From patchwork Fri Nov 4 01:33:00 2016 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Qu Wenruo X-Patchwork-Id: 9411737 Return-Path: Received: from mail.wl.linuxfoundation.org (pdx-wl-mail.web.codeaurora.org [172.30.200.125]) by pdx-korg-patchwork.web.codeaurora.org (Postfix) with ESMTP id C3681601C2 for ; Fri, 4 Nov 2016 01:34:09 +0000 (UTC) Received: from mail.wl.linuxfoundation.org (localhost [127.0.0.1]) by mail.wl.linuxfoundation.org (Postfix) with ESMTP id B432D2B030 for ; Fri, 4 Nov 2016 01:34:09 +0000 (UTC) Received: by mail.wl.linuxfoundation.org (Postfix, from userid 486) id A8CD22B039; Fri, 4 Nov 2016 01:34:09 +0000 (UTC) X-Spam-Checker-Version: SpamAssassin 3.3.1 (2010-03-16) on pdx-wl-mail.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-6.9 required=2.0 tests=BAYES_00,RCVD_IN_DNSWL_HI autolearn=ham version=3.3.1 Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.wl.linuxfoundation.org (Postfix) with ESMTP id 58AFC2B035 for ; Fri, 4 Nov 2016 01:34:07 +0000 (UTC) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S933641AbcKDBd6 (ORCPT ); Thu, 3 Nov 2016 21:33:58 -0400 Received: from cn.fujitsu.com ([222.73.24.84]:50610 "EHLO song.cn.fujitsu.com" rhost-flags-OK-FAIL-OK-OK) by vger.kernel.org with ESMTP id S933311AbcKDBdz (ORCPT ); Thu, 3 Nov 2016 21:33:55 -0400 X-IronPort-AV: E=Sophos;i="5.20,367,1444665600"; d="scan'208";a="943838" Received: from unknown (HELO cn.fujitsu.com) ([10.167.250.3]) by song.cn.fujitsu.com with ESMTP; 04 Nov 2016 09:33:21 +0800 Received: from localhost.localdomain (unknown [10.167.226.34]) by cn.fujitsu.com (Postfix) with ESMTP id 67214406CF61; Fri, 4 Nov 2016 09:33:18 +0800 (CST) From: Qu Wenruo To: linux-btrfs@vger.kernel.org Cc: Wang Xiaoguang Subject: [PATCH v14 11/15] btrfs: dedupe: Inband in-memory only de-duplication implement Date: Fri, 4 Nov 2016 09:33:00 +0800 Message-Id: <20161104013304.11365-12-quwenruo@cn.fujitsu.com> X-Mailer: git-send-email 2.10.1 In-Reply-To: <20161104013304.11365-1-quwenruo@cn.fujitsu.com> References: <20161104013304.11365-1-quwenruo@cn.fujitsu.com> MIME-Version: 1.0 X-yoursite-MailScanner-ID: 67214406CF61.AEFE8 X-yoursite-MailScanner: Found to be clean X-yoursite-MailScanner-From: quwenruo@cn.fujitsu.com Sender: linux-btrfs-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-btrfs@vger.kernel.org X-Virus-Scanned: ClamAV using ClamSMTP Core implement for inband de-duplication. It reuse the async_cow_start() facility to do the calculate dedupe hash. And use dedupe hash to do inband de-duplication at extent level. The work flow is as below: 1) Run delalloc range for an inode 2) Calculate hash for the delalloc range at the unit of dedupe_bs 3) For hash match(duplicated) case, just increase source extent ref and insert file extent. For hash mismatch case, go through the normal cow_file_range() fallback, and add hash into dedupe_tree. Compress for hash miss case is not supported yet. Current implement restore all dedupe hash in memory rb-tree, with LRU behavior to control the limit. Signed-off-by: Wang Xiaoguang Signed-off-by: Qu Wenruo --- fs/btrfs/extent-tree.c | 20 ++++ fs/btrfs/inode.c | 256 ++++++++++++++++++++++++++++++++++++++++++------- fs/btrfs/relocation.c | 16 ++++ 3 files changed, 260 insertions(+), 32 deletions(-) diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 9aa6d2c..0a8f0c4 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -37,6 +37,7 @@ #include "math.h" #include "sysfs.h" #include "qgroup.h" +#include "dedupe.h" #undef SCRAMBLE_DELAYED_REFS @@ -2399,6 +2400,8 @@ static int run_one_delayed_ref(struct btrfs_trans_handle *trans, if (btrfs_delayed_ref_is_head(node)) { struct btrfs_delayed_ref_head *head; + struct btrfs_fs_info *fs_info = root->fs_info; + /* * we've hit the end of the chain and we were supposed * to insert this extent into the tree. But, it got @@ -2414,6 +2417,18 @@ static int run_one_delayed_ref(struct btrfs_trans_handle *trans, btrfs_pin_extent(root, node->bytenr, node->num_bytes, 1); if (head->is_data) { + /* + * If insert_reserved is given, it means + * a new extent is revered, then deleted + * in one tran, and inc/dec get merged to 0. + * + * In this case, we need to remove its dedupe + * hash. + */ + ret = btrfs_dedupe_del(trans, fs_info, + node->bytenr); + if (ret < 0) + return ret; ret = btrfs_del_csums(trans, root, node->bytenr, node->num_bytes); @@ -7087,6 +7102,11 @@ static int __btrfs_free_extent(struct btrfs_trans_handle *trans, btrfs_release_path(path); if (is_data) { + ret = btrfs_dedupe_del(trans, info, bytenr); + if (ret < 0) { + btrfs_abort_transaction(trans, ret); + goto out; + } ret = btrfs_del_csums(trans, root, bytenr, num_bytes); if (ret) { btrfs_abort_transaction(trans, ret); diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c index f364d19..ff35baa 100644 --- a/fs/btrfs/inode.c +++ b/fs/btrfs/inode.c @@ -337,6 +337,7 @@ struct async_extent { struct page **pages; unsigned long nr_pages; int compress_type; + struct btrfs_dedupe_hash *hash; struct list_head list; }; @@ -355,7 +356,8 @@ static noinline int add_async_extent(struct async_cow *cow, u64 compressed_size, struct page **pages, unsigned long nr_pages, - int compress_type) + int compress_type, + struct btrfs_dedupe_hash *hash) { struct async_extent *async_extent; @@ -367,6 +369,7 @@ static noinline int add_async_extent(struct async_cow *cow, async_extent->pages = pages; async_extent->nr_pages = nr_pages; async_extent->compress_type = compress_type; + async_extent->hash = hash; list_add_tail(&async_extent->list, &cow->extents); return 0; } @@ -599,7 +602,7 @@ cont: */ add_async_extent(async_cow, start, num_bytes, total_compressed, pages, nr_pages_ret, - compress_type); + compress_type, NULL); if (start + num_bytes < end) { start += num_bytes; @@ -645,7 +648,7 @@ cleanup_and_bail_uncompressed: if (redirty) extent_range_redirty_for_io(inode, start, end); add_async_extent(async_cow, start, end - start + 1, 0, NULL, 0, - BTRFS_COMPRESS_NONE); + BTRFS_COMPRESS_NONE, NULL); *num_added += 1; return; @@ -674,6 +677,38 @@ static void free_async_extent_pages(struct async_extent *async_extent) async_extent->pages = NULL; } +static void end_dedupe_extent(struct inode *inode, u64 start, + u32 len, unsigned long page_ops) +{ + int i; + unsigned int nr_pages = len / PAGE_SIZE; + struct page *page; + + for (i = 0; i < nr_pages; i++) { + page = find_get_page(inode->i_mapping, + start >> PAGE_SHIFT); + /* page should be already locked by caller */ + if (WARN_ON(!page)) + continue; + + /* We need to do this by ourselves as we skipped IO */ + if (page_ops & PAGE_CLEAR_DIRTY) + clear_page_dirty_for_io(page); + if (page_ops & PAGE_SET_WRITEBACK) + set_page_writeback(page); + + end_extent_writepage(page, 0, start, + start + PAGE_SIZE - 1); + if (page_ops & PAGE_END_WRITEBACK) + end_page_writeback(page); + if (page_ops & PAGE_UNLOCK) + unlock_page(page); + + start += PAGE_SIZE; + put_page(page); + } +} + /* * phase two of compressed writeback. This is the ordered portion * of the code, which only gets called in the order the work was @@ -690,6 +725,7 @@ static noinline void submit_compressed_extents(struct inode *inode, struct btrfs_root *root = BTRFS_I(inode)->root; struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree; struct extent_io_tree *io_tree; + struct btrfs_dedupe_hash *hash; int ret = 0; again: @@ -699,6 +735,7 @@ again: list_del(&async_extent->list); io_tree = &BTRFS_I(inode)->io_tree; + hash = async_extent->hash; retry: /* did the compression code fall back to uncompressed IO? */ @@ -728,7 +765,7 @@ retry: async_extent->start + async_extent->ram_size - 1, &page_started, &nr_written, 0, - NULL); + hash); /* JDM XXX */ @@ -738,15 +775,26 @@ retry: * and IO for us. Otherwise, we need to submit * all those pages down to the drive. */ - if (!page_started && !ret) - extent_write_locked_range(io_tree, - inode, async_extent->start, - async_extent->start + - async_extent->ram_size - 1, - btrfs_get_extent, - WB_SYNC_ALL); - else if (ret) + if (!page_started && !ret) { + /* Skip IO for dedupe async_extent */ + if (btrfs_dedupe_hash_hit(hash)) + end_dedupe_extent(inode, + async_extent->start, + async_extent->ram_size, + PAGE_CLEAR_DIRTY | + PAGE_SET_WRITEBACK | + PAGE_END_WRITEBACK | + PAGE_UNLOCK); + else + extent_write_locked_range(io_tree, + inode, async_extent->start, + async_extent->start + + async_extent->ram_size - 1, + btrfs_get_extent, + WB_SYNC_ALL); + } else if (ret) unlock_page(async_cow->locked_page); + kfree(hash); kfree(async_extent); cond_resched(); continue; @@ -876,6 +924,7 @@ retry: free_async_extent_pages(async_extent); } alloc_hint = ins.objectid + ins.offset; + kfree(hash); kfree(async_extent); cond_resched(); } @@ -895,6 +944,7 @@ out_free: PAGE_SET_WRITEBACK | PAGE_END_WRITEBACK | PAGE_SET_ERROR); free_async_extent_pages(async_extent); + kfree(hash); kfree(async_extent); goto again; } @@ -1009,11 +1059,17 @@ static noinline int cow_file_range(struct inode *inode, unsigned long op; cur_alloc_size = disk_num_bytes; - ret = btrfs_reserve_extent(root, cur_alloc_size, cur_alloc_size, + if (btrfs_dedupe_hash_hit(hash)) { + ins.objectid = hash->bytenr; + ins.offset = hash->num_bytes; + } else { + ret = btrfs_reserve_extent(root, cur_alloc_size, + cur_alloc_size, root->sectorsize, 0, alloc_hint, &ins, 1, 1); - if (ret < 0) - goto out_unlock; + if (ret < 0) + goto out_unlock; + } em = alloc_extent_map(); if (!em) { @@ -1050,8 +1106,9 @@ static noinline int cow_file_range(struct inode *inode, goto out_reserve; cur_alloc_size = ins.offset; - ret = btrfs_add_ordered_extent(inode, start, ins.objectid, - ram_size, cur_alloc_size, 0); + ret = btrfs_add_ordered_extent_dedupe(inode, start, + ins.objectid, cur_alloc_size, ins.offset, + 0, hash); if (ret) goto out_drop_extent_cache; @@ -1063,7 +1120,14 @@ static noinline int cow_file_range(struct inode *inode, goto out_drop_extent_cache; } - btrfs_dec_block_group_reservations(root->fs_info, ins.objectid); + /* + * Hash hit didn't allocate extent, no need to dec bg + * reservation. + * Or we will underflow reservations and block balance. + */ + if (!btrfs_dedupe_hash_hit(hash)) + btrfs_dec_block_group_reservations(root->fs_info, + ins.objectid); if (disk_num_bytes < cur_alloc_size) break; @@ -1106,6 +1170,79 @@ out_unlock: goto out; } +static int hash_file_ranges(struct inode *inode, u64 start, u64 end, + struct async_cow *async_cow, int *num_added) +{ + struct btrfs_root *root = BTRFS_I(inode)->root; + struct btrfs_fs_info *fs_info = root->fs_info; + struct btrfs_dedupe_info *dedupe_info = fs_info->dedupe_info; + struct page *locked_page = async_cow->locked_page; + u16 hash_algo; + u64 dedupe_bs; + u64 cur_offset = start; + int ret = 0; + + /* If dedupe is not enabled, don't split extent into dedupe_bs */ + if (fs_info->dedupe_enabled && dedupe_info) { + dedupe_bs = dedupe_info->blocksize; + hash_algo = dedupe_info->hash_algo; + } else { + dedupe_bs = SZ_128M; + /* Just dummy, to avoid access NULL pointer */ + hash_algo = BTRFS_DEDUPE_HASH_SHA256; + } + + while (cur_offset < end) { + struct btrfs_dedupe_hash *hash = NULL; + u64 len; + + len = min(end + 1 - cur_offset, dedupe_bs); + if (len < dedupe_bs) + goto next; + + hash = btrfs_dedupe_alloc_hash(hash_algo); + if (!hash) { + ret = -ENOMEM; + goto out; + } + ret = btrfs_dedupe_calc_hash(fs_info, inode, cur_offset, hash); + if (ret < 0) { + kfree(hash); + goto out; + } + + ret = btrfs_dedupe_search(fs_info, inode, cur_offset, hash); + if (ret < 0) { + kfree(hash); + goto out; + } + ret = 0; + +next: + /* Redirty the locked page if it corresponds to our extent */ + if (page_offset(locked_page) >= start && + page_offset(locked_page) <= end) + __set_page_dirty_nobuffers(locked_page); + + add_async_extent(async_cow, cur_offset, len, 0, NULL, 0, + BTRFS_COMPRESS_NONE, hash); + cur_offset += len; + (*num_added)++; + } +out: + /* + * Caller won't unlock pages, so if error happens, we must unlock + * pages by ourselves. + */ + if (ret) + extent_clear_unlock_delalloc(inode, cur_offset, + end, end, NULL, EXTENT_LOCKED | EXTENT_DO_ACCOUNTING | + EXTENT_DELALLOC | EXTENT_DEFRAG, PAGE_UNLOCK | + PAGE_CLEAR_DIRTY | PAGE_SET_WRITEBACK | + PAGE_END_WRITEBACK | PAGE_SET_ERROR); + return ret; +} + /* * work queue call back to started compression on a file and pages */ @@ -1113,11 +1250,17 @@ static noinline void async_cow_start(struct btrfs_work *work) { struct async_cow *async_cow; int num_added = 0; + int ret = 0; async_cow = container_of(work, struct async_cow, work); - compress_file_range(async_cow->inode, async_cow->locked_page, - async_cow->start, async_cow->end, async_cow, - &num_added); + if (inode_need_compress(async_cow->inode)) + compress_file_range(async_cow->inode, async_cow->locked_page, + async_cow->start, async_cow->end, async_cow, + &num_added); + else + ret = hash_file_ranges(async_cow->inode, async_cow->start, + async_cow->end, async_cow, &num_added); + if (num_added == 0) { btrfs_add_delayed_iput(async_cow->inode); async_cow->inode = NULL; @@ -1166,6 +1309,8 @@ static int cow_file_range_async(struct inode *inode, struct page *locked_page, { struct async_cow *async_cow; struct btrfs_root *root = BTRFS_I(inode)->root; + struct btrfs_fs_info *fs_info = root->fs_info; + struct btrfs_dedupe_info *dedupe_info = fs_info->dedupe_info; unsigned long nr_pages; u64 cur_end; int limit = 10 * SZ_1M; @@ -1180,11 +1325,16 @@ static int cow_file_range_async(struct inode *inode, struct page *locked_page, async_cow->locked_page = locked_page; async_cow->start = start; - if (BTRFS_I(inode)->flags & BTRFS_INODE_NOCOMPRESS && - !btrfs_test_opt(root->fs_info, FORCE_COMPRESS)) + if (fs_info->dedupe_enabled && dedupe_info) { + u64 len = max_t(u64, SZ_512K, dedupe_info->blocksize); + + cur_end = min(end, start + len - 1); + } else if (BTRFS_I(inode)->flags & BTRFS_INODE_NOCOMPRESS && + !btrfs_test_opt(root->fs_info, FORCE_COMPRESS)) { cur_end = end; - else + } else { cur_end = min(end, start + SZ_512K - 1); + } async_cow->end = cur_end; INIT_LIST_HEAD(&async_cow->extents); @@ -1585,6 +1735,8 @@ static int run_delalloc_range(struct inode *inode, struct page *locked_page, int force_cow = need_force_cow(inode, start, end); struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree; int need_compress; + struct btrfs_root *root = BTRFS_I(inode)->root; + struct btrfs_fs_info *fs_info = root->fs_info; need_compress = test_range_bit(io_tree, start, end, EXTENT_COMPRESS, 1, NULL); @@ -1600,7 +1752,7 @@ static int run_delalloc_range(struct inode *inode, struct page *locked_page, ret = run_delalloc_nocow(inode, locked_page, start, end, page_started, 0, nr_written); - } else if (!need_compress) { + } else if (!need_compress && !fs_info->dedupe_enabled) { ret = cow_file_range(inode, locked_page, start, end, end, page_started, nr_written, 1, NULL); } else { @@ -2245,7 +2397,8 @@ static int insert_reserved_file_extent(struct btrfs_trans_handle *trans, u64 disk_bytenr, u64 disk_num_bytes, u64 num_bytes, u64 ram_bytes, u8 compression, u8 encryption, - u16 other_encoding, int extent_type) + u16 other_encoding, int extent_type, + struct btrfs_dedupe_hash *hash) { struct btrfs_root *root = BTRFS_I(inode)->root; struct btrfs_file_extent_item *fi; @@ -2307,10 +2460,43 @@ static int insert_reserved_file_extent(struct btrfs_trans_handle *trans, ins.objectid = disk_bytenr; ins.offset = disk_num_bytes; ins.type = BTRFS_EXTENT_ITEM_KEY; - ret = btrfs_alloc_reserved_file_extent(trans, root, + + /* + * Only for no-dedupe or hash miss case, we need to increase + * extent reference + * For hash hit case, reference is already increased + */ + if (!hash || hash->bytenr == 0) + ret = btrfs_alloc_reserved_file_extent(trans, root, root->root_key.objectid, btrfs_ino(inode), file_pos, ram_bytes, &ins); + if (ret < 0) + goto out_qgroup; + + /* + * Hash hit won't create a new data extent, so its reserved + * space won't be freed by new delayed_ref_head. + * Need to free it here. + */ + if (btrfs_dedupe_hash_hit(hash)) + btrfs_free_reserved_data_space(inode, file_pos, ram_bytes); + + /* Add missed hash into dedupe tree */ + if (hash && hash->bytenr == 0) { + hash->bytenr = ins.objectid; + hash->num_bytes = ins.offset; + + /* + * Here we ignore dedupe_add error, as even it failed, + * it won't corrupt the filesystem. It will only only slightly + * reduce dedup rate + */ + btrfs_dedupe_add(trans, root->fs_info, hash); + } + +out_qgroup: + /* * Release the reserved range from inode dirty range map, as it is * already moved into delayed_ref_head @@ -3002,6 +3188,7 @@ static int btrfs_finish_ordered_io(struct btrfs_ordered_extent *ordered_extent) bool nolock; bool truncated = false; enum btrfs_metadata_reserve_type reserve_type = BTRFS_RESERVE_NORMAL; + int hash_hit = btrfs_dedupe_hash_hit(ordered_extent->hash); nolock = btrfs_is_free_space_inode(inode); @@ -3098,8 +3285,10 @@ static int btrfs_finish_ordered_io(struct btrfs_ordered_extent *ordered_extent) ordered_extent->disk_len, logical_len, logical_len, compress_type, 0, 0, - BTRFS_FILE_EXTENT_REG); - if (!ret) + BTRFS_FILE_EXTENT_REG, + ordered_extent->hash); + /* Hash hit case doesn't reserve delalloc bytes */ + if (!ret && !hash_hit) btrfs_release_delalloc_bytes(root, ordered_extent->start, ordered_extent->disk_len); @@ -3151,15 +3340,17 @@ out: * wrong we need to return the space for this ordered extent * back to the allocator. We only free the extent in the * truncated case if we didn't write out the extent at all. + * + * For hash hit case, never free that extent, as it's being used + * by others. */ - if ((ret || !logical_len) && + if ((ret || !logical_len) && !hash_hit && !test_bit(BTRFS_ORDERED_NOCOW, &ordered_extent->flags) && !test_bit(BTRFS_ORDERED_PREALLOC, &ordered_extent->flags)) btrfs_free_reserved_extent(root, ordered_extent->start, ordered_extent->disk_len, 1); } - /* * This needs to be done to make sure anybody waiting knows we are done * updating everything for this ordered extent. @@ -10513,7 +10704,8 @@ static int __btrfs_prealloc_file_range(struct inode *inode, int mode, cur_offset, ins.objectid, ins.offset, ins.offset, ins.offset, 0, 0, 0, - BTRFS_FILE_EXTENT_PREALLOC); + BTRFS_FILE_EXTENT_PREALLOC, + NULL); if (ret) { btrfs_free_reserved_extent(root, ins.objectid, ins.offset, 0); diff --git a/fs/btrfs/relocation.c b/fs/btrfs/relocation.c index db90b47..2f10441 100644 --- a/fs/btrfs/relocation.c +++ b/fs/btrfs/relocation.c @@ -32,6 +32,7 @@ #include "free-space-cache.h" #include "inode-map.h" #include "qgroup.h" +#include "dedupe.h" /* * backref_node, mapping_node and tree_block start with this @@ -4039,6 +4040,7 @@ static noinline_for_stack int relocate_block_group(struct reloc_control *rc) struct btrfs_trans_handle *trans = NULL; struct btrfs_path *path; struct btrfs_extent_item *ei; + struct btrfs_fs_info *fs_info = rc->extent_root->fs_info; u64 flags; u32 item_size; int ret; @@ -4161,6 +4163,20 @@ restart: rc->search_start = key.objectid; } } + /* + * This data extent will be replaced, but normal dedupe_del() + * will only happen at run_delayed_ref() time, which is too + * late, so delete dedupe_hash early to prevent its ref get + * increased during relocation + */ + if (rc->stage == MOVE_DATA_EXTENTS && + (flags & BTRFS_EXTENT_FLAG_DATA)) { + ret = btrfs_dedupe_del(trans, fs_info, key.objectid); + if (ret < 0) { + err = ret; + break; + } + } btrfs_end_transaction_throttle(trans, rc->extent_root); btrfs_btree_balance_dirty(rc->extent_root);