From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from synology.com ([59.124.61.242]:43363 "EHLO synology.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754366AbeEHKLy (ORCPT ); Tue, 8 May 2018 06:11:54 -0400 From: robbieko To: linux-btrfs@vger.kernel.org Cc: Robbie Ko Subject: [PATCH v2 2/2] btrfs: incremental send, improve rmdir performance for large directory Date: Tue, 8 May 2018 18:11:38 +0800 Message-Id: <1525774298-6919-3-git-send-email-robbieko@synology.com> In-Reply-To: <1525774298-6919-1-git-send-email-robbieko@synology.com> References: <1525774298-6919-1-git-send-email-robbieko@synology.com> Sender: linux-btrfs-owner@vger.kernel.org List-ID: From: Robbie Ko Currently when checking if we can delete a directory, we always check if all its children have been processed. Example: A directory with 2,000,000 files was deleted Result: original : 1994m57.071s patch : 1m38.554s [FIX] Instead of checking all children on all calls to can_rmdir(), we keep track of the directory index offset of the child last checked in the last call to can_rmdir(), and then use it as the starting point for future calls to can_rmdir(). Signed-off-by: Robbie Ko --- V2: fix comments split optimization allocations orphan_dir_info fs/btrfs/send.c | 31 ++++++++++++++++++++++--------- 1 file changed, 22 insertions(+), 9 deletions(-) diff --git a/fs/btrfs/send.c b/fs/btrfs/send.c index 2830871..a477268 100644 --- a/fs/btrfs/send.c +++ b/fs/btrfs/send.c @@ -247,6 +247,7 @@ struct orphan_dir_info { struct rb_node node; u64 ino; u64 gen; + u64 last_dir_index_offset; }; struct name_cache_entry { @@ -2872,6 +2873,7 @@ static int orphanize_inode(struct send_ctx *sctx, u64 ino, u64 gen, return ERR_PTR(-ENOMEM); odi->ino = dir_ino; odi->gen = 0; + odi->last_dir_index_offset = 0; rb_link_node(&odi->node, parent, p); rb_insert_color(&odi->node, &sctx->orphan_dirs); @@ -2927,6 +2929,7 @@ static int can_rmdir(struct send_ctx *sctx, u64 dir, u64 dir_gen, struct btrfs_key found_key; struct btrfs_key loc; struct btrfs_dir_item *di; + struct orphan_dir_info *odi = NULL; /* * Don't try to rmdir the top/root subvolume dir. @@ -2941,6 +2944,11 @@ static int can_rmdir(struct send_ctx *sctx, u64 dir, u64 dir_gen, key.objectid = dir; key.type = BTRFS_DIR_INDEX_KEY; key.offset = 0; + + odi = get_orphan_dir_info(sctx, dir); + if (odi) + key.offset = odi->last_dir_index_offset; + ret = btrfs_search_slot(NULL, root, &key, path, 0, 0); if (ret < 0) goto out; @@ -2968,30 +2976,33 @@ static int can_rmdir(struct send_ctx *sctx, u64 dir, u64 dir_gen, dm = get_waiting_dir_move(sctx, loc.objectid); if (dm) { - struct orphan_dir_info *odi; - odi = add_orphan_dir_info(sctx, dir); if (IS_ERR(odi)) { ret = PTR_ERR(odi); goto out; } odi->gen = dir_gen; + odi->last_dir_index_offset = found_key.offset; dm->rmdir_ino = dir; ret = 0; goto out; } if (loc.objectid > send_progress) { - struct orphan_dir_info *odi; - - odi = get_orphan_dir_info(sctx, dir); - free_orphan_dir_info(sctx, odi); + odi = add_orphan_dir_info(sctx, dir); + if (IS_ERR(odi)) { + ret = PTR_ERR(odi); + goto out; + } + odi->gen = dir_gen; + odi->last_dir_index_offset = found_key.offset; ret = 0; goto out; } path->slots[0]++; } + free_orphan_dir_info(sctx, odi); ret = 1; @@ -3269,13 +3280,16 @@ static int apply_dir_move(struct send_ctx *sctx, struct pending_dir_move *pm) if (rmdir_ino) { struct orphan_dir_info *odi; + u64 gen; odi = get_orphan_dir_info(sctx, rmdir_ino); if (!odi) { /* already deleted */ goto finish; } - ret = can_rmdir(sctx, rmdir_ino, odi->gen, sctx->cur_ino); + gen = odi->gen; + + ret = can_rmdir(sctx, rmdir_ino, gen, sctx->cur_ino); if (ret < 0) goto out; if (!ret) @@ -3286,13 +3300,12 @@ static int apply_dir_move(struct send_ctx *sctx, struct pending_dir_move *pm) ret = -ENOMEM; goto out; } - ret = get_cur_path(sctx, rmdir_ino, odi->gen, name); + ret = get_cur_path(sctx, rmdir_ino, gen, name); if (ret < 0) goto out; ret = send_rmdir(sctx, name); if (ret < 0) goto out; - free_orphan_dir_info(sctx, odi); } finish: -- 1.9.1