From patchwork Fri May 12 06:32:38 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Vincent Whitchurch X-Patchwork-Id: 92965 Return-Path: Delivered-To: ouuuleilei@gmail.com Received: by 2002:a59:b0ea:0:b0:3b6:4342:cba0 with SMTP id b10csp4890078vqo; Thu, 11 May 2023 23:36:53 -0700 (PDT) X-Google-Smtp-Source: ACHHUZ4MCUisa0S7fqSvtQxacmGXJvBDbbj4oHC5Lp47deUNQFWNaRPkuLAGj6lF4N10o/zQa+rs X-Received: by 2002:a05:6a00:b55:b0:648:c1be:496 with SMTP id p21-20020a056a000b5500b00648c1be0496mr10885989pfo.22.1683873412932; Thu, 11 May 2023 23:36:52 -0700 (PDT) ARC-Seal: i=1; a=rsa-sha256; t=1683873412; cv=none; d=google.com; s=arc-20160816; b=ovTbarCqW/ShTHqjE/CuXyRHClfr3Ra+y7AmVIvqyWR5noVMrtrW69PDxomJ67tNRs iudgb7cdsQzwUd7U2EJgxXhWckSfDVerhQC2NLzfmJlssGFaj1hbHLyMmmyYohgX10Og dJGbyREvumtPQxdY87q4c9tdS8pJ6KMgAFUjqp+ODlsSOzSkoKHjP3up7uuRuyjL90D6 ttkRmUvPaBJqLI0n+K4rz01/gYSBH8/rCJITAH35adRy1mp6mA+6g1VEkx82BPNc37Wz 8ZilCUiGsX5QGW3opVVF6grIeDSbDcmCDmGSpq0RFu9aiZKPUMcdXBJfPP6VCsyMo65D P63w== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=list-id:precedence:cc:to:message-id:content-transfer-encoding :mime-version:subject:date:from:dkim-signature; bh=LWIqxDinQ9Zd5p8i+VillZT6T19WStn1SG6Wb7L97ZM=; b=ND9eTyn4Om6eB5yDBjFb+FE/wtkFP9NDGYlUVGOOXMQgYnigPKxvXgBdtNpJDXfPUQ SmMUyhCl2HFaYNQS7hCTrbRbW++DcpRFK/W3B6IVvTOr35vh61NJOuPiOYwIy01VYYJY Mmk9YgYxY7iLWbQF5oXZ7YXXIZgg8ccZvAVHpkoyp3rnb1huyKqliDAooEakp2JWx5M3 ToRAwBNbwpAsNEAhEFNYr0nnCQ02BnDsk+NWwYicrUgSX77a6TmTZjiUpsbRX9XWJoG0 XeisYAi4wSBnbFNbXCvkJovGoK9D+EtwI0alJKJ+lgb6ETn4DE4tEVQJTYQ5qBBZYiZP zfwg== ARC-Authentication-Results: i=1; mx.google.com; dkim=pass (test mode) header.i=@axis.com header.s=axis-central1 header.b=NNq6ZNwk; 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=pass (p=NONE sp=NONE dis=NONE) header.from=axis.com Received: from out1.vger.email (out1.vger.email. [2620:137:e000::1:20]) by mx.google.com with ESMTP id 77-20020a621850000000b0063d2b810fcbsi9649364pfy.208.2023.05.11.23.36.37; Thu, 11 May 2023 23:36:52 -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; dkim=pass (test mode) header.i=@axis.com header.s=axis-central1 header.b=NNq6ZNwk; 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=pass (p=NONE sp=NONE dis=NONE) header.from=axis.com Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S239976AbjELGdp (ORCPT + 99 others); Fri, 12 May 2023 02:33:45 -0400 Received: from lindbergh.monkeyblade.net ([23.128.96.19]:54972 "EHLO lindbergh.monkeyblade.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S239719AbjELGdn (ORCPT ); Fri, 12 May 2023 02:33:43 -0400 Received: from smtp2.axis.com (smtp2.axis.com [195.60.68.18]) by lindbergh.monkeyblade.net (Postfix) with ESMTPS id 5460A49D6 for ; Thu, 11 May 2023 23:33:01 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=axis.com; q=dns/txt; s=axis-central1; t=1683873181; x=1715409181; h=from:date:subject:mime-version:content-transfer-encoding: message-id:to:cc; bh=LWIqxDinQ9Zd5p8i+VillZT6T19WStn1SG6Wb7L97ZM=; b=NNq6ZNwkV7ZYEvEpWH/KxmXC1yYbAaHs01m/I8W6pZoQJlLTs0M4yfKd wLwEhEumdvKOOtMDujhPeaPFExATNkfDmS4SZM198oUZ910P50VlLanY/ S0cLZI6k3Ja3VBoltqzCz//MBr8YxZuEgaGgbAEJyrExT7anXGMQMS0JI 5iiZsNPTGZnpfbZMMxR8PVn2vobxiysFKdWhkdbQqwiXP9A6vwt13bgIg zDMLcx7nnPLRBf/YtT59xIrq8PAvsU/oduqf1IsCfwCQE+e5AQ9NHP4oN Y0aTzhIMJcgqP1gdXQNirn6gtvZfzCSzSlDRiKenF/2x00uUKX88EPtrv A==; From: Vincent Whitchurch Date: Fri, 12 May 2023 08:32:38 +0200 Subject: [PATCH] squashfs: cache partial compressed blocks MIME-Version: 1.0 Message-ID: <20230510-squashfs-cache-v1-1-3b6bb0e7d952@axis.com> X-B4-Tracking: v=1; b=H4sIAIXdXWQC/z2NQQrCMBAAv1L27ErSEIt+RTxs1q3ZS6pZKoXSv zf14HFghlnBpKoY3LoVqnzVdCoN/KkDzlRegvpsDL3rg4veoX1msjwaMnEWHCikqwwhRvbQokQ mmCoVzkf2ty/ngEknO5x3lVGX3/T+2LYdf0Fkf4QAAAA= To: Phillip Lougher , Andrew Morton CC: Philippe Liard , , , , , Vincent Whitchurch X-Mailer: b4 0.12.1 X-Spam-Status: No, score=-4.4 required=5.0 tests=BAYES_00,DKIM_SIGNED, DKIM_VALID,DKIM_VALID_AU,DKIM_VALID_EF,RCVD_IN_DNSWL_MED,SPF_HELO_PASS, SPF_PASS,T_SCC_BODY_TEXT_LINE,URIBL_BLOCKED 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: =?utf-8?q?INBOX?= X-GMAIL-THRID: =?utf-8?q?1765669247720603735?= X-GMAIL-MSGID: =?utf-8?q?1765669247720603735?= Before commit 93e72b3c612adcaca1 ("squashfs: migrate from ll_rw_block usage to BIO"), compressed blocks read by squashfs were cached in the page cache, but that is not the case after that commit. That has lead to squashfs having to re-read a lot of sectors from disk/flash. For example, the first sectors of every metadata block need to be read twice from the disk. Once partially to read the length, and a second time to read the block itself. Also, in linear reads of large files, the last sectors of one data block are re-read from disk when reading the next data block, since the compressed blocks are of variable sizes and not aligned to device blocks. This extra I/O results in a degrade in read performance of, for example, ~16% in one scenario on my ARM platform using squashfs with dm-verity and NAND. Since the decompressed data is cached in the page cache or squashfs' internal metadata and fragment caches, caching _all_ compressed pages would lead to a lot of double caching and is undesirable. But make the code cache any disk blocks which were only partially requested, since these are the ones likely to include data which is needed by other file system blocks. This restores read performance in my test scenario. The compressed block caching is only applied when the disk block size is equal to the page size, to avoid having to deal with caching sub-page reads. Signed-off-by: Vincent Whitchurch --- fs/squashfs/block.c | 118 +++++++++++++++++++++++++++++++++++++++++-- fs/squashfs/squashfs_fs_sb.h | 1 + fs/squashfs/super.c | 12 +++++ 3 files changed, 127 insertions(+), 4 deletions(-) --- base-commit: 457391b0380335d5e9a5babdec90ac53928b23b4 change-id: 20230510-squashfs-cache-7a3b9e7355c1 Best regards, diff --git a/fs/squashfs/block.c b/fs/squashfs/block.c index bed3bb8b27fa..e565ff574e24 100644 --- a/fs/squashfs/block.c +++ b/fs/squashfs/block.c @@ -76,10 +76,101 @@ static int copy_bio_to_actor(struct bio *bio, return copied_bytes; } -static int squashfs_bio_read(struct super_block *sb, u64 index, int length, +static int squashfs_bio_read_cached(struct bio *fullbio, struct address_space *cache_mapping, + u64 index, int length, u64 read_start, u64 read_end, + int page_count) +{ + struct page *head_to_cache = NULL, *tail_to_cache = NULL; + struct block_device *bdev = fullbio->bi_bdev; + struct bvec_iter_all iter_all; + struct bio *bio = NULL; + int prev_io_idx = -1; + struct bio_vec *bv; + int idx = 0; + int err = 0; + + bio_for_each_segment_all(bv, fullbio, iter_all) { + struct page *page = bv->bv_page; + int retlen; + + if (page->mapping == cache_mapping && PageUptodate(page)) { + idx++; + continue; + } + + /* + * We only use this when the device block size is the same as + * the page size, so read_start and read_end cover full pages. + * + * Compare these to the original required index and length to + * only cache pages which were requested partially, since these + * are the ones which are likely to be needed when reading + * adjacent blocks. + */ + if (idx == 0 && index != read_start) + head_to_cache = page; + else if (idx == page_count - 1 && index + length != read_end) + tail_to_cache = page; + + if (!bio || idx != prev_io_idx + 1) { + unsigned int remaining_pages; + unsigned int this_nr_pages; + +submit_and_retry: + remaining_pages = page_count - idx; + this_nr_pages = min(remaining_pages, BIO_MAX_VECS); + bio = blk_next_bio(bio, bdev, this_nr_pages, REQ_OP_READ, + GFP_NOIO); + bio->bi_iter.bi_sector = fullbio->bi_iter.bi_sector + + idx * (PAGE_SIZE / SECTOR_SIZE); + } + + retlen = bio_add_page(bio, bv->bv_page, bv->bv_len, bv->bv_offset); + if (retlen != bv->bv_len) + goto submit_and_retry; + + prev_io_idx = idx; + idx++; + } + + if (bio) { + err = submit_bio_wait(bio); + bio_put(bio); + } + + if (err) + return err; + + if (head_to_cache) { + int ret = add_to_page_cache_lru(head_to_cache, cache_mapping, + read_start, GFP_NOIO); + + if (!ret) { + SetPageUptodate(head_to_cache); + unlock_page(head_to_cache); + } + + } + + if (tail_to_cache) { + int ret = add_to_page_cache_lru(tail_to_cache, cache_mapping, + read_end - PAGE_SIZE, GFP_NOIO); + + if (!ret) { + SetPageUptodate(tail_to_cache); + unlock_page(tail_to_cache); + } + } + + return 0; +} + +int squashfs_bio_read(struct super_block *sb, u64 index, int length, struct bio **biop, int *block_offset) { struct squashfs_sb_info *msblk = sb->s_fs_info; + struct inode *cache_inode = msblk->cache_inode; + struct address_space *cache_mapping = cache_inode ? cache_inode->i_mapping : NULL; const u64 read_start = round_down(index, msblk->devblksize); const sector_t block = read_start >> msblk->devblksize_log2; const u64 read_end = round_up(index + length, msblk->devblksize); @@ -99,13 +190,27 @@ static int squashfs_bio_read(struct super_block *sb, u64 index, int length, for (i = 0; i < page_count; ++i) { unsigned int len = min_t(unsigned int, PAGE_SIZE - offset, total_len); - struct page *page = alloc_page(GFP_NOIO); + struct page *page = NULL; + + if (cache_mapping) + page = find_get_page(cache_mapping, + read_start + i * PAGE_SIZE); + if (!page) + page = alloc_page(GFP_NOIO); if (!page) { error = -ENOMEM; goto out_free_bio; } - if (!bio_add_page(bio, page, len, offset)) { + + if (cache_mapping) { + /* + * Use the __ version to avoid merging since we need + * each page to be separate when we check for and avoid + * cached pages. + */ + __bio_add_page(bio, page, len, offset); + } else if (!bio_add_page(bio, page, len, offset)) { error = -EIO; goto out_free_bio; } @@ -113,7 +218,12 @@ static int squashfs_bio_read(struct super_block *sb, u64 index, int length, total_len -= len; } - error = submit_bio_wait(bio); + if (cache_mapping) + error = squashfs_bio_read_cached(bio, cache_mapping, index, + length, read_start, read_end, + page_count); + else + error = submit_bio_wait(bio); if (error) goto out_free_bio; diff --git a/fs/squashfs/squashfs_fs_sb.h b/fs/squashfs/squashfs_fs_sb.h index 72f6f4b37863..dfee65845d48 100644 --- a/fs/squashfs/squashfs_fs_sb.h +++ b/fs/squashfs/squashfs_fs_sb.h @@ -47,6 +47,7 @@ struct squashfs_sb_info { struct squashfs_cache *block_cache; struct squashfs_cache *fragment_cache; struct squashfs_cache *read_page; + struct inode *cache_inode; int next_meta_index; __le64 *id_table; __le64 *fragment_index; diff --git a/fs/squashfs/super.c b/fs/squashfs/super.c index e090fae48e68..64d6bc95950b 100644 --- a/fs/squashfs/super.c +++ b/fs/squashfs/super.c @@ -329,6 +329,16 @@ static int squashfs_fill_super(struct super_block *sb, struct fs_context *fc) goto failed_mount; } + if (msblk->devblksize == PAGE_SIZE) { + msblk->cache_inode = new_inode(sb); + if (msblk->cache_inode == NULL) + goto failed_mount; + + set_nlink(msblk->cache_inode, 1); + msblk->cache_inode->i_size = OFFSET_MAX; + mapping_set_gfp_mask(msblk->cache_inode->i_mapping, GFP_NOFS); + } + msblk->stream = squashfs_decompressor_setup(sb, flags); if (IS_ERR(msblk->stream)) { err = PTR_ERR(msblk->stream); @@ -454,6 +464,7 @@ static int squashfs_fill_super(struct super_block *sb, struct fs_context *fc) squashfs_cache_delete(msblk->block_cache); squashfs_cache_delete(msblk->fragment_cache); squashfs_cache_delete(msblk->read_page); + iput(msblk->cache_inode); msblk->thread_ops->destroy(msblk); kfree(msblk->inode_lookup_table); kfree(msblk->fragment_index); @@ -572,6 +583,7 @@ static void squashfs_put_super(struct super_block *sb) squashfs_cache_delete(sbi->block_cache); squashfs_cache_delete(sbi->fragment_cache); squashfs_cache_delete(sbi->read_page); + iput(sbi->cache_inode); sbi->thread_ops->destroy(sbi); kfree(sbi->id_table); kfree(sbi->fragment_index);