From patchwork Thu Aug 10 10:33:29 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Ryan Roberts X-Patchwork-Id: 133883 Return-Path: Delivered-To: ouuuleilei@gmail.com Received: by 2002:a59:b824:0:b0:3f2:4152:657d with SMTP id z4csp338086vqi; Thu, 10 Aug 2023 04:08:19 -0700 (PDT) X-Google-Smtp-Source: AGHT+IE0se9ekG9ey0YRh/rd1BxihLwPMfM+NOCgm5LopeQ6rpSySYQe1e9YTcKoF+PnfFkN/Nmq X-Received: by 2002:a54:4e94:0:b0:3a3:76c6:a46f with SMTP id c20-20020a544e94000000b003a376c6a46fmr2198235oiy.38.1691665699241; Thu, 10 Aug 2023 04:08:19 -0700 (PDT) ARC-Seal: i=1; a=rsa-sha256; t=1691665699; cv=none; d=google.com; s=arc-20160816; b=VEWnTEF2qTJBbpj+rlj+fiOIzTVsGzRBKpL+M3qsj8VCCQ/vfH+F5KukfL3dW10yLs SUSBsJNCh9n1h15toY0083JUXZoZ/RYZ3r2DH5gD4RgjW68YFPe1yburk+SZQ8WtooS3 3fHJxm0hP4Ul9FdkKLgW+dSgSc/b2+2nQuUfaJQmArgWbJZYrmXWnANlAtPWuIMRcFIS K+tUpVMD/wIc6Dum3gP6HG4X5xAgAyZqjjCzzr4DXEQ5M0ZiBf/untrOf22ycPo0H3wQ 310i5hntoTNkDjh3eQ4CmhAPUJGZPFtfYegpY1gUhEVhLFOCqjOM1qZvAgT5YlvZ9Boj QzCQ== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=list-id:precedence:content-transfer-encoding:mime-version :references:in-reply-to:message-id:date:subject:cc:to:from; bh=LuWe/KwZnwtyrwm231Z4pl36eOk0QjHBRP62MyaAfj0=; fh=y3Sw7VG+9eLsYabwWUWsSnvNS303XOkFtMtRzGTCOfU=; b=QVtnRG91iWvUY0uwIgJVX50Tyz9/64iyGwk76Gb0/rgMZwlCL2QCdXk4IE6yQmet5Q w2Hy4HitCjgZXnMs2Xu3HC7vliMyTndaDhjEd0/AUOnptRe5KnVsBYL8s61WOhug5qPz 2scVOyxLvfx+UUWbl54lEFtb0JiNjMQ+TLbey0M1nJH8vAvQ2fw8X+dHfr/3mjkS9Kcd kzE5XTFoNLu4wtiDYJJ8z3XzLL6VcWkYFwKB7NddkXkw0YmPIgzWyuIo7oa56EqB/W6y 0TP1Q5gqqdzx4T/S7P8tRBTTUkYv9nvNUVU5wUdUQDff4s6R+ZNZCoCOvNn2/ta8gxfF UZzg== ARC-Authentication-Results: i=1; mx.google.com; spf=pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) smtp.mailfrom=linux-kernel-owner@vger.kernel.org; dmarc=fail (p=NONE sp=NONE dis=NONE) header.from=arm.com Received: from out1.vger.email (out1.vger.email. [2620:137:e000::1:20]) by mx.google.com with ESMTP id v10-20020a63b64a000000b00564730be155si1351730pgt.6.2023.08.10.04.08.04; Thu, 10 Aug 2023 04:08:19 -0700 (PDT) Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20; Authentication-Results: mx.google.com; spf=pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) smtp.mailfrom=linux-kernel-owner@vger.kernel.org; dmarc=fail (p=NONE sp=NONE dis=NONE) header.from=arm.com Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S233844AbjHJKdy (ORCPT + 99 others); Thu, 10 Aug 2023 06:33:54 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:58844 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S229890AbjHJKdu (ORCPT ); Thu, 10 Aug 2023 06:33:50 -0400 Received: from foss.arm.com (foss.arm.com [217.140.110.172]) by lindbergh.monkeyblade.net (Postfix) with ESMTP id 0AD1AE4D for ; Thu, 10 Aug 2023 03:33:49 -0700 (PDT) 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 1EA50113E; Thu, 10 Aug 2023 03:34:31 -0700 (PDT) Received: from e125769.cambridge.arm.com (e125769.cambridge.arm.com [10.1.196.26]) by usa-sjc-imap-foss1.foss.arm.com (Postfix) with ESMTPSA id 396243F6C4; Thu, 10 Aug 2023 03:33:46 -0700 (PDT) From: Ryan Roberts To: Will Deacon , "Aneesh Kumar K.V" , Andrew Morton , Nick Piggin , Peter Zijlstra , Christian Borntraeger , Sven Schnelle , Arnd Bergmann , "Matthew Wilcox (Oracle)" , David Hildenbrand , Yu Zhao , "Kirill A. Shutemov" , Yin Fengwei , Yang Shi , "Huang, Ying" , Zi Yan Cc: Ryan Roberts , linux-mm@kvack.org, linux-kernel@vger.kernel.org Subject: [PATCH v1 1/4] mm: Implement folio_remove_rmap_range() Date: Thu, 10 Aug 2023 11:33:29 +0100 Message-Id: <20230810103332.3062143-2-ryan.roberts@arm.com> X-Mailer: git-send-email 2.25.1 In-Reply-To: <20230810103332.3062143-1-ryan.roberts@arm.com> References: <20230810103332.3062143-1-ryan.roberts@arm.com> MIME-Version: 1.0 X-Spam-Status: No, score=-1.9 required=5.0 tests=BAYES_00, RCVD_IN_DNSWL_BLOCKED,SPF_HELO_NONE,SPF_NONE autolearn=ham autolearn_force=no version=3.4.6 X-Spam-Checker-Version: SpamAssassin 3.4.6 (2021-04-09) on lindbergh.monkeyblade.net Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org X-getmail-retrieved-from-mailbox: INBOX X-GMAIL-THRID: 1773840051990733849 X-GMAIL-MSGID: 1773840051990733849 Like page_remove_rmap() but batch-removes the rmap for a range of pages belonging to a folio. This can provide a small speedup due to less manipuation of the various counters. But more crucially, if removing the rmap for all pages of a folio in a batch, there is no need to (spuriously) add it to the deferred split list, which saves significant cost when there is contention for the split queue lock. All contained pages are accounted using the order-0 folio (or base page) scheme. page_remove_rmap() is refactored so that it forwards to folio_remove_rmap_range() for !compound cases, and both functions now share a common epilogue function. The intention here is to avoid duplication of code. Signed-off-by: Ryan Roberts --- include/linux/rmap.h | 2 + mm/rmap.c | 125 ++++++++++++++++++++++++++++++++----------- 2 files changed, 97 insertions(+), 30 deletions(-) diff --git a/include/linux/rmap.h b/include/linux/rmap.h index a3825ce81102..d442d1e5425d 100644 --- a/include/linux/rmap.h +++ b/include/linux/rmap.h @@ -202,6 +202,8 @@ void folio_add_file_rmap_range(struct folio *, struct page *, unsigned int nr, struct vm_area_struct *, bool compound); void page_remove_rmap(struct page *, struct vm_area_struct *, bool compound); +void folio_remove_rmap_range(struct folio *folio, struct page *page, + int nr, struct vm_area_struct *vma); void hugepage_add_anon_rmap(struct page *, struct vm_area_struct *, unsigned long address, rmap_t flags); diff --git a/mm/rmap.c b/mm/rmap.c index 1f04debdc87a..d82d52ebf3a6 100644 --- a/mm/rmap.c +++ b/mm/rmap.c @@ -1379,6 +1379,94 @@ void page_add_file_rmap(struct page *page, struct vm_area_struct *vma, folio_add_file_rmap_range(folio, page, nr_pages, vma, compound); } +/** + * __remove_rmap_finish - common operations when taking down a mapping. + * @folio: Folio containing all pages taken down. + * @vma: The VM area containing the range. + * @compound: True if pages were taken down from PMD or false if from PTE(s). + * @nr_unmapped: Number of pages within folio that are now unmapped. + * @nr_mapped: Number of pages within folio that are still mapped. + */ +static void __remove_rmap_finish(struct folio *folio, + struct vm_area_struct *vma, bool compound, + int nr_unmapped, int nr_mapped) +{ + enum node_stat_item idx; + + if (nr_unmapped) { + idx = folio_test_anon(folio) ? NR_ANON_MAPPED : NR_FILE_MAPPED; + __lruvec_stat_mod_folio(folio, idx, -nr_unmapped); + + /* + * Queue large anon folio for deferred split if at least one + * page of the folio is unmapped and at least one page is still + * mapped. + */ + if (folio_test_large(folio) && + folio_test_anon(folio) && nr_mapped) + deferred_split_folio(folio); + } + + /* + * It would be tidy to reset folio_test_anon mapping when fully + * unmapped, but that might overwrite a racing page_add_anon_rmap + * which increments mapcount after us but sets mapping before us: + * so leave the reset to free_pages_prepare, and remember that + * it's only reliable while mapped. + */ + + munlock_vma_folio(folio, vma, compound); +} + +/** + * folio_remove_rmap_range - Take down PTE mappings from a range of pages. + * @folio: Folio containing all pages in range. + * @page: First page in range to unmap. + * @nr: Number of pages to unmap. + * @vma: The VM area containing the range. + * + * All pages in the range must belong to the same VMA & folio. They must be + * mapped with PTEs, not a PMD. + * + * Context: Caller holds the pte lock. + */ +void folio_remove_rmap_range(struct folio *folio, struct page *page, + int nr, struct vm_area_struct *vma) +{ + atomic_t *mapped = &folio->_nr_pages_mapped; + int nr_unmapped = 0; + int nr_mapped = 0; + bool last; + + if (unlikely(folio_test_hugetlb(folio))) { + VM_WARN_ON_FOLIO(1, folio); + return; + } + + VM_WARN_ON_ONCE(page < &folio->page || + page + nr > (&folio->page + folio_nr_pages(folio))); + + if (!folio_test_large(folio)) { + /* Is this the page's last map to be removed? */ + last = atomic_add_negative(-1, &page->_mapcount); + nr_unmapped = last; + } else { + for (; nr != 0; nr--, page++) { + /* Is this the page's last map to be removed? */ + last = atomic_add_negative(-1, &page->_mapcount); + if (last) + nr_unmapped++; + } + + /* Pages still mapped if folio mapped entirely */ + nr_mapped = atomic_sub_return_relaxed(nr_unmapped, mapped); + if (nr_mapped >= COMPOUND_MAPPED) + nr_unmapped = 0; + } + + __remove_rmap_finish(folio, vma, false, nr_unmapped, nr_mapped); +} + /** * page_remove_rmap - take down pte mapping from a page * @page: page to remove mapping from @@ -1405,15 +1493,13 @@ void page_remove_rmap(struct page *page, struct vm_area_struct *vma, return; } - /* Is page being unmapped by PTE? Is this its last map to be removed? */ + /* Is page being unmapped by PTE? */ if (likely(!compound)) { - last = atomic_add_negative(-1, &page->_mapcount); - nr = last; - if (last && folio_test_large(folio)) { - nr = atomic_dec_return_relaxed(mapped); - nr = (nr < COMPOUND_MAPPED); - } - } else if (folio_test_pmd_mappable(folio)) { + folio_remove_rmap_range(folio, page, 1, vma); + return; + } + + if (folio_test_pmd_mappable(folio)) { /* That test is redundant: it's for safety or to optimize out */ last = atomic_add_negative(-1, &folio->_entire_mapcount); @@ -1441,29 +1527,8 @@ void page_remove_rmap(struct page *page, struct vm_area_struct *vma, idx = NR_FILE_PMDMAPPED; __lruvec_stat_mod_folio(folio, idx, -nr_pmdmapped); } - if (nr) { - idx = folio_test_anon(folio) ? NR_ANON_MAPPED : NR_FILE_MAPPED; - __lruvec_stat_mod_folio(folio, idx, -nr); - - /* - * Queue anon THP for deferred split if at least one - * page of the folio is unmapped and at least one page - * is still mapped. - */ - if (folio_test_pmd_mappable(folio) && folio_test_anon(folio)) - if (!compound || nr < nr_pmdmapped) - deferred_split_folio(folio); - } - - /* - * It would be tidy to reset folio_test_anon mapping when fully - * unmapped, but that might overwrite a racing page_add_anon_rmap - * which increments mapcount after us but sets mapping before us: - * so leave the reset to free_pages_prepare, and remember that - * it's only reliable while mapped. - */ - munlock_vma_folio(folio, vma, compound); + __remove_rmap_finish(folio, vma, compound, nr, nr_pmdmapped - nr); } /* From patchwork Thu Aug 10 10:33:30 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Ryan Roberts X-Patchwork-Id: 133925 Return-Path: Delivered-To: ouuuleilei@gmail.com Received: by 2002:a59:b824:0:b0:3f2:4152:657d with SMTP id z4csp348057vqi; Thu, 10 Aug 2023 04:26:56 -0700 (PDT) X-Google-Smtp-Source: AGHT+IFsSbI4sbYn1uqaR0ClQOqAs+f062iHIHSoVwbSXed1om4xj5SkTylOVeM41fMStX5/NRft X-Received: by 2002:a05:6a20:7f84:b0:140:61f8:53f6 with SMTP id d4-20020a056a207f8400b0014061f853f6mr2081557pzj.29.1691666816495; Thu, 10 Aug 2023 04:26:56 -0700 (PDT) ARC-Seal: i=1; a=rsa-sha256; t=1691666816; cv=none; d=google.com; s=arc-20160816; b=UF6N3ETeqxvAwH1SHoPD3afoe8kSaC9qZ3SpUFP330MbU6M/WxqbjdFLkYBkwP0tLB 7iUTSoHyjwRamYvqYTOB40vfHPuYxle2MssXRAm2Gj/GGJSpa7q4FMhee6Y/U37OZwQe 8v4BI0g8Jwmm+iTzFQnI7/D9mNmycjl6tzERcERkFb1inw//5EgiiWmlG45nale7rY2w E6n7e4VKbuAowfWC/zGx2lirGKX7iu85w5DKIAtP87L+DvtQjS5wG/3RbkF0E99huzaR Dna8owq+/wkFEJfLzEFGIwSGsmmX2DpiK/AjqaOE/F0N6RIf5o67Otjlo5Ca3xz7zeaC boBQ== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=list-id:precedence:content-transfer-encoding:mime-version :references:in-reply-to:message-id:date:subject:cc:to:from; bh=1bKY+YcGJdzrroBDyOKPdXYYkkK61aMopDGMBl3xXFI=; fh=y3Sw7VG+9eLsYabwWUWsSnvNS303XOkFtMtRzGTCOfU=; b=pUmIopaW10UvSNZnNEdYUE3/PPaueTloLbct79v42qjr/2qs62xHU+jG7iuqFjiXHJ idnefLSAvF13Lb0Mgo6RDJDtSM80rPICJh02tcY3+87WQOY6+jWZQbx4GS9PxWtwTCC/ hhqgbTekmJZebDLBTyxeUQeOqlAP6BDVqz0wqroWMyof5iJjMjAlkFTKQm9+csnFPzK4 7VPAoksjuGc+6RxKyrjkCGB7n+rLzq7ePeCTqolfeiYad5D07510xKP/JHyVsife1oaq hMpilFL+U6LiQU9ZXWJmbB2qOaNa82K8lH2hhUDyxbFORPw+yD3yAwc8d1KoEKODITit VOlg== ARC-Authentication-Results: i=1; mx.google.com; spf=pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) smtp.mailfrom=linux-kernel-owner@vger.kernel.org; dmarc=fail (p=NONE sp=NONE dis=NONE) header.from=arm.com Received: from out1.vger.email (out1.vger.email. [2620:137:e000::1:20]) by mx.google.com with ESMTP id bv68-20020a632e47000000b00563ee39a642si1388281pgb.308.2023.08.10.04.26.42; Thu, 10 Aug 2023 04:26:56 -0700 (PDT) Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20; Authentication-Results: mx.google.com; spf=pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) smtp.mailfrom=linux-kernel-owner@vger.kernel.org; dmarc=fail (p=NONE sp=NONE dis=NONE) header.from=arm.com Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S234938AbjHJKd4 (ORCPT + 99 others); Thu, 10 Aug 2023 06:33:56 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:58886 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S234728AbjHJKdx (ORCPT ); Thu, 10 Aug 2023 06:33:53 -0400 Received: from foss.arm.com (foss.arm.com [217.140.110.172]) by lindbergh.monkeyblade.net (Postfix) with ESMTP id EF87110DE for ; Thu, 10 Aug 2023 03:33:51 -0700 (PDT) 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 1237C11FB; Thu, 10 Aug 2023 03:34:34 -0700 (PDT) Received: from e125769.cambridge.arm.com (e125769.cambridge.arm.com [10.1.196.26]) by usa-sjc-imap-foss1.foss.arm.com (Postfix) with ESMTPSA id 0EA363F6C4; Thu, 10 Aug 2023 03:33:48 -0700 (PDT) From: Ryan Roberts To: Will Deacon , "Aneesh Kumar K.V" , Andrew Morton , Nick Piggin , Peter Zijlstra , Christian Borntraeger , Sven Schnelle , Arnd Bergmann , "Matthew Wilcox (Oracle)" , David Hildenbrand , Yu Zhao , "Kirill A. Shutemov" , Yin Fengwei , Yang Shi , "Huang, Ying" , Zi Yan Cc: Ryan Roberts , linux-mm@kvack.org, linux-kernel@vger.kernel.org Subject: [PATCH v1 2/4] mm/mmu_gather: generalize mmu_gather rmap removal mechanism Date: Thu, 10 Aug 2023 11:33:30 +0100 Message-Id: <20230810103332.3062143-3-ryan.roberts@arm.com> X-Mailer: git-send-email 2.25.1 In-Reply-To: <20230810103332.3062143-1-ryan.roberts@arm.com> References: <20230810103332.3062143-1-ryan.roberts@arm.com> MIME-Version: 1.0 X-Spam-Status: No, score=-1.9 required=5.0 tests=BAYES_00, RCVD_IN_DNSWL_BLOCKED,SPF_HELO_NONE,SPF_NONE autolearn=ham autolearn_force=no version=3.4.6 X-Spam-Checker-Version: SpamAssassin 3.4.6 (2021-04-09) on lindbergh.monkeyblade.net Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org X-getmail-retrieved-from-mailbox: INBOX X-GMAIL-THRID: 1773841223479415817 X-GMAIL-MSGID: 1773841223479415817 Commit 5df397dec7c4 ("mm: delay page_remove_rmap() until after the TLB has been flushed") added a mechanism whereby pages added to the mmu_gather buffer could indicate whether they should also be removed from the rmap. Then a call to the new tlb_flush_rmaps() API would iterate though the buffer and remove each flagged page from the rmap. This mechanism was intended for use with !PageAnon(page) pages only. Let's generalize this rmap removal mechanism so that any type of page can be removed from the rmap. This is done as preparation for batching rmap removals with folio_remove_rmap_range(), whereby we will pass a contiguous range of pages belonging to the same folio to be removed in one shot for a performance improvement. The mmu_gather now maintains a "pointer" that points to batch and index within that batch of the next page in the queue that is yet to be removed from the rmap. tlb_discard_rmaps() resets this "pointer" to the first empty location in the queue. Whenever tlb_flush_rmaps() is called, every page from "pointer" to the end of the queue is removed from the rmap. Once the mmu is flushed (tlb_flush_mmu()/tlb_finish_mmu()) any pending rmap removals are discarded. This pointer mechanism ensures that tlb_flush_rmaps() only has to walk the part of the queue for which rmap removal is pending, avoids the (potentially large) early portion of the queue for which rmap removal has already been performed but for which tlb invalidation/page freeing is still pending. tlb_flush_rmaps() must always be called under the same PTL as was used to clear the corresponding PTEs. So in practice rmap removal will be done in a batch for each PTE table, while the tlbi/freeing can continue to be done in much bigger batches outside the PTL. See this example flow: tlb_gather_mmu() for each pte table { with ptl held { for each pte { tlb_remove_tlb_entry() __tlb_remove_page() } if (any removed pages require rmap after tlbi) tlb_flush_mmu_tlbonly() tlb_flush_rmaps() } if (full) tlb_flush_mmu() } tlb_finish_mmu() So this more general mechanism is no longer just for delaying rmap removal until after tlbi, but can be used that way when required. Note that s390 does not gather pages, but does immediate tlbi and page freeing. In this case we continue to do the rmap removal page-by-page without gathering them in the mmu_gather. Signed-off-by: Ryan Roberts --- include/asm-generic/tlb.h | 34 ++++++++++++------------ mm/memory.c | 24 ++++++++++------- mm/mmu_gather.c | 55 +++++++++++++++++++++++---------------- 3 files changed, 66 insertions(+), 47 deletions(-) diff --git a/include/asm-generic/tlb.h b/include/asm-generic/tlb.h index 129a3a759976..f339d68cf44f 100644 --- a/include/asm-generic/tlb.h +++ b/include/asm-generic/tlb.h @@ -266,25 +266,30 @@ extern bool __tlb_remove_page_size(struct mmu_gather *tlb, #ifdef CONFIG_SMP /* - * This both sets 'delayed_rmap', and returns true. It would be an inline - * function, except we define it before the 'struct mmu_gather'. + * For configurations that support batching the rmap removal, the removal is + * triggered by calling tlb_flush_rmaps(), which must be called after the pte(s) + * are cleared and the page has been added to the mmu_gather, and before the ptl + * lock that was held for clearing the pte is released. */ -#define tlb_delay_rmap(tlb) (((tlb)->delayed_rmap = 1), true) +#define tlb_batch_rmap(tlb) (true) extern void tlb_flush_rmaps(struct mmu_gather *tlb, struct vm_area_struct *vma); +extern void tlb_discard_rmaps(struct mmu_gather *tlb); #endif #endif /* - * We have a no-op version of the rmap removal that doesn't - * delay anything. That is used on S390, which flushes remote - * TLBs synchronously, and on UP, which doesn't have any - * remote TLBs to flush and is not preemptible due to this - * all happening under the page table lock. + * We have a no-op version of the rmap removal that doesn't do anything. That is + * used on S390, which flushes remote TLBs synchronously, and on UP, which + * doesn't have any remote TLBs to flush and is not preemptible due to this all + * happening under the page table lock. Here, the caller must manage each rmap + * removal separately. */ -#ifndef tlb_delay_rmap -#define tlb_delay_rmap(tlb) (false) -static inline void tlb_flush_rmaps(struct mmu_gather *tlb, struct vm_area_struct *vma) { } +#ifndef tlb_batch_rmap +#define tlb_batch_rmap(tlb) (false) +static inline void tlb_flush_rmaps(struct mmu_gather *tlb, + struct vm_area_struct *vma) { } +static inline void tlb_discard_rmaps(struct mmu_gather *tlb) { } #endif /* @@ -317,11 +322,6 @@ struct mmu_gather { */ unsigned int freed_tables : 1; - /* - * Do we have pending delayed rmap removals? - */ - unsigned int delayed_rmap : 1; - /* * at which levels have we cleared entries? */ @@ -343,6 +343,8 @@ struct mmu_gather { struct mmu_gather_batch *active; struct mmu_gather_batch local; struct page *__pages[MMU_GATHER_BUNDLE]; + struct mmu_gather_batch *rmap_pend; + unsigned int rmap_pend_first; #ifdef CONFIG_MMU_GATHER_PAGE_SIZE unsigned int page_size; diff --git a/mm/memory.c b/mm/memory.c index d003076b218d..94a6ebd409a6 100644 --- a/mm/memory.c +++ b/mm/memory.c @@ -1405,6 +1405,7 @@ static unsigned long zap_pte_range(struct mmu_gather *tlb, swp_entry_t entry; tlb_change_page_size(tlb, PAGE_SIZE); + tlb_discard_rmaps(tlb); init_rss_vec(rss); start_pte = pte = pte_offset_map_lock(mm, pmd, addr, &ptl); if (!pte) @@ -1423,7 +1424,7 @@ static unsigned long zap_pte_range(struct mmu_gather *tlb, break; if (pte_present(ptent)) { - unsigned int delay_rmap; + unsigned int batch_rmap; page = vm_normal_page(vma, addr, ptent); if (unlikely(!should_zap_page(details, page))) @@ -1438,12 +1439,15 @@ static unsigned long zap_pte_range(struct mmu_gather *tlb, continue; } - delay_rmap = 0; + batch_rmap = tlb_batch_rmap(tlb); if (!PageAnon(page)) { if (pte_dirty(ptent)) { set_page_dirty(page); - if (tlb_delay_rmap(tlb)) { - delay_rmap = 1; + if (batch_rmap) { + /* + * Ensure tlb flush happens + * before rmap remove. + */ force_flush = 1; } } @@ -1451,12 +1455,12 @@ static unsigned long zap_pte_range(struct mmu_gather *tlb, mark_page_accessed(page); } rss[mm_counter(page)]--; - if (!delay_rmap) { + if (!batch_rmap) { page_remove_rmap(page, vma, false); if (unlikely(page_mapcount(page) < 0)) print_bad_pte(vma, addr, ptent, page); } - if (unlikely(__tlb_remove_page(tlb, page, delay_rmap))) { + if (unlikely(__tlb_remove_page(tlb, page, 0))) { force_flush = 1; addr += PAGE_SIZE; break; @@ -1517,10 +1521,12 @@ static unsigned long zap_pte_range(struct mmu_gather *tlb, arch_leave_lazy_mmu_mode(); /* Do the actual TLB flush before dropping ptl */ - if (force_flush) { + if (force_flush) tlb_flush_mmu_tlbonly(tlb); - tlb_flush_rmaps(tlb, vma); - } + + /* Rmap removal must always happen before dropping ptl */ + tlb_flush_rmaps(tlb, vma); + pte_unmap_unlock(start_pte, ptl); /* diff --git a/mm/mmu_gather.c b/mm/mmu_gather.c index ea9683e12936..ca328ecef5c2 100644 --- a/mm/mmu_gather.c +++ b/mm/mmu_gather.c @@ -19,10 +19,6 @@ static bool tlb_next_batch(struct mmu_gather *tlb) { struct mmu_gather_batch *batch; - /* Limit batching if we have delayed rmaps pending */ - if (tlb->delayed_rmap && tlb->active != &tlb->local) - return false; - batch = tlb->active; if (batch->next) { tlb->active = batch->next; @@ -48,36 +44,49 @@ static bool tlb_next_batch(struct mmu_gather *tlb) } #ifdef CONFIG_SMP -static void tlb_flush_rmap_batch(struct mmu_gather_batch *batch, struct vm_area_struct *vma) +static void tlb_flush_rmap_batch(struct mmu_gather_batch *batch, + unsigned int first, + struct vm_area_struct *vma) { - for (int i = 0; i < batch->nr; i++) { + for (int i = first; i < batch->nr; i++) { struct encoded_page *enc = batch->encoded_pages[i]; + struct page *page = encoded_page_ptr(enc); - if (encoded_page_flags(enc)) { - struct page *page = encoded_page_ptr(enc); - page_remove_rmap(page, vma, false); - } + page_remove_rmap(page, vma, false); } } /** - * tlb_flush_rmaps - do pending rmap removals after we have flushed the TLB + * tlb_flush_rmaps - do pending rmap removals * @tlb: the current mmu_gather + * @vma: vm area from which all pages are removed * - * Note that because of how tlb_next_batch() above works, we will - * never start multiple new batches with pending delayed rmaps, so - * we only need to walk through the current active batch and the - * original local one. + * Removes rmap from all pages added via (e.g.) __tlb_remove_page_size() since + * the last call to tlb_discard_rmaps() or tlb_flush_rmaps(). All of those pages + * must have been mapped by vma. Must be called after the pte(s) are cleared, + * and before the ptl lock that was held for clearing the pte is released. Pages + * are accounted using the order-0 folio (or base page) scheme. */ void tlb_flush_rmaps(struct mmu_gather *tlb, struct vm_area_struct *vma) { - if (!tlb->delayed_rmap) - return; + struct mmu_gather_batch *batch = tlb->rmap_pend; - tlb_flush_rmap_batch(&tlb->local, vma); - if (tlb->active != &tlb->local) - tlb_flush_rmap_batch(tlb->active, vma); - tlb->delayed_rmap = 0; + tlb_flush_rmap_batch(batch, tlb->rmap_pend_first, vma); + + for (batch = batch->next; batch && batch->nr; batch = batch->next) + tlb_flush_rmap_batch(batch, 0, vma); + + tlb_discard_rmaps(tlb); +} + +/** + * tlb_discard_rmaps - discard any pending rmap removals + * @tlb: the current mmu_gather + */ +void tlb_discard_rmaps(struct mmu_gather *tlb) +{ + tlb->rmap_pend = tlb->active; + tlb->rmap_pend_first = tlb->active->nr; } #endif @@ -102,6 +111,7 @@ static void tlb_batch_pages_flush(struct mmu_gather *tlb) } while (batch->nr); } tlb->active = &tlb->local; + tlb_discard_rmaps(tlb); } static void tlb_batch_list_free(struct mmu_gather *tlb) @@ -312,8 +322,9 @@ static void __tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, tlb->local.max = ARRAY_SIZE(tlb->__pages); tlb->active = &tlb->local; tlb->batch_count = 0; + tlb->rmap_pend = &tlb->local; + tlb->rmap_pend_first = 0; #endif - tlb->delayed_rmap = 0; tlb_table_init(tlb); #ifdef CONFIG_MMU_GATHER_PAGE_SIZE From patchwork Thu Aug 10 10:33:31 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Ryan Roberts X-Patchwork-Id: 133890 Return-Path: Delivered-To: ouuuleilei@gmail.com Received: by 2002:a59:b824:0:b0:3f2:4152:657d with SMTP id z4csp341398vqi; Thu, 10 Aug 2023 04:13:56 -0700 (PDT) X-Google-Smtp-Source: AGHT+IEgAHs6jKLrShGUqi0kDHFOUaeKszf8j+zl7VRE7jseohQyEs8ud0lFF6bCq6YNu92caXjF X-Received: by 2002:a17:902:b187:b0:1bb:a8f7:a4bb with SMTP id s7-20020a170902b18700b001bba8f7a4bbmr1769654plr.5.1691666035797; Thu, 10 Aug 2023 04:13:55 -0700 (PDT) ARC-Seal: i=1; a=rsa-sha256; t=1691666035; cv=none; d=google.com; s=arc-20160816; b=Vxa7EZ3J38zB4c30d2B1gtAyscHCXbtBVk3SwME9HiltuuL/bgkwcG+0Swdiy290zO XukWLjYTY5/DWT5vyvKKFXGrVPi3goJ9AM9vIZzGLwk9iQAIfM4LLTzUvlTLHQq76RKA AZAttcRmBSTU08UQ5YDx15gt8llV8sWyfF3ducy1EDrz8SvP4kOLjxzhbm3TLPc5ATXY m2wgQrnMYfPmH/V1bd0hN7ramI9vB58wbaqb+GycSQOubX7+rJqUhH6DbEJvZkecLPTO E1IT6v1rYcI2dvIwAiajodBVC0UG/6MnyOC8onQ6XNr+NxiUoJaVWHurHG3ScM/fSJn6 k/Xg== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=list-id:precedence:content-transfer-encoding:mime-version :references:in-reply-to:message-id:date:subject:cc:to:from; bh=asXfQ6ze4/2x2LKEtV7V8p75FuuLzfR6PdeJcLg+esQ=; fh=y3Sw7VG+9eLsYabwWUWsSnvNS303XOkFtMtRzGTCOfU=; b=PSk/PrtBYs0EOOVYg0Db2KEwGoAqLxG2p0lOvSADxnfNjI6Z++su7OOdY7fyRbYJxj /jbj+fWl8dtav3PXj7AyD76/H4nOviUiOWpGpm0cAwfAFvcTOZdG3g+0PqfmLGuW0V32 qKxaF8zMo3i26eNo8rizVheSwkAG3gCS6MUJ46x0rwt6w6e16kz4ZiVRWrwBI0FJbxkb U19URHNGLeop2FbdWfn5iNs5rSo1b1n6XpXTTaYKUkjtFBQb5Hfv/XnLhIRfSP4AEhKL NUZqihjC1zkGvuQRWBuF4dTN6tC6eh/diy1KPK2XECIRUKNGLSVHQH4gd2BXX+lkFmEH uyjw== ARC-Authentication-Results: i=1; mx.google.com; spf=pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) smtp.mailfrom=linux-kernel-owner@vger.kernel.org; dmarc=fail (p=NONE sp=NONE dis=NONE) header.from=arm.com Received: from out1.vger.email (out1.vger.email. [2620:137:e000::1:20]) by mx.google.com with ESMTP id p1-20020a170902e74100b001bbaaf979a4si1322708plf.360.2023.08.10.04.13.42; Thu, 10 Aug 2023 04:13:55 -0700 (PDT) Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20; Authentication-Results: mx.google.com; spf=pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) smtp.mailfrom=linux-kernel-owner@vger.kernel.org; dmarc=fail (p=NONE sp=NONE dis=NONE) header.from=arm.com Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S235090AbjHJKeG (ORCPT + 99 others); Thu, 10 Aug 2023 06:34:06 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:58884 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S234945AbjHJKd4 (ORCPT ); Thu, 10 Aug 2023 06:33:56 -0400 Received: from foss.arm.com (foss.arm.com [217.140.110.172]) by lindbergh.monkeyblade.net (Postfix) with ESMTP id C6194211C for ; Thu, 10 Aug 2023 03:33:54 -0700 (PDT) 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 DBA19139F; Thu, 10 Aug 2023 03:34:36 -0700 (PDT) Received: from e125769.cambridge.arm.com (e125769.cambridge.arm.com [10.1.196.26]) by usa-sjc-imap-foss1.foss.arm.com (Postfix) with ESMTPSA id 0251A3F6C4; Thu, 10 Aug 2023 03:33:51 -0700 (PDT) From: Ryan Roberts To: Will Deacon , "Aneesh Kumar K.V" , Andrew Morton , Nick Piggin , Peter Zijlstra , Christian Borntraeger , Sven Schnelle , Arnd Bergmann , "Matthew Wilcox (Oracle)" , David Hildenbrand , Yu Zhao , "Kirill A. Shutemov" , Yin Fengwei , Yang Shi , "Huang, Ying" , Zi Yan Cc: Ryan Roberts , linux-mm@kvack.org, linux-kernel@vger.kernel.org Subject: [PATCH v1 3/4] mm/mmu_gather: Remove encoded_page infrastructure Date: Thu, 10 Aug 2023 11:33:31 +0100 Message-Id: <20230810103332.3062143-4-ryan.roberts@arm.com> X-Mailer: git-send-email 2.25.1 In-Reply-To: <20230810103332.3062143-1-ryan.roberts@arm.com> References: <20230810103332.3062143-1-ryan.roberts@arm.com> MIME-Version: 1.0 X-Spam-Status: No, score=-1.9 required=5.0 tests=BAYES_00, RCVD_IN_DNSWL_BLOCKED,SPF_HELO_NONE,SPF_NONE autolearn=ham autolearn_force=no version=3.4.6 X-Spam-Checker-Version: SpamAssassin 3.4.6 (2021-04-09) on lindbergh.monkeyblade.net Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org X-getmail-retrieved-from-mailbox: INBOX X-GMAIL-THRID: 1773840405110738651 X-GMAIL-MSGID: 1773840405110738651 commit 70fb4fdff582 ("mm: introduce 'encoded' page pointers with embedded extra bits") and commit 7cc8f9c7146a ("mm: mmu_gather: prepare to gather encoded page pointers with flags") converted mmu_gather for dealing with encoded_page, where the bottom 2 bits could encode extra flags. Only 1 bit was ever used; to flag whether the page should participate in a delayed rmap removal. Now that the mmu_gather batched rmap removal mechanism has been generalized, all pages participate and therefore the flag is unused. So let's remove encoded_page to simplify the code. It also gets in the way of further optimization which will be done in a follow up patch. Signed-off-by: Ryan Roberts --- arch/s390/include/asm/tlb.h | 9 +++------ include/asm-generic/tlb.h | 10 +++++----- include/linux/mm.h | 4 +--- include/linux/mm_types.h | 34 +--------------------------------- include/linux/swap.h | 2 +- mm/memory.c | 2 +- mm/mmu_gather.c | 11 +++++------ mm/swap.c | 8 +++----- mm/swap_state.c | 4 ++-- 9 files changed, 22 insertions(+), 62 deletions(-) diff --git a/arch/s390/include/asm/tlb.h b/arch/s390/include/asm/tlb.h index 383b1f91442c..c40b44f6a31b 100644 --- a/arch/s390/include/asm/tlb.h +++ b/arch/s390/include/asm/tlb.h @@ -25,7 +25,7 @@ void __tlb_remove_table(void *_table); static inline void tlb_flush(struct mmu_gather *tlb); static inline bool __tlb_remove_page_size(struct mmu_gather *tlb, - struct encoded_page *page, + struct page *page, int page_size); #define tlb_flush tlb_flush @@ -41,15 +41,12 @@ static inline bool __tlb_remove_page_size(struct mmu_gather *tlb, * Release the page cache reference for a pte removed by * tlb_ptep_clear_flush. In both flush modes the tlb for a page cache page * has already been freed, so just do free_page_and_swap_cache. - * - * s390 doesn't delay rmap removal, so there is nothing encoded in - * the page pointer. */ static inline bool __tlb_remove_page_size(struct mmu_gather *tlb, - struct encoded_page *page, + struct page *page, int page_size) { - free_page_and_swap_cache(encoded_page_ptr(page)); + free_page_and_swap_cache(page); return false; } diff --git a/include/asm-generic/tlb.h b/include/asm-generic/tlb.h index f339d68cf44f..d874415aaa33 100644 --- a/include/asm-generic/tlb.h +++ b/include/asm-generic/tlb.h @@ -246,7 +246,7 @@ struct mmu_gather_batch { struct mmu_gather_batch *next; unsigned int nr; unsigned int max; - struct encoded_page *encoded_pages[]; + struct page *pages[]; }; #define MAX_GATHER_BATCH \ @@ -261,7 +261,7 @@ struct mmu_gather_batch { #define MAX_GATHER_BATCH_COUNT (10000UL/MAX_GATHER_BATCH) extern bool __tlb_remove_page_size(struct mmu_gather *tlb, - struct encoded_page *page, + struct page *page, int page_size); #ifdef CONFIG_SMP @@ -464,13 +464,13 @@ static inline void tlb_flush_mmu_tlbonly(struct mmu_gather *tlb) static inline void tlb_remove_page_size(struct mmu_gather *tlb, struct page *page, int page_size) { - if (__tlb_remove_page_size(tlb, encode_page(page, 0), page_size)) + if (__tlb_remove_page_size(tlb, page, page_size)) tlb_flush_mmu(tlb); } -static __always_inline bool __tlb_remove_page(struct mmu_gather *tlb, struct page *page, unsigned int flags) +static __always_inline bool __tlb_remove_page(struct mmu_gather *tlb, struct page *page) { - return __tlb_remove_page_size(tlb, encode_page(page, flags), PAGE_SIZE); + return __tlb_remove_page_size(tlb, page, PAGE_SIZE); } /* tlb_remove_page diff --git a/include/linux/mm.h b/include/linux/mm.h index 6a95dfed4957..914e08185272 100644 --- a/include/linux/mm.h +++ b/include/linux/mm.h @@ -1518,8 +1518,7 @@ static inline void folio_put_refs(struct folio *folio, int refs) * * release_pages() releases a simple array of multiple pages, and * accepts various different forms of said page array: either - * a regular old boring array of pages, an array of folios, or - * an array of encoded page pointers. + * a regular old boring array of pages or an array of folios. * * The transparent union syntax for this kind of "any of these * argument types" is all kinds of ugly, so look away. @@ -1527,7 +1526,6 @@ static inline void folio_put_refs(struct folio *folio, int refs) typedef union { struct page **pages; struct folio **folios; - struct encoded_page **encoded_pages; } release_pages_arg __attribute__ ((__transparent_union__)); void release_pages(release_pages_arg, int nr); diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h index 291c05cacd48..b2cf57f9134c 100644 --- a/include/linux/mm_types.h +++ b/include/linux/mm_types.h @@ -68,7 +68,7 @@ struct mem_cgroup; #ifdef CONFIG_HAVE_ALIGNED_STRUCT_PAGE #define _struct_page_alignment __aligned(2 * sizeof(unsigned long)) #else -#define _struct_page_alignment __aligned(sizeof(unsigned long)) +#define _struct_page_alignment #endif struct page { @@ -216,38 +216,6 @@ struct page { #endif } _struct_page_alignment; -/* - * struct encoded_page - a nonexistent type marking this pointer - * - * An 'encoded_page' pointer is a pointer to a regular 'struct page', but - * with the low bits of the pointer indicating extra context-dependent - * information. Not super-common, but happens in mmu_gather and mlock - * handling, and this acts as a type system check on that use. - * - * We only really have two guaranteed bits in general, although you could - * play with 'struct page' alignment (see CONFIG_HAVE_ALIGNED_STRUCT_PAGE) - * for more. - * - * Use the supplied helper functions to endcode/decode the pointer and bits. - */ -struct encoded_page; -#define ENCODE_PAGE_BITS 3ul -static __always_inline struct encoded_page *encode_page(struct page *page, unsigned long flags) -{ - BUILD_BUG_ON(flags > ENCODE_PAGE_BITS); - return (struct encoded_page *)(flags | (unsigned long)page); -} - -static inline unsigned long encoded_page_flags(struct encoded_page *page) -{ - return ENCODE_PAGE_BITS & (unsigned long)page; -} - -static inline struct page *encoded_page_ptr(struct encoded_page *page) -{ - return (struct page *)(~ENCODE_PAGE_BITS & (unsigned long)page); -} - /** * struct folio - Represents a contiguous set of bytes. * @flags: Identical to the page flags. diff --git a/include/linux/swap.h b/include/linux/swap.h index bb5adc604144..f199df803b33 100644 --- a/include/linux/swap.h +++ b/include/linux/swap.h @@ -453,7 +453,7 @@ static inline unsigned long total_swapcache_pages(void) extern void free_swap_cache(struct page *page); extern void free_page_and_swap_cache(struct page *); -extern void free_pages_and_swap_cache(struct encoded_page **, int); +extern void free_pages_and_swap_cache(struct page **, int); /* linux/mm/swapfile.c */ extern atomic_long_t nr_swap_pages; extern long total_swap_pages; diff --git a/mm/memory.c b/mm/memory.c index 94a6ebd409a6..b4f757171cf9 100644 --- a/mm/memory.c +++ b/mm/memory.c @@ -1460,7 +1460,7 @@ static unsigned long zap_pte_range(struct mmu_gather *tlb, if (unlikely(page_mapcount(page) < 0)) print_bad_pte(vma, addr, ptent, page); } - if (unlikely(__tlb_remove_page(tlb, page, 0))) { + if (unlikely(__tlb_remove_page(tlb, page))) { force_flush = 1; addr += PAGE_SIZE; break; diff --git a/mm/mmu_gather.c b/mm/mmu_gather.c index ca328ecef5c2..5d100ac85e21 100644 --- a/mm/mmu_gather.c +++ b/mm/mmu_gather.c @@ -49,8 +49,7 @@ static void tlb_flush_rmap_batch(struct mmu_gather_batch *batch, struct vm_area_struct *vma) { for (int i = first; i < batch->nr; i++) { - struct encoded_page *enc = batch->encoded_pages[i]; - struct page *page = encoded_page_ptr(enc); + struct page *page = batch->pages[i]; page_remove_rmap(page, vma, false); } @@ -95,7 +94,7 @@ static void tlb_batch_pages_flush(struct mmu_gather *tlb) struct mmu_gather_batch *batch; for (batch = &tlb->local; batch && batch->nr; batch = batch->next) { - struct encoded_page **pages = batch->encoded_pages; + struct page **pages = batch->pages; do { /* @@ -125,7 +124,7 @@ static void tlb_batch_list_free(struct mmu_gather *tlb) tlb->local.next = NULL; } -bool __tlb_remove_page_size(struct mmu_gather *tlb, struct encoded_page *page, int page_size) +bool __tlb_remove_page_size(struct mmu_gather *tlb, struct page *page, int page_size) { struct mmu_gather_batch *batch; @@ -140,13 +139,13 @@ bool __tlb_remove_page_size(struct mmu_gather *tlb, struct encoded_page *page, i * Add the page and check if we are full. If so * force a flush. */ - batch->encoded_pages[batch->nr++] = page; + batch->pages[batch->nr++] = page; if (batch->nr == batch->max) { if (!tlb_next_batch(tlb)) return true; batch = tlb->active; } - VM_BUG_ON_PAGE(batch->nr > batch->max, encoded_page_ptr(page)); + VM_BUG_ON_PAGE(batch->nr > batch->max, page); return false; } diff --git a/mm/swap.c b/mm/swap.c index cd8f0150ba3a..b05cce475202 100644 --- a/mm/swap.c +++ b/mm/swap.c @@ -953,14 +953,12 @@ void lru_cache_disable(void) * Decrement the reference count on all the pages in @arg. If it * fell to zero, remove the page from the LRU and free it. * - * Note that the argument can be an array of pages, encoded pages, - * or folio pointers. We ignore any encoded bits, and turn any of - * them into just a folio that gets free'd. + * Note that the argument can be an array of pages or folio pointers. */ void release_pages(release_pages_arg arg, int nr) { int i; - struct encoded_page **encoded = arg.encoded_pages; + struct page **pages = arg.pages; LIST_HEAD(pages_to_free); struct lruvec *lruvec = NULL; unsigned long flags = 0; @@ -970,7 +968,7 @@ void release_pages(release_pages_arg arg, int nr) struct folio *folio; /* Turn any of the argument types into a folio */ - folio = page_folio(encoded_page_ptr(encoded[i])); + folio = page_folio(pages[i]); /* * Make sure the IRQ-safe lock-holding time does not get diff --git a/mm/swap_state.c b/mm/swap_state.c index 01f15139b7d9..73b16795b0ff 100644 --- a/mm/swap_state.c +++ b/mm/swap_state.c @@ -307,11 +307,11 @@ void free_page_and_swap_cache(struct page *page) * Passed an array of pages, drop them all from swapcache and then release * them. They are removed from the LRU and freed if this is their last use. */ -void free_pages_and_swap_cache(struct encoded_page **pages, int nr) +void free_pages_and_swap_cache(struct page **pages, int nr) { lru_add_drain(); for (int i = 0; i < nr; i++) - free_swap_cache(encoded_page_ptr(pages[i])); + free_swap_cache(pages[i]); release_pages(pages, nr); } From patchwork Thu Aug 10 10:33:32 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Ryan Roberts X-Patchwork-Id: 133905 Return-Path: Delivered-To: ouuuleilei@gmail.com Received: by 2002:a59:b824:0:b0:3f2:4152:657d with SMTP id z4csp342849vqi; Thu, 10 Aug 2023 04:16:26 -0700 (PDT) X-Google-Smtp-Source: AGHT+IEi6ht7J6urlwxeuQSLGh8gHTiWHassAEplXLY8HEs+QqFxZ+qJhPantBlLJ5C3zTx55ahZ X-Received: by 2002:a05:6a00:8ce:b0:682:759c:644d with SMTP id s14-20020a056a0008ce00b00682759c644dmr2299020pfu.27.1691666185834; Thu, 10 Aug 2023 04:16:25 -0700 (PDT) ARC-Seal: i=1; a=rsa-sha256; t=1691666185; cv=none; d=google.com; s=arc-20160816; b=r1noWaM6OlFJiqTG+QbHeezfSU85+8oXHWdHVAA7/xtDEd8tsDLD4DMpSc3QEffqqG 5PmkpXVGtpD5NMVlOapIE9qEtbfb3lnQa06vW5jCEpT8XkTo7HgGmSEk5qL+tC6bu4xj Eg0wUvUiPQOuTbwB+UIVKMZKOBAoLz0IoOSBQLJny2yqpns36c/CoVZI7B8NjKFV2YBc nZasGYE8HwPqUILRYf5zm4PEKD/JjaSJthd2Zvi951lkhEgc90PfyOf8janNtURhrLMG g7jldNbbEWSyb+Oj2L1fveq7ELI+5YGMvW+SplVKPgWToYtInOSjJ3rOpD0GlX75qudt Y/iw== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=list-id:precedence:content-transfer-encoding:mime-version :references:in-reply-to:message-id:date:subject:cc:to:from; bh=cTjJh/XBkZJweU8b0ZiTSp4VgIZOyE5KVvVb0SxJfBk=; fh=y3Sw7VG+9eLsYabwWUWsSnvNS303XOkFtMtRzGTCOfU=; b=i2SmLN8eLCZKzCKuiPsvyPfeHWt6HsTPJuI1/emoql25v0MbZ06vG4w91qW1ROT9ZJ jH4ebWLyn2L5KnCEx7Sgn7nS0UgeSWvvbooTqeuhQ0zIr2+djdEQad5QwuLeymMZhK4+ SMb1aQixGCK/MbVGM/lZwU1y8M1i2fOxQRNoYwlaaYhbmejkIIys+MgA27xZShGG/mY9 +pJw7jnS1kGdDeUE/eeaSs08A4927AZ090FDebHlYE0EnqRngsHco0t1B9XqWhdz1Esn CLDXk7uVtvBve0JnTu7a5uPxukxfyayrUbi98OEZPq/0AzMCOeCxKiRhHazyZgxd9FcX uz+A== ARC-Authentication-Results: i=1; mx.google.com; spf=pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) smtp.mailfrom=linux-kernel-owner@vger.kernel.org; dmarc=fail (p=NONE sp=NONE dis=NONE) header.from=arm.com Received: from out1.vger.email (out1.vger.email. [2620:137:e000::1:20]) by mx.google.com with ESMTP id w29-20020a63491d000000b0056507c25042si1372141pga.239.2023.08.10.04.16.12; Thu, 10 Aug 2023 04:16:25 -0700 (PDT) Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20; Authentication-Results: mx.google.com; spf=pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) smtp.mailfrom=linux-kernel-owner@vger.kernel.org; dmarc=fail (p=NONE sp=NONE dis=NONE) header.from=arm.com Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S235040AbjHJKeL (ORCPT + 99 others); Thu, 10 Aug 2023 06:34:11 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:56926 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S235033AbjHJKeE (ORCPT ); Thu, 10 Aug 2023 06:34:04 -0400 Received: from foss.arm.com (foss.arm.com [217.140.110.172]) by lindbergh.monkeyblade.net (Postfix) with ESMTP id BA0192136 for ; Thu, 10 Aug 2023 03:33:57 -0700 (PDT) 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 B0EE51480; Thu, 10 Aug 2023 03:34:39 -0700 (PDT) Received: from e125769.cambridge.arm.com (e125769.cambridge.arm.com [10.1.196.26]) by usa-sjc-imap-foss1.foss.arm.com (Postfix) with ESMTPSA id CBE3B3F6C4; Thu, 10 Aug 2023 03:33:54 -0700 (PDT) From: Ryan Roberts To: Will Deacon , "Aneesh Kumar K.V" , Andrew Morton , Nick Piggin , Peter Zijlstra , Christian Borntraeger , Sven Schnelle , Arnd Bergmann , "Matthew Wilcox (Oracle)" , David Hildenbrand , Yu Zhao , "Kirill A. Shutemov" , Yin Fengwei , Yang Shi , "Huang, Ying" , Zi Yan Cc: Ryan Roberts , linux-mm@kvack.org, linux-kernel@vger.kernel.org Subject: [PATCH v1 4/4] mm/mmu_gather: Store and process pages in contig ranges Date: Thu, 10 Aug 2023 11:33:32 +0100 Message-Id: <20230810103332.3062143-5-ryan.roberts@arm.com> X-Mailer: git-send-email 2.25.1 In-Reply-To: <20230810103332.3062143-1-ryan.roberts@arm.com> References: <20230810103332.3062143-1-ryan.roberts@arm.com> MIME-Version: 1.0 X-Spam-Status: No, score=-1.9 required=5.0 tests=BAYES_00, RCVD_IN_DNSWL_BLOCKED,SPF_HELO_NONE,SPF_NONE autolearn=ham autolearn_force=no version=3.4.6 X-Spam-Checker-Version: SpamAssassin 3.4.6 (2021-04-09) on lindbergh.monkeyblade.net Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org X-getmail-retrieved-from-mailbox: INBOX X-GMAIL-THRID: 1773840562521072043 X-GMAIL-MSGID: 1773840562521072043 mmu_gather accumulates a set of pages into a buffer for later rmap removal and freeing. Page pointers were previously stored in a "linked list of arrays", then at flush time, each page in the buffer was removed from the rmap, removed from the swapcache and its refcount was decremented; if the refcount reached 0, then it was freed. With increasing numbers of large folios (or at least contiguous parts of large folios) mapped into userspace processes (pagecache pages for supporting filesystems currently, but in future also large anonymous folios), we can measurably improve performance of process teardown: - For rmap removal, we can batch-remove a range of pages belonging to the same folio with folio_remove_rmap_range(), which is more efficient because atomics can be manipulated just once per range. In the common case, it also allows us to elide adding the (anon) folio to the deferred split queue, only to remove it a bit later, once all pages of the folio have been removed fro mthe rmap. - For swapcache removal, we only need to check and remove the folio from the swap cache once, rather than trying for each individual page. - For page release, we can batch-decrement the refcount for each page in the folio and free it if it hits zero. Change the page pointer storage format within the mmu_gather batch structure to store "folio_range"s; a [start, end) page pointer pair. This allows us to run length encode a contiguous range of pages that all belong to the same folio. This likely allows us to improve cache locality a bit. But it also gives us a convenient format for implementing the above 3 optimizations. Of course if running on a system that does not extensively use large pte-mapped folios, then the RLE approach uses twice as much memory, because each range is 1 page long and uses 2 pointers. But performance measurements show no impact in terms of performance. Macro Performance Results ------------------------- Test: Timed kernel compilation on Ampere Altra (arm64), 80 jobs Configs: Comparing with and without large anon folios Without large anon folios: | kernel | real-time | kern-time | user-time | |:-----------------|------------:|------------:|------------:| | baseline-laf-off | 0.0% | 0.0% | 0.0% | | mmugather-range | -0.3% | -0.3% | 0.1% | With large anon folios (order-3): | kernel | real-time | kern-time | user-time | |:-----------------|------------:|------------:|------------:| | baseline-laf-on | 0.0% | 0.0% | 0.0% | | mmugather-range | -0.7% | -3.9% | -0.1% | Test: Timed kernel compilation in VM on Apple M2 MacBook Pro, 8 jobs Configs: Comparing with and without large anon folios Without large anon folios: | kernel | real-time | kern-time | user-time | |:-----------------|------------:|------------:|------------:| | baseline-laf-off | 0.0% | 0.0% | 0.0% | | mmugather-range | -0.9% | -2.9% | -0.6% | With large anon folios (order-3): | kernel | real-time | kern-time | user-time | |:-----------------|------------:|------------:|------------:| | baseline-laf-on | 0.0% | 0.0% | 0.0% | | mmugather-range | -0.4% | -3.7% | -0.2% | Micro Performance Results ------------------------- Flame graphs for kernel compilation on Ampere Altra show reduction in cycles consumed by __arm64_sys_exit_group syscall: Without large anon folios: -2% With large anon folios: -26% For the large anon folios case, it also shows a big difference in cost of rmap removal: baseline: cycles in page_remove_rmap(): 24.7B mmugather-range: cycles in folio_remove_rmap_range(): 5.5B Furthermore, the baseline shows 5.2B cycles used by deferred_split_folio() which has completely disappeared after applying this series. Signed-off-by: Ryan Roberts --- include/asm-generic/tlb.h | 7 +-- include/linux/mm.h | 7 +++ include/linux/swap.h | 6 +-- mm/mmu_gather.c | 56 ++++++++++++++++++++---- mm/swap.c | 91 +++++++++++++++++++++++++++++++++++++++ mm/swap_state.c | 11 ++--- 6 files changed, 158 insertions(+), 20 deletions(-) diff --git a/include/asm-generic/tlb.h b/include/asm-generic/tlb.h index d874415aaa33..fe300a64e59d 100644 --- a/include/asm-generic/tlb.h +++ b/include/asm-generic/tlb.h @@ -246,11 +246,11 @@ struct mmu_gather_batch { struct mmu_gather_batch *next; unsigned int nr; unsigned int max; - struct page *pages[]; + struct folio_range ranges[]; }; #define MAX_GATHER_BATCH \ - ((PAGE_SIZE - sizeof(struct mmu_gather_batch)) / sizeof(void *)) + ((PAGE_SIZE - sizeof(struct mmu_gather_batch)) / sizeof(struct folio_range)) /* * Limit the maximum number of mmu_gather batches to reduce a risk of soft @@ -342,7 +342,8 @@ struct mmu_gather { #ifndef CONFIG_MMU_GATHER_NO_GATHER struct mmu_gather_batch *active; struct mmu_gather_batch local; - struct page *__pages[MMU_GATHER_BUNDLE]; + struct folio_range __ranges[MMU_GATHER_BUNDLE]; + struct page *range_limit; struct mmu_gather_batch *rmap_pend; unsigned int rmap_pend_first; diff --git a/include/linux/mm.h b/include/linux/mm.h index 914e08185272..f86c905a065d 100644 --- a/include/linux/mm.h +++ b/include/linux/mm.h @@ -1513,6 +1513,13 @@ static inline void folio_put_refs(struct folio *folio, int refs) __folio_put(folio); } +struct folio_range { + struct page *start; + struct page *end; +}; + +void folios_put_refs(struct folio_range *folios, int nr); + /* * union release_pages_arg - an array of pages or folios * diff --git a/include/linux/swap.h b/include/linux/swap.h index f199df803b33..06a7cf3ad6c9 100644 --- a/include/linux/swap.h +++ b/include/linux/swap.h @@ -453,7 +453,7 @@ static inline unsigned long total_swapcache_pages(void) extern void free_swap_cache(struct page *page); extern void free_page_and_swap_cache(struct page *); -extern void free_pages_and_swap_cache(struct page **, int); +extern void free_folios_and_swap_cache(struct folio_range *, int); /* linux/mm/swapfile.c */ extern atomic_long_t nr_swap_pages; extern long total_swap_pages; @@ -530,8 +530,8 @@ static inline void put_swap_device(struct swap_info_struct *si) * so leave put_page and release_pages undeclared... */ #define free_page_and_swap_cache(page) \ put_page(page) -#define free_pages_and_swap_cache(pages, nr) \ - release_pages((pages), (nr)); +#define free_folios_and_swap_cache(folios, nr) \ + folios_put_refs((folios), (nr)) /* used to sanity check ptes in zap_pte_range when CONFIG_SWAP=0 */ #define free_swap_and_cache(e) is_pfn_swap_entry(e) diff --git a/mm/mmu_gather.c b/mm/mmu_gather.c index 5d100ac85e21..fd2ea7577817 100644 --- a/mm/mmu_gather.c +++ b/mm/mmu_gather.c @@ -22,6 +22,7 @@ static bool tlb_next_batch(struct mmu_gather *tlb) batch = tlb->active; if (batch->next) { tlb->active = batch->next; + tlb->range_limit = NULL; return true; } @@ -39,6 +40,7 @@ static bool tlb_next_batch(struct mmu_gather *tlb) tlb->active->next = batch; tlb->active = batch; + tlb->range_limit = NULL; return true; } @@ -49,9 +51,11 @@ static void tlb_flush_rmap_batch(struct mmu_gather_batch *batch, struct vm_area_struct *vma) { for (int i = first; i < batch->nr; i++) { - struct page *page = batch->pages[i]; + struct folio_range *range = &batch->ranges[i]; + int nr = range->end - range->start; + struct folio *folio = page_folio(range->start); - page_remove_rmap(page, vma, false); + folio_remove_rmap_range(folio, range->start, nr, vma); } } @@ -75,6 +79,11 @@ void tlb_flush_rmaps(struct mmu_gather *tlb, struct vm_area_struct *vma) for (batch = batch->next; batch && batch->nr; batch = batch->next) tlb_flush_rmap_batch(batch, 0, vma); + /* + * Move to the next range on next page insertion to prevent any future + * pages from being accumulated into the range we just did the rmap for. + */ + tlb->range_limit = NULL; tlb_discard_rmaps(tlb); } @@ -94,7 +103,7 @@ static void tlb_batch_pages_flush(struct mmu_gather *tlb) struct mmu_gather_batch *batch; for (batch = &tlb->local; batch && batch->nr; batch = batch->next) { - struct page **pages = batch->pages; + struct folio_range *ranges = batch->ranges; do { /* @@ -102,14 +111,15 @@ static void tlb_batch_pages_flush(struct mmu_gather *tlb) */ unsigned int nr = min(512U, batch->nr); - free_pages_and_swap_cache(pages, nr); - pages += nr; + free_folios_and_swap_cache(ranges, nr); + ranges += nr; batch->nr -= nr; cond_resched(); } while (batch->nr); } tlb->active = &tlb->local; + tlb->range_limit = NULL; tlb_discard_rmaps(tlb); } @@ -127,6 +137,7 @@ static void tlb_batch_list_free(struct mmu_gather *tlb) bool __tlb_remove_page_size(struct mmu_gather *tlb, struct page *page, int page_size) { struct mmu_gather_batch *batch; + struct folio_range *range; VM_BUG_ON(!tlb->end); @@ -135,11 +146,37 @@ bool __tlb_remove_page_size(struct mmu_gather *tlb, struct page *page, int page_ #endif batch = tlb->active; + range = &batch->ranges[batch->nr - 1]; + + /* + * If there is a range being accumulated, add the page to the range if + * its contiguous, else start the next range. range_limit is always NULL + * when nr is 0, which protects the batch->ranges[-1] case. + */ + if (tlb->range_limit && page == range->end) { + range->end++; + } else { + struct folio *folio = page_folio(page); + + range = &batch->ranges[batch->nr++]; + range->start = page; + range->end = page + 1; + + tlb->range_limit = &folio->page + folio_nr_pages(folio); + } + + /* + * If we have reached the end of the folio, move to the next range when + * we add the next page; Never span multiple folios in the same range. + */ + if (range->end == tlb->range_limit) + tlb->range_limit = NULL; + /* - * Add the page and check if we are full. If so - * force a flush. + * Check if we are full. If so force a flush. In order to ensure we + * always have a free range for the next added page, the last range in a + * batch always only has a single page. */ - batch->pages[batch->nr++] = page; if (batch->nr == batch->max) { if (!tlb_next_batch(tlb)) return true; @@ -318,8 +355,9 @@ static void __tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, tlb->need_flush_all = 0; tlb->local.next = NULL; tlb->local.nr = 0; - tlb->local.max = ARRAY_SIZE(tlb->__pages); + tlb->local.max = ARRAY_SIZE(tlb->__ranges); tlb->active = &tlb->local; + tlb->range_limit = NULL; tlb->batch_count = 0; tlb->rmap_pend = &tlb->local; tlb->rmap_pend_first = 0; diff --git a/mm/swap.c b/mm/swap.c index b05cce475202..e238d3623fcb 100644 --- a/mm/swap.c +++ b/mm/swap.c @@ -1041,6 +1041,97 @@ void release_pages(release_pages_arg arg, int nr) } EXPORT_SYMBOL(release_pages); +/** + * folios_put_refs - batched folio_put_refs() + * @folios: array of `struct folio_range`s to release + * @nr: number of folio ranges + * + * Each `struct folio_range` describes the start and end page of a range within + * a folio. The folio reference count is decremented once for each page in the + * range. If it fell to zero, remove the page from the LRU and free it. + */ +void folios_put_refs(struct folio_range *folios, int nr) +{ + int i; + LIST_HEAD(pages_to_free); + struct lruvec *lruvec = NULL; + unsigned long flags = 0; + unsigned int lock_batch; + + for (i = 0; i < nr; i++) { + struct folio *folio = page_folio(folios[i].start); + int refs = folios[i].end - folios[i].start; + + /* + * Make sure the IRQ-safe lock-holding time does not get + * excessive with a continuous string of pages from the + * same lruvec. The lock is held only if lruvec != NULL. + */ + if (lruvec && ++lock_batch == SWAP_CLUSTER_MAX) { + unlock_page_lruvec_irqrestore(lruvec, flags); + lruvec = NULL; + } + + if (is_huge_zero_page(&folio->page)) + continue; + + if (folio_is_zone_device(folio)) { + if (lruvec) { + unlock_page_lruvec_irqrestore(lruvec, flags); + lruvec = NULL; + } + if (put_devmap_managed_page(&folio->page)) + continue; + if (folio_put_testzero(folio)) + free_zone_device_page(&folio->page); + continue; + } + + if (!folio_ref_sub_and_test(folio, refs)) + continue; + + if (folio_test_large(folio)) { + if (lruvec) { + unlock_page_lruvec_irqrestore(lruvec, flags); + lruvec = NULL; + } + __folio_put_large(folio); + continue; + } + + if (folio_test_lru(folio)) { + struct lruvec *prev_lruvec = lruvec; + + lruvec = folio_lruvec_relock_irqsave(folio, lruvec, + &flags); + if (prev_lruvec != lruvec) + lock_batch = 0; + + lruvec_del_folio(lruvec, folio); + __folio_clear_lru_flags(folio); + } + + /* + * In rare cases, when truncation or holepunching raced with + * munlock after VM_LOCKED was cleared, Mlocked may still be + * found set here. This does not indicate a problem, unless + * "unevictable_pgs_cleared" appears worryingly large. + */ + if (unlikely(folio_test_mlocked(folio))) { + __folio_clear_mlocked(folio); + zone_stat_sub_folio(folio, NR_MLOCK); + count_vm_event(UNEVICTABLE_PGCLEARED); + } + + list_add(&folio->lru, &pages_to_free); + } + if (lruvec) + unlock_page_lruvec_irqrestore(lruvec, flags); + + mem_cgroup_uncharge_list(&pages_to_free); + free_unref_page_list(&pages_to_free); +} + /* * The folios which we're about to release may be in the deferred lru-addition * queues. That would prevent them from really being freed right now. That's diff --git a/mm/swap_state.c b/mm/swap_state.c index 73b16795b0ff..526bbd5a2ce1 100644 --- a/mm/swap_state.c +++ b/mm/swap_state.c @@ -304,15 +304,16 @@ void free_page_and_swap_cache(struct page *page) } /* - * Passed an array of pages, drop them all from swapcache and then release - * them. They are removed from the LRU and freed if this is their last use. + * Passed an array of folio ranges, drop all folios from swapcache and then put + * a folio reference for each page in the range. They are removed from the LRU + * and freed if this is their last use. */ -void free_pages_and_swap_cache(struct page **pages, int nr) +void free_folios_and_swap_cache(struct folio_range *folios, int nr) { lru_add_drain(); for (int i = 0; i < nr; i++) - free_swap_cache(pages[i]); - release_pages(pages, nr); + free_swap_cache(folios[i].start); + folios_put_refs(folios, nr); } static inline bool swap_use_vma_readahead(void)