362 X-HE-Meta: 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 mLhis1sn 475dRn/wwLtZDqZvc8iaa6qKjQuuC+xq7ARLpSAq5Trf6WNuPW4l0LpqG59dS138j1RK6haCtv1dQInNa6t5AwERf+ieuuIiIZPLKuaWNEQw7HQNcR6Z+rogicAIbVEdAbBBFgLweTkwVqpezMO92cimAOJnc3Y3kCsCOheTI2Q3ty5eHJw7Dmz4FJw/VhDR6x07P6aOxFhDkQEI/y8MXb709xset+FrBfH3O/Ef931jX1A99TzqAjI+ixjZz2Eha3FYyROPoAJpqEfOJhsuKuNGxYZ3O5qjp9XPGRqlLwz6pNjCsActFQX9TRJPaABIP0AmcmkPI4w0CSNunW+KxClVuqDZ+FS1AjNKsyu6nsEnsISrf4pIrOpu/r46XEmoM++dS Sender: owner-linux-mm@kvack.org Precedence: bulk X-Loop: owner-majordomo@kvack.org List-ID: List-Subscribe: List-Unsubscribe: On 26/05/26 8:20 pm, Yin Tirui wrote: > Classify the source PMD via pmd_present() and vm_normal_folio_pmd(), > matching the way the PTE path uses pte_present() and vm_normal_page(). > This moves the present-PMD decision from VMA identity checks to the > actual PMD/folio state. > > Drop the defensive "if (!pmd_trans_huge(pmd)) goto out_unlock" branch: > with mmap_write_lock held during fork, it should not occur. Split this from this patch? This is a functional change so will be better to review it separately. > > Extract the present-PMD side of copy_huge_pmd() into > copy_present_huge_pmd(). The helper owns the child pgtable passed by the > caller: it either deposits the pgtable when installing a copied PMD, or > frees it on paths that do not install one. > > The child pgtable is now allocated once up front and freed on every skip > path. This makes file/shmem and PFNMAP/special skip paths take the PMD > locks and free the preallocated pgtable before returning. These are not > expected to be hot paths, and the PFNMAP case is reused by the follow-up > PMD PFNMAP copy support. > > Signed-off-by: Yin Tirui > --- Since the series is still marked RFC, I think it would be better to send the refactor patches separately? > mm/huge_memory.c | 175 +++++++++++++++++++++++++---------------------- > 1 file changed, 95 insertions(+), 80 deletions(-) > > diff --git a/mm/huge_memory.c b/mm/huge_memory.c > index 9832ee910d5e..3964258ff91d 100644 > --- a/mm/huge_memory.c > +++ b/mm/huge_memory.c > @@ -1879,6 +1879,82 @@ bool touch_pmd(struct vm_area_struct *vma, unsigned long addr, > return false; > } > > +static int copy_present_huge_pmd( > + struct mm_struct *dst_mm, struct mm_struct *src_mm, > + pmd_t *dst_pmd, pmd_t *src_pmd, unsigned long addr, > + struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma, > + pmd_t pmd, pgtable_t pgtable, bool *need_split) > +{ > + struct folio *src_folio; > + bool wrprotect = true; > + > + src_folio = vm_normal_folio_pmd(src_vma, addr, pmd); > + if (!src_folio) { > + /* > + * When page table lock is held, the huge zero pmd should not be > + * under splitting since we don't split the page itself, only pmd to > + * a page table. > + */ > + if (is_huge_zero_pmd(pmd)) { > + /* > + * mm_get_huge_zero_folio() will never allocate a new > + * folio here, since we already have a zero page to > + * copy. It just takes a reference. > + */ > + mm_get_huge_zero_folio(dst_mm); > + goto set_pmd; > + } > + > + /* > + * Making sure it's not a CoW VMA with writable > + * mapping, otherwise it means either the anon page wrongly > + * applied special bit, or we made the PRIVATE mapping be > + * able to wrongly write to the backend MMIO. > + */ > + VM_WARN_ON_ONCE(is_cow_mapping(src_vma->vm_flags) && pmd_write(pmd)); > + pte_free(dst_mm, pgtable); > + pgtable = NULL; > + wrprotect = false; > + goto set_pmd; > + } > + > + /* File THPs are copied lazily by refaulting. */ > + if (!folio_test_anon(src_folio)) { > + pte_free(dst_mm, pgtable); > + return 0; > + } You removed !vma_is_anonymous to condense it into !folio_test_anon. For private-file mappings that is not true, but okay since PMD mapping is not supported. > + > + folio_get(src_folio); > + if (unlikely(folio_try_dup_anon_rmap_pmd(src_folio, > + &src_folio->page, > + dst_vma, src_vma))) { > + /* Page maybe pinned: split and retry the fault on PTEs. */ > + folio_put(src_folio); > + pte_free(dst_mm, pgtable); > + *need_split = true; > + return -EAGAIN; > + } > + add_mm_counter(dst_mm, MM_ANONPAGES, HPAGE_PMD_NR); > + > +set_pmd: > + if (pgtable) { > + mm_inc_nr_ptes(dst_mm); > + pgtable_trans_huge_deposit(dst_mm, dst_pmd, pgtable); > + } > + > + if (wrprotect) { > + pmdp_set_wrprotect(src_mm, addr, src_pmd); > + if (!userfaultfd_wp(dst_vma)) > + pmd = pmd_clear_uffd_wp(pmd); > + pmd = pmd_wrprotect(pmd); > + } > + > + pmd = pmd_mkold(pmd); > + set_pmd_at(dst_mm, addr, dst_pmd, pmd); > + > + return 0; > +} > + > static void copy_huge_non_present_pmd( > struct mm_struct *dst_mm, struct mm_struct *src_mm, > pmd_t *dst_pmd, pmd_t *src_pmd, unsigned long addr, > @@ -1940,104 +2016,43 @@ int copy_huge_pmd(struct mm_struct *dst_mm, struct mm_struct *src_mm, > struct vm_area_struct *dst_vma, struct vm_area_struct *src_vma) > { > spinlock_t *dst_ptl, *src_ptl; > - struct page *src_page; > - struct folio *src_folio; > - pmd_t pmd; > pgtable_t pgtable = NULL; > - int ret = -ENOMEM; > - > - pmd = pmdp_get_lockless(src_pmd); > - if (unlikely(pmd_present(pmd) && pmd_special(pmd) && > - !is_huge_zero_pmd(pmd))) { > - dst_ptl = pmd_lock(dst_mm, dst_pmd); > - src_ptl = pmd_lockptr(src_mm, src_pmd); > - spin_lock_nested(src_ptl, SINGLE_DEPTH_NESTING); > - /* > - * No need to recheck the pmd, it can't change with write > - * mmap lock held here. > - * > - * Meanwhile, making sure it's not a CoW VMA with writable > - * mapping, otherwise it means either the anon page wrongly > - * applied special bit, or we made the PRIVATE mapping be > - * able to wrongly write to the backend MMIO. > - */ > - VM_WARN_ON_ONCE(is_cow_mapping(src_vma->vm_flags) && pmd_write(pmd)); > - goto set_pmd; > - } > - > - /* Skip if can be re-fill on fault */ > - if (!vma_is_anonymous(dst_vma)) > - return 0; > + bool need_split = false; > + int ret = 0; > + pmd_t pmd; > > pgtable = pte_alloc_one(dst_mm); > if (unlikely(!pgtable)) > - goto out; > + return -ENOMEM; > > dst_ptl = pmd_lock(dst_mm, dst_pmd); > src_ptl = pmd_lockptr(src_mm, src_pmd); > spin_lock_nested(src_ptl, SINGLE_DEPTH_NESTING); > > - ret = -EAGAIN; > pmd = *src_pmd; > > - if (unlikely(thp_migration_supported() && > - pmd_is_valid_softleaf(pmd))) { > - copy_huge_non_present_pmd(dst_mm, src_mm, dst_pmd, src_pmd, addr, > + if (likely(pmd_present(pmd))) { > + ret = copy_present_huge_pmd(dst_mm, src_mm, dst_pmd, src_pmd, addr, > + dst_vma, src_vma, pmd, pgtable, &need_split); > + } else if (unlikely(thp_migration_supported() && pmd_is_valid_softleaf(pmd))) { > + if (unlikely(!vma_is_anonymous(dst_vma))) > + pte_free(dst_mm, pgtable); > + else > + copy_huge_non_present_pmd(dst_mm, src_mm, dst_pmd, src_pmd, addr, > dst_vma, src_vma, pmd, pgtable); > - ret = 0; > - goto out_unlock; > - } > - > - if (unlikely(!pmd_trans_huge(pmd))) { > + } else { > + VM_WARN_ONCE(1, "unexpected non-present PMD %llx\n", > + (unsigned long long)pmd_val(pmd)); > pte_free(dst_mm, pgtable); > - goto out_unlock; > - } > - /* > - * When page table lock is held, the huge zero pmd should not be > - * under splitting since we don't split the page itself, only pmd to > - * a page table. > - */ > - if (is_huge_zero_pmd(pmd)) { > - /* > - * mm_get_huge_zero_folio() will never allocate a new > - * folio here, since we already have a zero page to > - * copy. It just takes a reference. > - */ > - mm_get_huge_zero_folio(dst_mm); > - goto out_zero_page; > + ret = -EAGAIN; > } > > - src_page = pmd_page(pmd); > - VM_BUG_ON_PAGE(!PageHead(src_page), src_page); > - src_folio = page_folio(src_page); > + spin_unlock(src_ptl); > + spin_unlock(dst_ptl); > > - folio_get(src_folio); > - if (unlikely(folio_try_dup_anon_rmap_pmd(src_folio, src_page, dst_vma, src_vma))) { > - /* Page maybe pinned: split and retry the fault on PTEs. */ > - folio_put(src_folio); > - pte_free(dst_mm, pgtable); > - spin_unlock(src_ptl); > - spin_unlock(dst_ptl); > + if (unlikely(need_split)) > __split_huge_pmd(src_vma, src_pmd, addr, false); > - return -EAGAIN; > - } > - add_mm_counter(dst_mm, MM_ANONPAGES, HPAGE_PMD_NR); > -out_zero_page: > - mm_inc_nr_ptes(dst_mm); > - pgtable_trans_huge_deposit(dst_mm, dst_pmd, pgtable); > - pmdp_set_wrprotect(src_mm, addr, src_pmd); > - if (!userfaultfd_wp(dst_vma)) > - pmd = pmd_clear_uffd_wp(pmd); > - pmd = pmd_wrprotect(pmd); > -set_pmd: > - pmd = pmd_mkold(pmd); > - set_pmd_at(dst_mm, addr, dst_pmd, pmd); > > - ret = 0; > -out_unlock: > - spin_unlock(src_ptl); > - spin_unlock(dst_ptl); > -out: > return ret; > } >