From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org X-Spam-Level: X-Spam-Status: No, score=-2.6 required=3.0 tests=DKIM_SIGNED,DKIM_VALID, DKIM_VALID_AU,MAILING_LIST_MULTI,SPF_PASS,T_DKIMWL_WL_HIGH,USER_AGENT_MUTT autolearn=ham autolearn_force=no version=3.4.0 Received: from mail.kernel.org (mail.kernel.org [198.145.29.99]) by smtp.lore.kernel.org (Postfix) with ESMTP id D5DF8C43142 for ; Sun, 29 Jul 2018 01:44:23 +0000 (UTC) Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by mail.kernel.org (Postfix) with ESMTP id 59C7120893 for ; Sun, 29 Jul 2018 01:44:23 +0000 (UTC) Authentication-Results: mail.kernel.org; dkim=pass (1024-bit key) header.d=kernel.org header.i=@kernel.org header.b="nIaBgAHH" DMARC-Filter: OpenDMARC Filter v1.3.2 mail.kernel.org 59C7120893 Authentication-Results: mail.kernel.org; dmarc=fail (p=none dis=none) header.from=kernel.org Authentication-Results: mail.kernel.org; spf=none smtp.mailfrom=linux-kernel-owner@vger.kernel.org Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1731816AbeG2DCZ (ORCPT ); Sat, 28 Jul 2018 23:02:25 -0400 Received: from mail.kernel.org ([198.145.29.99]:53782 "EHLO mail.kernel.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1731693AbeG2DCY (ORCPT ); Sat, 28 Jul 2018 23:02:24 -0400 Received: from localhost (c-67-160-202-76.hsd1.ca.comcast.net [67.160.202.76]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by mail.kernel.org (Postfix) with ESMTPSA id 46739204EC; Sun, 29 Jul 2018 01:33:52 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=kernel.org; s=default; t=1532828032; bh=6dRghJk+5M4ZnLYhnmZe2xrup9Js1aaUlaLrdtO2pX8=; h=Date:From:To:Cc:Subject:References:In-Reply-To:From; b=nIaBgAHHN8ZqffGI6QEDqqqGB9tiNQCUXSEMQjsZ1upPNB/09eSMrsMEGMLBHoiQ7 SsZhw7aheDuS7lvcc4dGBjjPV/eb6PJoxI+3bgNOW2drTGuFiF7U8LCsfDD5n57cae 0v3iiUGsyIMMZEIjU/NDaHiYnEsBHkL+rLWGYwoA= Date: Sat, 28 Jul 2018 18:33:51 -0700 From: Jaegeuk Kim To: Chao Yu Cc: linux-f2fs-devel@lists.sourceforge.net, linux-kernel@vger.kernel.org, chao@kernel.org Subject: Re: [PATCH v5 1/2] f2fs: fix to avoid broken of dnode block list Message-ID: <20180729013351.GE83620@jaegeuk-macbookpro.roam.corp.google.com> References: <20180728013613.91304-1-yuchao0@huawei.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20180728013613.91304-1-yuchao0@huawei.com> User-Agent: Mutt/1.8.2 (2017-04-18) Sender: linux-kernel-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On 07/28, Chao Yu wrote: > f2fs recovery flow is relying on dnode block link list, it means fsynced > file recovery depends on previous dnode's persistence in the list, so > during fsync() we should wait on all regular inode's dnode writebacked > before issuing flush. > > By this way, we can avoid dnode block list being broken by out-of-order > IO submission due to IO scheduler or driver. > > Sheng Yong helps to do the test with this patch: > > Target:/data (f2fs, -) > 64MB / 32768KB / 4KB / 8 > > 1 / PERSIST / Index > > Base: > SEQ-RD(MB/s) SEQ-WR(MB/s) RND-RD(IOPS) RND-WR(IOPS) Insert(TPS) Update(TPS) Delete(TPS) > 1 867.82 204.15 41440.03 41370.54 680.8 1025.94 1031.08 > 2 871.87 205.87 41370.3 40275.2 791.14 1065.84 1101.7 > 3 866.52 205.69 41795.67 40596.16 694.69 1037.16 1031.48 > Avg 868.7366667 205.2366667 41535.33333 40747.3 722.21 1042.98 1054.753333 > > After: > SEQ-RD(MB/s) SEQ-WR(MB/s) RND-RD(IOPS) RND-WR(IOPS) Insert(TPS) Update(TPS) Delete(TPS) > 1 798.81 202.5 41143 40613.87 602.71 838.08 913.83 > 2 805.79 206.47 40297.2 41291.46 604.44 840.75 924.27 > 3 814.83 206.17 41209.57 40453.62 602.85 834.66 927.91 > Avg 806.4766667 205.0466667 40883.25667 40786.31667 603.3333333 837.83 922.0033333 > > Patched/Original: > 0.928332713 0.999074239 0.984300676 1.000957528 0.835398753 0.803303994 0.874141189 I expect Sheng will provide more test results tho, at least it seems SEQ-RD in Base shows better than two After results, even if it doesn't matter with the issue. Please confirm it first in order for anybody to say there is no regression. > > It looks like atomic write will suffer performance regression. > > I suspect that the criminal is that we forcing to wait all dnode being in > storage cache before we issue PREFLUSH+FUA. > > BTW, will commit ("f2fs: don't need to wait for node writes for atomic write") > cause the problem: we will lose data of last transaction after SPO, even if > atomic write return no error: > > - atomic_open(); > - write() P1, P2, P3; > - atomic_commit(); > - writeback data: P1, P2, P3; > - writeback node: N1, N2, N3; <--- If N1, N2 is not writebacked, N3 with fsync_mark is > writebacked, In SPOR, we won't find N3 since node chain is broken, turns out that losing > last transaction. > - preflush + fua; > - power-cut > > If we don't wait dnode writeback for atomic_write: > > SEQ-RD(MB/s) SEQ-WR(MB/s) RND-RD(IOPS) RND-WR(IOPS) Insert(TPS) Update(TPS) Delete(TPS) > 1 779.91 206.03 41621.5 40333.16 716.9 1038.21 1034.85 > 2 848.51 204.35 40082.44 39486.17 791.83 1119.96 1083.77 > 3 772.12 206.27 41335.25 41599.65 723.29 1055.07 971.92 > Avg 800.18 205.55 41013.06333 40472.99333 744.0066667 1071.08 1030.18 > > Patched/Original: > 0.92108464 1.001526693 0.987425886 0.993268102 1.030180511 1.026942031 0.976702294 > > SQLite's performance recovers. > > Jaegeuk: > "Practically, I don't see db corruption becase of this. We can excuse to lose > the last transaction." > > Finally, we decide to keep original implementation of atomic write interface > sematics that we don't wait all dnode writeback before preflush+fua submission. > > Tested-by: Sheng Yong > Signed-off-by: Chao Yu > --- > v5: > - add missing Tested-by. > - fix f2fs_reset_fsync_node_info() to reset sbi->fsync_seg_id correctly. > fs/f2fs/checkpoint.c | 2 + > fs/f2fs/data.c | 2 + > fs/f2fs/f2fs.h | 21 ++++++- > fs/f2fs/file.c | 5 +- > fs/f2fs/node.c | 144 +++++++++++++++++++++++++++++++++++-------- > fs/f2fs/super.c | 4 ++ > 6 files changed, 150 insertions(+), 28 deletions(-) > > diff --git a/fs/f2fs/checkpoint.c b/fs/f2fs/checkpoint.c > index 581710760ba6..2136430f9f0d 100644 > --- a/fs/f2fs/checkpoint.c > +++ b/fs/f2fs/checkpoint.c > @@ -1410,6 +1410,8 @@ static int do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) > > f2fs_release_ino_entry(sbi, false); > > + f2fs_reset_fsync_node_info(sbi); > + > clear_sbi_flag(sbi, SBI_IS_DIRTY); > clear_sbi_flag(sbi, SBI_NEED_CP); > __set_cp_next_pack(sbi); > diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c > index 6b8ca5011bfd..572c91e43337 100644 > --- a/fs/f2fs/data.c > +++ b/fs/f2fs/data.c > @@ -177,6 +177,8 @@ static void f2fs_write_end_io(struct bio *bio) > page->index != nid_of_node(page)); > > dec_page_count(sbi, type); > + if (f2fs_in_warm_node_list(sbi, page)) > + f2fs_del_fsync_node_entry(sbi, page); > clear_cold_data(page); > end_page_writeback(page); > } > diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h > index 0374f069520c..6627fef9ae38 100644 > --- a/fs/f2fs/f2fs.h > +++ b/fs/f2fs/f2fs.h > @@ -228,6 +228,12 @@ struct inode_entry { > struct inode *inode; /* vfs inode pointer */ > }; > > +struct fsync_node_entry { > + struct list_head list; /* list head */ > + struct page *page; /* warm node page pointer */ > + unsigned int seq_id; /* sequence id */ > +}; > + > /* for the bitmap indicate blocks to be discarded */ > struct discard_entry { > struct list_head list; /* list head */ > @@ -1156,6 +1162,11 @@ struct f2fs_sb_info { > > struct inode_management im[MAX_INO_ENTRY]; /* manage inode cache */ > > + spinlock_t fsync_node_lock; /* for node entry lock */ > + struct list_head fsync_node_list; /* node list head */ > + unsigned int fsync_seg_id; /* sequence id */ > + unsigned int fsync_node_num; /* number of node entries */ > + > /* for orphan inode, use 0'th array */ > unsigned int max_orphans; /* max orphan inodes */ > > @@ -2822,6 +2833,10 @@ struct node_info; > > int f2fs_check_nid_range(struct f2fs_sb_info *sbi, nid_t nid); > bool f2fs_available_free_memory(struct f2fs_sb_info *sbi, int type); > +bool f2fs_in_warm_node_list(struct f2fs_sb_info *sbi, struct page *page); > +void f2fs_init_fsync_node_info(struct f2fs_sb_info *sbi); > +void f2fs_del_fsync_node_entry(struct f2fs_sb_info *sbi, struct page *page); > +void f2fs_reset_fsync_node_info(struct f2fs_sb_info *sbi); > int f2fs_need_dentry_mark(struct f2fs_sb_info *sbi, nid_t nid); > bool f2fs_is_checkpointed_node(struct f2fs_sb_info *sbi, nid_t nid); > bool f2fs_need_inode_block_update(struct f2fs_sb_info *sbi, nid_t ino); > @@ -2831,7 +2846,8 @@ pgoff_t f2fs_get_next_page_offset(struct dnode_of_data *dn, pgoff_t pgofs); > int f2fs_get_dnode_of_data(struct dnode_of_data *dn, pgoff_t index, int mode); > int f2fs_truncate_inode_blocks(struct inode *inode, pgoff_t from); > int f2fs_truncate_xattr_node(struct inode *inode); > -int f2fs_wait_on_node_pages_writeback(struct f2fs_sb_info *sbi, nid_t ino); > +int f2fs_wait_on_node_pages_writeback(struct f2fs_sb_info *sbi, > + unsigned int seq_id); > int f2fs_remove_inode_page(struct inode *inode); > struct page *f2fs_new_inode_page(struct inode *inode); > struct page *f2fs_new_node_page(struct dnode_of_data *dn, unsigned int ofs); > @@ -2840,7 +2856,8 @@ struct page *f2fs_get_node_page(struct f2fs_sb_info *sbi, pgoff_t nid); > struct page *f2fs_get_node_page_ra(struct page *parent, int start); > void f2fs_move_node_page(struct page *node_page, int gc_type); > int f2fs_fsync_node_pages(struct f2fs_sb_info *sbi, struct inode *inode, > - struct writeback_control *wbc, bool atomic); > + struct writeback_control *wbc, bool atomic, > + unsigned int *seq_id); > int f2fs_sync_node_pages(struct f2fs_sb_info *sbi, > struct writeback_control *wbc, > bool do_balance, enum iostat_type io_type); > diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c > index 8f9ab66858ca..7bd2412a8c37 100644 > --- a/fs/f2fs/file.c > +++ b/fs/f2fs/file.c > @@ -213,6 +213,7 @@ static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end, > .nr_to_write = LONG_MAX, > .for_reclaim = 0, > }; > + unsigned int seq_id = 0; > > if (unlikely(f2fs_readonly(inode->i_sb))) > return 0; > @@ -275,7 +276,7 @@ static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end, > } > sync_nodes: > atomic_inc(&sbi->wb_sync_req[NODE]); > - ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic); > + ret = f2fs_fsync_node_pages(sbi, inode, &wbc, atomic, &seq_id); > atomic_dec(&sbi->wb_sync_req[NODE]); > if (ret) > goto out; > @@ -301,7 +302,7 @@ static int f2fs_do_sync_file(struct file *file, loff_t start, loff_t end, > * given fsync mark. > */ > if (!atomic) { > - ret = f2fs_wait_on_node_pages_writeback(sbi, ino); > + ret = f2fs_wait_on_node_pages_writeback(sbi, seq_id); > if (ret) > goto out; > } > diff --git a/fs/f2fs/node.c b/fs/f2fs/node.c > index 9d9f4c9750c4..e109c671cd84 100644 > --- a/fs/f2fs/node.c > +++ b/fs/f2fs/node.c > @@ -28,6 +28,7 @@ > static struct kmem_cache *nat_entry_slab; > static struct kmem_cache *free_nid_slab; > static struct kmem_cache *nat_entry_set_slab; > +static struct kmem_cache *fsync_node_entry_slab; > > /* > * Check whether the given nid is within node id range. > @@ -264,6 +265,72 @@ static unsigned int __gang_lookup_nat_set(struct f2fs_nm_info *nm_i, > start, nr); > } > > +bool f2fs_in_warm_node_list(struct f2fs_sb_info *sbi, struct page *page) > +{ > + return NODE_MAPPING(sbi) == page->mapping && > + IS_DNODE(page) && is_cold_node(page); > +} > + > +void f2fs_init_fsync_node_info(struct f2fs_sb_info *sbi) > +{ > + spin_lock_init(&sbi->fsync_node_lock); > + INIT_LIST_HEAD(&sbi->fsync_node_list); > + sbi->fsync_seg_id = 0; > + sbi->fsync_node_num = 0; > +} > + > +static unsigned int f2fs_add_fsync_node_entry(struct f2fs_sb_info *sbi, > + struct page *page) > +{ > + struct fsync_node_entry *fn; > + unsigned long flags; > + unsigned int seq_id; > + > + fn = f2fs_kmem_cache_alloc(fsync_node_entry_slab, GFP_NOFS); > + > + get_page(page); > + fn->page = page; > + INIT_LIST_HEAD(&fn->list); > + > + spin_lock_irqsave(&sbi->fsync_node_lock, flags); > + list_add_tail(&fn->list, &sbi->fsync_node_list); > + fn->seq_id = sbi->fsync_seg_id++; > + seq_id = fn->seq_id; > + sbi->fsync_node_num++; > + spin_unlock_irqrestore(&sbi->fsync_node_lock, flags); > + > + return seq_id; > +} > + > +void f2fs_del_fsync_node_entry(struct f2fs_sb_info *sbi, struct page *page) > +{ > + struct fsync_node_entry *fn; > + unsigned long flags; > + > + spin_lock_irqsave(&sbi->fsync_node_lock, flags); > + list_for_each_entry(fn, &sbi->fsync_node_list, list) { > + if (fn->page == page) { > + list_del(&fn->list); > + sbi->fsync_node_num--; > + spin_unlock_irqrestore(&sbi->fsync_node_lock, flags); > + kmem_cache_free(fsync_node_entry_slab, fn); > + put_page(page); > + return; > + } > + } > + spin_unlock_irqrestore(&sbi->fsync_node_lock, flags); > + f2fs_bug_on(sbi, 1); > +} > + > +void f2fs_reset_fsync_node_info(struct f2fs_sb_info *sbi) > +{ > + unsigned long flags; > + > + spin_lock_irqsave(&sbi->fsync_node_lock, flags); > + sbi->fsync_seg_id = 0; > + spin_unlock_irqrestore(&sbi->fsync_node_lock, flags); > +} > + > int f2fs_need_dentry_mark(struct f2fs_sb_info *sbi, nid_t nid) > { > struct f2fs_nm_info *nm_i = NM_I(sbi); > @@ -1384,7 +1451,7 @@ static struct page *last_fsync_dnode(struct f2fs_sb_info *sbi, nid_t ino) > > static int __write_node_page(struct page *page, bool atomic, bool *submitted, > struct writeback_control *wbc, bool do_balance, > - enum iostat_type io_type) > + enum iostat_type io_type, unsigned int *seq_id) > { > struct f2fs_sb_info *sbi = F2FS_P_SB(page); > nid_t nid; > @@ -1401,6 +1468,7 @@ static int __write_node_page(struct page *page, bool atomic, bool *submitted, > .io_type = io_type, > .io_wbc = wbc, > }; > + unsigned int seq; > > trace_f2fs_writepage(page, NODE); > > @@ -1442,6 +1510,13 @@ static int __write_node_page(struct page *page, bool atomic, bool *submitted, > > set_page_writeback(page); > ClearPageError(page); > + > + if (f2fs_in_warm_node_list(sbi, page)) { > + seq = f2fs_add_fsync_node_entry(sbi, page); > + if (seq_id) > + *seq_id = seq; > + } > + > fio.old_blkaddr = ni.blk_addr; > f2fs_do_write_node_page(nid, &fio); > set_node_addr(sbi, &ni, fio.new_blkaddr, is_fsync_dnode(page)); > @@ -1489,7 +1564,7 @@ void f2fs_move_node_page(struct page *node_page, int gc_type) > goto out_page; > > if (__write_node_page(node_page, false, NULL, > - &wbc, false, FS_GC_NODE_IO)) > + &wbc, false, FS_GC_NODE_IO, NULL)) > unlock_page(node_page); > goto release_page; > } else { > @@ -1506,11 +1581,13 @@ void f2fs_move_node_page(struct page *node_page, int gc_type) > static int f2fs_write_node_page(struct page *page, > struct writeback_control *wbc) > { > - return __write_node_page(page, false, NULL, wbc, false, FS_NODE_IO); > + return __write_node_page(page, false, NULL, wbc, false, > + FS_NODE_IO, NULL); > } > > int f2fs_fsync_node_pages(struct f2fs_sb_info *sbi, struct inode *inode, > - struct writeback_control *wbc, bool atomic) > + struct writeback_control *wbc, bool atomic, > + unsigned int *seq_id) > { > pgoff_t index; > pgoff_t last_idx = ULONG_MAX; > @@ -1591,7 +1668,7 @@ int f2fs_fsync_node_pages(struct f2fs_sb_info *sbi, struct inode *inode, > ret = __write_node_page(page, atomic && > page == last_page, > &submitted, wbc, true, > - FS_NODE_IO); > + FS_NODE_IO, seq_id); > if (ret) { > unlock_page(page); > f2fs_put_page(last_page, 0); > @@ -1708,7 +1785,7 @@ int f2fs_sync_node_pages(struct f2fs_sb_info *sbi, > set_dentry_mark(page, 0); > > ret = __write_node_page(page, false, &submitted, > - wbc, do_balance, io_type); > + wbc, do_balance, io_type, NULL); > if (ret) > unlock_page(page); > else if (submitted) > @@ -1739,35 +1816,46 @@ int f2fs_sync_node_pages(struct f2fs_sb_info *sbi, > return ret; > } > > -int f2fs_wait_on_node_pages_writeback(struct f2fs_sb_info *sbi, nid_t ino) > +int f2fs_wait_on_node_pages_writeback(struct f2fs_sb_info *sbi, > + unsigned int seq_id) > { > - pgoff_t index = 0; > - struct pagevec pvec; > + struct fsync_node_entry *fn; > + struct page *page; > + struct list_head *head = &sbi->fsync_node_list; > + unsigned long flags; > + unsigned int cur_seq_id = 0; > int ret2, ret = 0; > - int nr_pages; > > - pagevec_init(&pvec); > + while (seq_id && cur_seq_id < seq_id) { > + spin_lock_irqsave(&sbi->fsync_node_lock, flags); > + if (list_empty(head)) { > + spin_unlock_irqrestore(&sbi->fsync_node_lock, flags); > + break; > + } > + fn = list_first_entry(head, struct fsync_node_entry, list); > + if (fn->seq_id > seq_id) { > + spin_unlock_irqrestore(&sbi->fsync_node_lock, flags); > + break; > + } > + cur_seq_id = fn->seq_id; > + page = fn->page; > + get_page(page); > + spin_unlock_irqrestore(&sbi->fsync_node_lock, flags); > > - while ((nr_pages = pagevec_lookup_tag(&pvec, NODE_MAPPING(sbi), &index, > - PAGECACHE_TAG_WRITEBACK))) { > - int i; > + f2fs_wait_on_page_writeback(page, NODE, true); > + if (TestClearPageError(page)) > + ret = -EIO; > > - for (i = 0; i < nr_pages; i++) { > - struct page *page = pvec.pages[i]; > + put_page(page); > > - if (ino && ino_of_node(page) == ino) { > - f2fs_wait_on_page_writeback(page, NODE, true); > - if (TestClearPageError(page)) > - ret = -EIO; > - } > - } > - pagevec_release(&pvec); > - cond_resched(); > + if (ret) > + break; > } > > ret2 = filemap_check_errors(NODE_MAPPING(sbi)); > if (!ret) > ret = ret2; > + > return ret; > } > > @@ -2982,8 +3070,15 @@ int __init f2fs_create_node_manager_caches(void) > sizeof(struct nat_entry_set)); > if (!nat_entry_set_slab) > goto destroy_free_nid; > + > + fsync_node_entry_slab = f2fs_kmem_cache_create("fsync_node_entry", > + sizeof(struct fsync_node_entry)); > + if (!fsync_node_entry_slab) > + goto destroy_nat_entry_set; > return 0; > > +destroy_nat_entry_set: > + kmem_cache_destroy(nat_entry_set_slab); > destroy_free_nid: > kmem_cache_destroy(free_nid_slab); > destroy_nat_entry: > @@ -2994,6 +3089,7 @@ int __init f2fs_create_node_manager_caches(void) > > void f2fs_destroy_node_manager_caches(void) > { > + kmem_cache_destroy(fsync_node_entry_slab); > kmem_cache_destroy(nat_entry_set_slab); > kmem_cache_destroy(free_nid_slab); > kmem_cache_destroy(nat_entry_slab); > diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c > index 0eff4637fe55..cb1ba9f18353 100644 > --- a/fs/f2fs/super.c > +++ b/fs/f2fs/super.c > @@ -1030,6 +1030,8 @@ static void f2fs_put_super(struct super_block *sb) > */ > f2fs_release_ino_entry(sbi, true); > > + f2fs_bug_on(sbi, sbi->fsync_node_num); > + > f2fs_leave_shrinker(sbi); > mutex_unlock(&sbi->umount_mutex); > > @@ -2923,6 +2925,8 @@ static int f2fs_fill_super(struct super_block *sb, void *data, int silent) > > f2fs_init_ino_entry_info(sbi); > > + f2fs_init_fsync_node_info(sbi); > + > /* setup f2fs internal modules */ > err = f2fs_build_segment_manager(sbi); > if (err) { > -- > 2.18.0.rc1