Hi Yunlei, When testing the original patch, I've got some panics from checkpoint. I've fixed them, and revised your patch like below. Please take a look at below.
>From a0e4f7bfdd5665643c5323c480b95a852d141a2b Mon Sep 17 00:00:00 2001 From: Yunlei He <[email protected]> Date: Mon, 13 Mar 2017 20:22:18 +0800 Subject: [PATCH] f2fs: allow write page cache when writting cp This patch allow write data to normal file when writting new checkpoint. We relax three limitations for write_begin path: 1. data allocation 2. node allocation 3. variables in checkpoint Signed-off-by: Yunlei He <[email protected]> Signed-off-by: Jaegeuk Kim <[email protected]> --- fs/f2fs/checkpoint.c | 31 +++++++++++++++++++++++-------- fs/f2fs/data.c | 28 ++++++++++++++++++++++------ fs/f2fs/f2fs.h | 1 + fs/f2fs/node.c | 12 ++++++------ fs/f2fs/super.c | 1 + 5 files changed, 53 insertions(+), 20 deletions(-) diff --git a/fs/f2fs/checkpoint.c b/fs/f2fs/checkpoint.c index 22348c7df67d..bd755468915b 100644 --- a/fs/f2fs/checkpoint.c +++ b/fs/f2fs/checkpoint.c @@ -942,6 +942,19 @@ int f2fs_sync_inode_meta(struct f2fs_sb_info *sbi) return 0; } +static void __prepare_cp_block(struct f2fs_sb_info *sbi) +{ + struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); + struct f2fs_nm_info *nm_i = NM_I(sbi); + nid_t last_nid = nm_i->next_scan_nid; + + next_free_nid(sbi, &last_nid); + ckpt->valid_block_count = cpu_to_le64(valid_user_blocks(sbi)); + ckpt->valid_node_count = cpu_to_le32(valid_node_count(sbi)); + ckpt->valid_inode_count = cpu_to_le32(valid_inode_count(sbi)); + ckpt->next_free_nid = cpu_to_le32(last_nid); +} + /* * Freeze all the FS-operations for checkpoint. */ @@ -980,6 +993,8 @@ static int block_operations(struct f2fs_sb_info *sbi) * POR: we should ensure that there are no dirty node pages * until finishing nat/sit flush. */ + down_write(&sbi->node_change); + retry_flush_nodes: down_write(&sbi->node_write); @@ -987,11 +1002,19 @@ static int block_operations(struct f2fs_sb_info *sbi) up_write(&sbi->node_write); err = sync_node_pages(sbi, &wbc); if (err) { + up_write(&sbi->node_change); f2fs_unlock_all(sbi); goto out; } goto retry_flush_nodes; } + + /* + * sbi->node_change is used only for AIO write_begin path which produces + * dirty node blocks and some checkpoint values by block allocation. + */ + __prepare_cp_block(sbi); + up_write(&sbi->node_change); out: blk_finish_plug(&plug); return err; @@ -1059,7 +1082,6 @@ static int do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) struct f2fs_checkpoint *ckpt = F2FS_CKPT(sbi); struct f2fs_nm_info *nm_i = NM_I(sbi); unsigned long orphan_num = sbi->im[ORPHAN_INO].ino_num; - nid_t last_nid = nm_i->next_scan_nid; block_t start_blk; unsigned int data_sum_blocks, orphan_blocks; __u32 crc32 = 0; @@ -1076,14 +1098,11 @@ static int do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) return -EIO; } - next_free_nid(sbi, &last_nid); - /* * modify checkpoint * version number is already updated */ ckpt->elapsed_time = cpu_to_le64(get_mtime(sbi)); - ckpt->valid_block_count = cpu_to_le64(valid_user_blocks(sbi)); ckpt->free_segment_count = cpu_to_le32(free_segments(sbi)); for (i = 0; i < NR_CURSEG_NODE_TYPE; i++) { ckpt->cur_node_segno[i] = @@ -1102,10 +1121,6 @@ static int do_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc) curseg_alloc_type(sbi, i + CURSEG_HOT_DATA); } - ckpt->valid_node_count = cpu_to_le32(valid_node_count(sbi)); - ckpt->valid_inode_count = cpu_to_le32(valid_inode_count(sbi)); - ckpt->next_free_nid = cpu_to_le32(last_nid); - /* 2 cp + n data seg summary + orphan inode blocks */ data_sum_blocks = npages_for_summary_flush(sbi, false); spin_lock(&sbi->cp_lock); diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c index 2ab5ca06c334..e07c60fb1033 100644 --- a/fs/f2fs/data.c +++ b/fs/f2fs/data.c @@ -786,6 +786,21 @@ int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *from) return err; } +static inline void __do_map_lock(struct f2fs_sb_info *sbi, int flag, bool lock) +{ + if (flag == F2FS_GET_BLOCK_PRE_AIO) { + if (lock) + down_read(&sbi->node_change); + else + up_read(&sbi->node_change); + } else { + if (lock) + f2fs_lock_op(sbi); + else + f2fs_unlock_op(sbi); + } +} + /* * f2fs_map_blocks() now supported readahead/bmap/rw direct_IO with * f2fs_map_blocks structure. @@ -828,7 +843,7 @@ int f2fs_map_blocks(struct inode *inode, struct f2fs_map_blocks *map, next_dnode: if (create) - f2fs_lock_op(sbi); + __do_map_lock(sbi, flag, true); /* When reading holes, we need its node page */ set_new_dnode(&dn, inode, NULL, NULL, 0); @@ -938,7 +953,7 @@ int f2fs_map_blocks(struct inode *inode, struct f2fs_map_blocks *map, f2fs_put_dnode(&dn); if (create) { - f2fs_unlock_op(sbi); + __do_map_lock(sbi, flag, false); f2fs_balance_fs(sbi, dn.node_changed); } goto next_dnode; @@ -947,7 +962,7 @@ int f2fs_map_blocks(struct inode *inode, struct f2fs_map_blocks *map, f2fs_put_dnode(&dn); unlock_out: if (create) { - f2fs_unlock_op(sbi); + __do_map_lock(sbi, flag, false); f2fs_balance_fs(sbi, dn.node_changed); } out: @@ -1687,7 +1702,7 @@ static int prepare_write_begin(struct f2fs_sb_info *sbi, if (f2fs_has_inline_data(inode) || (pos & PAGE_MASK) >= i_size_read(inode)) { - f2fs_lock_op(sbi); + __do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, true); locked = true; } restart: @@ -1723,7 +1738,8 @@ static int prepare_write_begin(struct f2fs_sb_info *sbi, err = get_dnode_of_data(&dn, index, LOOKUP_NODE); if (err || dn.data_blkaddr == NULL_ADDR) { f2fs_put_dnode(&dn); - f2fs_lock_op(sbi); + __do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, + true); locked = true; goto restart; } @@ -1737,7 +1753,7 @@ static int prepare_write_begin(struct f2fs_sb_info *sbi, f2fs_put_dnode(&dn); unlock_out: if (locked) - f2fs_unlock_op(sbi); + __do_map_lock(sbi, F2FS_GET_BLOCK_PRE_AIO, false); return err; } diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h index f2ad3ab98c8f..84161b2160b9 100644 --- a/fs/f2fs/f2fs.h +++ b/fs/f2fs/f2fs.h @@ -829,6 +829,7 @@ struct f2fs_sb_info { struct mutex cp_mutex; /* checkpoint procedure lock */ struct rw_semaphore cp_rwsem; /* blocking FS operations */ struct rw_semaphore node_write; /* locking node writes */ + struct rw_semaphore node_change; /* locking node change */ wait_queue_head_t cp_wait; unsigned long last_time[MAX_TIME]; /* to store time in jiffies */ long interval_time[MAX_TIME]; /* to store thresholds */ diff --git a/fs/f2fs/node.c b/fs/f2fs/node.c index 3bfffd744f87..cc7ef45b3f8d 100644 --- a/fs/f2fs/node.c +++ b/fs/f2fs/node.c @@ -2420,10 +2420,11 @@ static void __flush_nat_entry_set(struct f2fs_sb_info *sbi, f2fs_put_page(page, 1); } - f2fs_bug_on(sbi, set->entry_cnt); - - radix_tree_delete(&NM_I(sbi)->nat_set_root, set->set); - kmem_cache_free(nat_entry_set_slab, set); + /* Allow dirty nats by node block allocation in write_begin */ + if (!set->entry_cnt) { + radix_tree_delete(&NM_I(sbi)->nat_set_root, set->set); + kmem_cache_free(nat_entry_set_slab, set); + } } /* @@ -2468,8 +2469,7 @@ void flush_nat_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc) __flush_nat_entry_set(sbi, set, cpc); up_write(&nm_i->nat_tree_lock); - - f2fs_bug_on(sbi, nm_i->dirty_nat_cnt); + /* Allow dirty nats by node block allocation in write_begin */ } static int __get_nat_bitmaps(struct f2fs_sb_info *sbi) diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c index 778b6c449d47..f86a2021c2d1 100644 --- a/fs/f2fs/super.c +++ b/fs/f2fs/super.c @@ -1920,6 +1920,7 @@ static int f2fs_fill_super(struct super_block *sb, void *data, int silent) mutex_init(&sbi->gc_mutex); mutex_init(&sbi->cp_mutex); init_rwsem(&sbi->node_write); + init_rwsem(&sbi->node_change); /* disallow all the data/node/meta page writes */ set_sbi_flag(sbi, SBI_POR_DOING); -- 2.11.0 ------------------------------------------------------------------------------ Check out the vibrant tech community on one of the world's most engaging tech sites, Slashdot.org! http://sdm.link/slashdot _______________________________________________ Linux-f2fs-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
