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=-9.8 required=3.0 tests=HEADER_FROM_DIFFERENT_DOMAINS, INCLUDES_PATCH,MAILING_LIST_MULTI,SIGNED_OFF_BY,SPF_HELO_NONE,SPF_PASS, USER_AGENT_GIT 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 910C2C2BB86 for ; Mon, 13 Apr 2020 12:52:40 +0000 (UTC) Received: from kanga.kvack.org (kanga.kvack.org [205.233.56.17]) by mail.kernel.org (Postfix) with ESMTP id 44FB72073E for ; Mon, 13 Apr 2020 12:52:40 +0000 (UTC) DMARC-Filter: OpenDMARC Filter v1.3.2 mail.kernel.org 44FB72073E Authentication-Results: mail.kernel.org; dmarc=fail (p=none dis=none) header.from=linux.intel.com Authentication-Results: mail.kernel.org; spf=pass smtp.mailfrom=owner-linux-mm@kvack.org Received: by kanga.kvack.org (Postfix) id 1078B8E0109; Mon, 13 Apr 2020 08:52:31 -0400 (EDT) Received: by kanga.kvack.org (Postfix, from userid 40) id 0403C8E0104; Mon, 13 Apr 2020 08:52:30 -0400 (EDT) X-Delivered-To: int-list-linux-mm@kvack.org Received: by kanga.kvack.org (Postfix, from userid 63042) id DFA0C8E0109; Mon, 13 Apr 2020 08:52:30 -0400 (EDT) X-Delivered-To: linux-mm@kvack.org Received: from forelay.hostedemail.com (smtprelay0232.hostedemail.com [216.40.44.232]) by kanga.kvack.org (Postfix) with ESMTP id BF1438E0104 for ; Mon, 13 Apr 2020 08:52:30 -0400 (EDT) Received: from smtpin30.hostedemail.com (10.5.19.251.rfc1918.com [10.5.19.251]) by forelay01.hostedemail.com (Postfix) with ESMTP id 7EDE3180AD807 for ; Mon, 13 Apr 2020 12:52:30 +0000 (UTC) X-FDA: 76702820460.30.force68_32ea0407af562 X-HE-Tag: force68_32ea0407af562 X-Filterd-Recvd-Size: 11425 Received: from mga09.intel.com (mga09.intel.com [134.134.136.24]) by imf33.hostedemail.com (Postfix) with ESMTP for ; Mon, 13 Apr 2020 12:52:29 +0000 (UTC) IronPort-SDR: 3pHmK2sf+oD9urWiKKNqsTgnoYV6p1p5vkOl/VzGzyaXdcLFXEYDhiL0xCAET7+4NaIFxwYTzH 85H0yxNizTaQ== X-Amp-Result: SKIPPED(no attachment in message) X-Amp-File-Uploaded: False Received: from fmsmga007.fm.intel.com ([10.253.24.52]) by orsmga102.jf.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 13 Apr 2020 05:52:28 -0700 IronPort-SDR: 0wGN6UiggZoH8hQIHsCNUhXlwygtfdaiAN9tzmqvmAtGHGxKCsojsh3SHGtgVrqSthTenhaERq +4BiwxUX6Hhw== X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.72,378,1580803200"; d="scan'208";a="243529249" Received: from black.fi.intel.com ([10.237.72.28]) by fmsmga007.fm.intel.com with ESMTP; 13 Apr 2020 05:52:25 -0700 Received: by black.fi.intel.com (Postfix, from userid 1000) id 0A3872B6; Mon, 13 Apr 2020 15:52:22 +0300 (EEST) From: "Kirill A. Shutemov" To: akpm@linux-foundation.org, Andrea Arcangeli Cc: Zi Yan , Yang Shi , Ralph Campbell , John Hubbard , William Kucharski , linux-mm@kvack.org, linux-kernel@vger.kernel.org, "Kirill A. Shutemov" Subject: [PATCHv3, RESEND 7/8] thp: Change CoW semantics for anon-THP Date: Mon, 13 Apr 2020 15:52:19 +0300 Message-Id: <20200413125220.663-8-kirill.shutemov@linux.intel.com> X-Mailer: git-send-email 2.25.1 In-Reply-To: <20200413125220.663-1-kirill.shutemov@linux.intel.com> References: <20200413125220.663-1-kirill.shutemov@linux.intel.com> MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable X-Bogosity: Ham, tests=bogofilter, spamicity=0.000000, version=1.2.4 Sender: owner-linux-mm@kvack.org Precedence: bulk X-Loop: owner-majordomo@kvack.org List-ID: Currently we have different copy-on-write semantics for anon- and file-THP. For anon-THP we try to allocate huge page on the write fault, but on file-THP we split PMD and allocate 4k page. Arguably, file-THP semantics is more desirable: we don't necessary want to unshare full PMD range from the parent on the first access. This is the primary reason THP is unusable for some workloads, like Redis. The original THP refcounting didn't allow to have PTE-mapped compound pages, so we had no options, but to allocate huge page on CoW (with fallback to 512 4k pages). The current refcounting doesn't have such limitations and we can cut a lot of complex code out of fault path. khugepaged is now able to recover THP from such ranges if the configuration allows. Signed-off-by: Kirill A. Shutemov --- mm/huge_memory.c | 249 +++++------------------------------------------ 1 file changed, 25 insertions(+), 224 deletions(-) diff --git a/mm/huge_memory.c b/mm/huge_memory.c index 24ad53b4dfc0..25b84cc0f17d 100644 --- a/mm/huge_memory.c +++ b/mm/huge_memory.c @@ -1206,262 +1206,63 @@ void huge_pmd_set_accessed(struct vm_fault *vmf,= pmd_t orig_pmd) spin_unlock(vmf->ptl); } =20 -static vm_fault_t do_huge_pmd_wp_page_fallback(struct vm_fault *vmf, - pmd_t orig_pmd, struct page *page) -{ - struct vm_area_struct *vma =3D vmf->vma; - unsigned long haddr =3D vmf->address & HPAGE_PMD_MASK; - struct mem_cgroup *memcg; - pgtable_t pgtable; - pmd_t _pmd; - int i; - vm_fault_t ret =3D 0; - struct page **pages; - struct mmu_notifier_range range; - - pages =3D kmalloc_array(HPAGE_PMD_NR, sizeof(struct page *), - GFP_KERNEL); - if (unlikely(!pages)) { - ret |=3D VM_FAULT_OOM; - goto out; - } - - for (i =3D 0; i < HPAGE_PMD_NR; i++) { - pages[i] =3D alloc_page_vma_node(GFP_HIGHUSER_MOVABLE, vma, - vmf->address, page_to_nid(page)); - if (unlikely(!pages[i] || - mem_cgroup_try_charge_delay(pages[i], vma->vm_mm, - GFP_KERNEL, &memcg, false))) { - if (pages[i]) - put_page(pages[i]); - while (--i >=3D 0) { - memcg =3D (void *)page_private(pages[i]); - set_page_private(pages[i], 0); - mem_cgroup_cancel_charge(pages[i], memcg, - false); - put_page(pages[i]); - } - kfree(pages); - ret |=3D VM_FAULT_OOM; - goto out; - } - set_page_private(pages[i], (unsigned long)memcg); - } - - for (i =3D 0; i < HPAGE_PMD_NR; i++) { - copy_user_highpage(pages[i], page + i, - haddr + PAGE_SIZE * i, vma); - __SetPageUptodate(pages[i]); - cond_resched(); - } - - mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, vma, vma->vm_mm, - haddr, haddr + HPAGE_PMD_SIZE); - mmu_notifier_invalidate_range_start(&range); - - vmf->ptl =3D pmd_lock(vma->vm_mm, vmf->pmd); - if (unlikely(!pmd_same(*vmf->pmd, orig_pmd))) - goto out_free_pages; - VM_BUG_ON_PAGE(!PageHead(page), page); - - /* - * Leave pmd empty until pte is filled note we must notify here as - * concurrent CPU thread might write to new page before the call to - * mmu_notifier_invalidate_range_end() happens which can lead to a - * device seeing memory write in different order than CPU. - * - * See Documentation/vm/mmu_notifier.rst - */ - pmdp_huge_clear_flush_notify(vma, haddr, vmf->pmd); - - pgtable =3D pgtable_trans_huge_withdraw(vma->vm_mm, vmf->pmd); - pmd_populate(vma->vm_mm, &_pmd, pgtable); - - for (i =3D 0; i < HPAGE_PMD_NR; i++, haddr +=3D PAGE_SIZE) { - pte_t entry; - entry =3D mk_pte(pages[i], vma->vm_page_prot); - entry =3D maybe_mkwrite(pte_mkdirty(entry), vma); - memcg =3D (void *)page_private(pages[i]); - set_page_private(pages[i], 0); - page_add_new_anon_rmap(pages[i], vmf->vma, haddr, false); - mem_cgroup_commit_charge(pages[i], memcg, false, false); - lru_cache_add_active_or_unevictable(pages[i], vma); - vmf->pte =3D pte_offset_map(&_pmd, haddr); - VM_BUG_ON(!pte_none(*vmf->pte)); - set_pte_at(vma->vm_mm, haddr, vmf->pte, entry); - pte_unmap(vmf->pte); - } - kfree(pages); - - smp_wmb(); /* make pte visible before pmd */ - pmd_populate(vma->vm_mm, vmf->pmd, pgtable); - page_remove_rmap(page, true); - spin_unlock(vmf->ptl); - - /* - * No need to double call mmu_notifier->invalidate_range() callback as - * the above pmdp_huge_clear_flush_notify() did already call it. - */ - mmu_notifier_invalidate_range_only_end(&range); - - ret |=3D VM_FAULT_WRITE; - put_page(page); - -out: - return ret; - -out_free_pages: - spin_unlock(vmf->ptl); - mmu_notifier_invalidate_range_end(&range); - for (i =3D 0; i < HPAGE_PMD_NR; i++) { - memcg =3D (void *)page_private(pages[i]); - set_page_private(pages[i], 0); - mem_cgroup_cancel_charge(pages[i], memcg, false); - put_page(pages[i]); - } - kfree(pages); - goto out; -} - vm_fault_t do_huge_pmd_wp_page(struct vm_fault *vmf, pmd_t orig_pmd) { struct vm_area_struct *vma =3D vmf->vma; - struct page *page =3D NULL, *new_page; - struct mem_cgroup *memcg; + struct page *page; unsigned long haddr =3D vmf->address & HPAGE_PMD_MASK; - struct mmu_notifier_range range; - gfp_t huge_gfp; /* for allocation and charge */ - vm_fault_t ret =3D 0; =20 vmf->ptl =3D pmd_lockptr(vma->vm_mm, vmf->pmd); VM_BUG_ON_VMA(!vma->anon_vma, vma); + if (is_huge_zero_pmd(orig_pmd)) - goto alloc; + goto fallback; + spin_lock(vmf->ptl); - if (unlikely(!pmd_same(*vmf->pmd, orig_pmd))) - goto out_unlock; + + if (unlikely(!pmd_same(*vmf->pmd, orig_pmd))) { + spin_unlock(vmf->ptl); + return 0; + } =20 page =3D pmd_page(orig_pmd); VM_BUG_ON_PAGE(!PageCompound(page) || !PageHead(page), page); - /* - * We can only reuse the page if nobody else maps the huge page or it's - * part. - */ + + /* Lock page for reuse_swap_page() */ if (!trylock_page(page)) { get_page(page); spin_unlock(vmf->ptl); lock_page(page); spin_lock(vmf->ptl); if (unlikely(!pmd_same(*vmf->pmd, orig_pmd))) { + spin_unlock(vmf->ptl); unlock_page(page); put_page(page); - goto out_unlock; + return 0; } put_page(page); } + + /* + * We can only reuse the page if nobody else maps the huge page or it's + * part. + */ if (reuse_swap_page(page, NULL)) { pmd_t entry; entry =3D pmd_mkyoung(orig_pmd); entry =3D maybe_pmd_mkwrite(pmd_mkdirty(entry), vma); - if (pmdp_set_access_flags(vma, haddr, vmf->pmd, entry, 1)) + if (pmdp_set_access_flags(vma, haddr, vmf->pmd, entry, 1)) update_mmu_cache_pmd(vma, vmf->address, vmf->pmd); - ret |=3D VM_FAULT_WRITE; unlock_page(page); - goto out_unlock; - } - unlock_page(page); - get_page(page); - spin_unlock(vmf->ptl); -alloc: - if (__transparent_hugepage_enabled(vma) && - !transparent_hugepage_debug_cow()) { - huge_gfp =3D alloc_hugepage_direct_gfpmask(vma); - new_page =3D alloc_hugepage_vma(huge_gfp, vma, haddr, HPAGE_PMD_ORDER)= ; - } else - new_page =3D NULL; - - if (likely(new_page)) { - prep_transhuge_page(new_page); - } else { - if (!page) { - split_huge_pmd(vma, vmf->pmd, vmf->address); - ret |=3D VM_FAULT_FALLBACK; - } else { - ret =3D do_huge_pmd_wp_page_fallback(vmf, orig_pmd, page); - if (ret & VM_FAULT_OOM) { - split_huge_pmd(vma, vmf->pmd, vmf->address); - ret |=3D VM_FAULT_FALLBACK; - } - put_page(page); - } - count_vm_event(THP_FAULT_FALLBACK); - goto out; - } - - if (unlikely(mem_cgroup_try_charge_delay(new_page, vma->vm_mm, - huge_gfp, &memcg, true))) { - put_page(new_page); - split_huge_pmd(vma, vmf->pmd, vmf->address); - if (page) - put_page(page); - ret |=3D VM_FAULT_FALLBACK; - count_vm_event(THP_FAULT_FALLBACK); - goto out; - } - - count_vm_event(THP_FAULT_ALLOC); - count_memcg_events(memcg, THP_FAULT_ALLOC, 1); - - if (!page) - clear_huge_page(new_page, vmf->address, HPAGE_PMD_NR); - else - copy_user_huge_page(new_page, page, vmf->address, - vma, HPAGE_PMD_NR); - __SetPageUptodate(new_page); - - mmu_notifier_range_init(&range, MMU_NOTIFY_CLEAR, 0, vma, vma->vm_mm, - haddr, haddr + HPAGE_PMD_SIZE); - mmu_notifier_invalidate_range_start(&range); - - spin_lock(vmf->ptl); - if (page) - put_page(page); - if (unlikely(!pmd_same(*vmf->pmd, orig_pmd))) { spin_unlock(vmf->ptl); - mem_cgroup_cancel_charge(new_page, memcg, true); - put_page(new_page); - goto out_mn; - } else { - pmd_t entry; - entry =3D mk_huge_pmd(new_page, vma->vm_page_prot); - entry =3D maybe_pmd_mkwrite(pmd_mkdirty(entry), vma); - pmdp_huge_clear_flush_notify(vma, haddr, vmf->pmd); - page_add_new_anon_rmap(new_page, vma, haddr, true); - mem_cgroup_commit_charge(new_page, memcg, false, true); - lru_cache_add_active_or_unevictable(new_page, vma); - set_pmd_at(vma->vm_mm, haddr, vmf->pmd, entry); - update_mmu_cache_pmd(vma, vmf->address, vmf->pmd); - if (!page) { - add_mm_counter(vma->vm_mm, MM_ANONPAGES, HPAGE_PMD_NR); - } else { - VM_BUG_ON_PAGE(!PageHead(page), page); - page_remove_rmap(page, true); - put_page(page); - } - ret |=3D VM_FAULT_WRITE; + return VM_FAULT_WRITE; } + + unlock_page(page); spin_unlock(vmf->ptl); -out_mn: - /* - * No need to double call mmu_notifier->invalidate_range() callback as - * the above pmdp_huge_clear_flush_notify() did already call it. - */ - mmu_notifier_invalidate_range_only_end(&range); -out: - return ret; -out_unlock: - spin_unlock(vmf->ptl); - return ret; +fallback: + __split_huge_pmd(vma, vmf->pmd, vmf->address, false, NULL); + return VM_FAULT_FALLBACK; } =20 /* --=20 2.26.0