From 0c872e2dedfc09f41a5604d1c5010f800c0bd8f1 Mon Sep 17 00:00:00 2001 From: Chao Yu Date: Thu, 5 Feb 2015 17:46:29 +0800 Subject: f2fs: move ext_lock out of struct extent_info Move ext_lock out of struct extent_info, then in the following patches we can use variables with struct extent_info type as a parameter to pass pure data. Signed-off-by: Chao Yu Signed-off-by: Jaegeuk Kim --- fs/f2fs/inode.c | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'fs/f2fs/inode.c') diff --git a/fs/f2fs/inode.c b/fs/f2fs/inode.c index 2d002e3738a7..28dd26a50276 100644 --- a/fs/f2fs/inode.c +++ b/fs/f2fs/inode.c @@ -130,7 +130,10 @@ static int do_read_inode(struct inode *inode) fi->i_pino = le32_to_cpu(ri->i_pino); fi->i_dir_level = ri->i_dir_level; + write_lock(&fi->ext_lock); get_extent_info(&fi->ext, ri->i_ext); + write_unlock(&fi->ext_lock); + get_inline_info(fi, ri); /* check data exist */ @@ -220,7 +223,11 @@ void update_inode(struct inode *inode, struct page *node_page) ri->i_links = cpu_to_le32(inode->i_nlink); ri->i_size = cpu_to_le64(i_size_read(inode)); ri->i_blocks = cpu_to_le64(inode->i_blocks); + + read_lock(&F2FS_I(inode)->ext_lock); set_raw_extent(&F2FS_I(inode)->ext, &ri->i_ext); + read_unlock(&F2FS_I(inode)->ext_lock); + set_raw_inline(F2FS_I(inode), ri); ri->i_atime = cpu_to_le64(inode->i_atime.tv_sec); -- cgit v1.2.3 From 1dcc336b02bff3d38f173feac55a2b6c25a5fb54 Mon Sep 17 00:00:00 2001 From: Chao Yu Date: Thu, 5 Feb 2015 17:57:31 +0800 Subject: f2fs: enable rb-tree extent cache This patch enables rb-tree based extent cache in f2fs. When we mount with "-o extent_cache", f2fs will try to add recently accessed page-block mappings into rb-tree based extent cache as much as possible, instead of original one extent info cache. By this way, f2fs can support more effective cache between dnode page cache and disk. It will supply high hit ratio in the cache with fewer memory when dnode page cache are reclaimed in environment of low memory. Storage: Sandisk sd card 64g 1.append write file (offset: 0, size: 128M); 2.override write file (offset: 2M, size: 1M); 3.override write file (offset: 4M, size: 1M); ... 4.override write file (offset: 48M, size: 1M); ... 5.override write file (offset: 112M, size: 1M); 6.sync 7.echo 3 > /proc/sys/vm/drop_caches 8.read file (size:128M, unit: 4k, count: 32768) (time dd if=/mnt/f2fs/128m bs=4k count=32768) Extent Hit Ratio: before patched Hit Ratio 121 / 1071 1071 / 1071 Performance: before patched real 0m37.051s 0m35.556s user 0m0.040s 0m0.026s sys 0m2.990s 0m2.251s Memory Cost: before patched Tree Count: 0 1 (size: 24 bytes) Node Count: 0 45 (size: 1440 bytes) v3: o retest and given more details of test result. Signed-off-by: Chao Yu Signed-off-by: Jaegeuk Kim --- fs/f2fs/data.c | 13 +++++++++++++ fs/f2fs/inode.c | 1 + fs/f2fs/segment.c | 3 +++ fs/f2fs/super.c | 9 ++++++++- 4 files changed, 25 insertions(+), 1 deletion(-) (limited to 'fs/f2fs/inode.c') diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c index f52745346205..acdc0767f77c 100644 --- a/fs/f2fs/data.c +++ b/fs/f2fs/data.c @@ -666,6 +666,9 @@ void f2fs_shrink_extent_tree(struct f2fs_sb_info *sbi, int nr_shrink) void **slot; unsigned int found; + if (!test_opt(sbi, EXTENT_CACHE)) + return; + if (available_free_memory(sbi, EXTENT_CACHE)) return; @@ -714,6 +717,9 @@ void f2fs_destroy_extent_tree(struct inode *inode) struct f2fs_sb_info *sbi = F2FS_I_SB(inode); struct extent_tree *et; + if (!test_opt(sbi, EXTENT_CACHE)) + return; + down_read(&sbi->extent_tree_lock); et = radix_tree_lookup(&sbi->extent_tree_root, inode->i_ino); if (!et) { @@ -749,6 +755,9 @@ out: static bool f2fs_lookup_extent_cache(struct inode *inode, pgoff_t pgofs, struct extent_info *ei) { + if (test_opt(F2FS_I_SB(inode), EXTENT_CACHE)) + return f2fs_lookup_extent_tree(inode, pgofs, ei); + return lookup_extent_info(inode, pgofs, ei); } @@ -765,6 +774,10 @@ void f2fs_update_extent_cache(struct dnode_of_data *dn) fofs = start_bidx_of_node(ofs_of_node(dn->node_page), fi) + dn->ofs_in_node; + if (test_opt(F2FS_I_SB(dn->inode), EXTENT_CACHE)) + return f2fs_update_extent_tree(dn->inode, fofs, + dn->data_blkaddr); + if (update_extent_info(dn->inode, fofs, dn->data_blkaddr)) sync_inode_page(dn); } diff --git a/fs/f2fs/inode.c b/fs/f2fs/inode.c index 28dd26a50276..b5087443be41 100644 --- a/fs/f2fs/inode.c +++ b/fs/f2fs/inode.c @@ -335,6 +335,7 @@ void f2fs_evict_inode(struct inode *inode) no_delete: stat_dec_inline_dir(inode); stat_dec_inline_inode(inode); + f2fs_destroy_extent_tree(inode); invalidate_mapping_pages(NODE_MAPPING(sbi), inode->i_ino, inode->i_ino); if (xnid) invalidate_mapping_pages(NODE_MAPPING(sbi), xnid, xnid); diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c index daee4ab913da..29c04edcb843 100644 --- a/fs/f2fs/segment.c +++ b/fs/f2fs/segment.c @@ -277,6 +277,9 @@ void f2fs_balance_fs(struct f2fs_sb_info *sbi) void f2fs_balance_fs_bg(struct f2fs_sb_info *sbi) { + /* try to shrink extent cache when there is no enough memory */ + f2fs_shrink_extent_tree(sbi, EXTENT_CACHE_SHRINK_NUMBER); + /* check the # of cached NAT entries and prefree segments */ if (try_to_free_nats(sbi, NAT_ENTRY_PER_BLOCK) || excess_prefree_segs(sbi) || diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c index 3ce4d7c9a29d..e649f21121d2 100644 --- a/fs/f2fs/super.c +++ b/fs/f2fs/super.c @@ -1079,6 +1079,8 @@ try_onemore: INIT_LIST_HEAD(&sbi->dir_inode_list); spin_lock_init(&sbi->dir_inode_lock); + init_extent_cache_info(sbi); + init_ino_entry_info(sbi); /* setup f2fs internal modules */ @@ -1285,10 +1287,13 @@ static int __init init_f2fs_fs(void) err = create_checkpoint_caches(); if (err) goto free_segment_manager_caches; + err = create_extent_cache(); + if (err) + goto free_checkpoint_caches; f2fs_kset = kset_create_and_add("f2fs", NULL, fs_kobj); if (!f2fs_kset) { err = -ENOMEM; - goto free_checkpoint_caches; + goto free_extent_cache; } err = register_filesystem(&f2fs_fs_type); if (err) @@ -1299,6 +1304,8 @@ static int __init init_f2fs_fs(void) free_kset: kset_unregister(f2fs_kset); +free_extent_cache: + destroy_extent_cache(); free_checkpoint_caches: destroy_checkpoint_caches(); free_segment_manager_caches: -- cgit v1.2.3 From 3c6c2bebef79999b1827041696dc1881e637e3af Mon Sep 17 00:00:00 2001 From: Jaegeuk Kim Date: Tue, 17 Mar 2015 17:16:35 -0700 Subject: f2fs: avoid punch_hole overhead when releasing volatile data This patch is to avoid some punch_hole overhead when releasing volatile data. If volatile data was not written yet, we just can make the first page as zero. Signed-off-by: Jaegeuk Kim --- fs/f2fs/data.c | 2 ++ fs/f2fs/f2fs.h | 6 ++++++ fs/f2fs/file.c | 6 ++++++ fs/f2fs/inode.c | 10 ++++++++++ 4 files changed, 24 insertions(+) (limited to 'fs/f2fs/inode.c') diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c index f0a18a005bda..486113db97b1 100644 --- a/fs/f2fs/data.c +++ b/fs/f2fs/data.c @@ -1291,6 +1291,8 @@ int do_write_data_page(struct page *page, struct f2fs_io_info *fio) write_data_page(page, &dn, fio); f2fs_update_extent_cache(&dn); set_inode_flag(F2FS_I(inode), FI_APPEND_WRITE); + if (page->index == 0) + set_inode_flag(F2FS_I(inode), FI_FIRST_BLOCK_WRITTEN); } out_writepage: f2fs_put_dnode(&dn); diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h index dbe114463a18..9e378179ff93 100644 --- a/fs/f2fs/f2fs.h +++ b/fs/f2fs/f2fs.h @@ -1232,6 +1232,7 @@ enum { FI_NEED_IPU, /* used for ipu per file */ FI_ATOMIC_FILE, /* indicate atomic file */ FI_VOLATILE_FILE, /* indicate volatile file */ + FI_FIRST_BLOCK_WRITTEN, /* indicate #0 data block was written */ FI_DROP_CACHE, /* drop dirty page cache */ FI_DATA_EXIST, /* indicate data exists */ }; @@ -1340,6 +1341,11 @@ static inline bool f2fs_is_volatile_file(struct inode *inode) return is_inode_flag_set(F2FS_I(inode), FI_VOLATILE_FILE); } +static inline bool f2fs_is_first_block_written(struct inode *inode) +{ + return is_inode_flag_set(F2FS_I(inode), FI_FIRST_BLOCK_WRITTEN); +} + static inline bool f2fs_is_drop_cache(struct inode *inode) { return is_inode_flag_set(F2FS_I(inode), FI_DROP_CACHE); diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c index 36dc7581a28b..39a07a59b134 100644 --- a/fs/f2fs/file.c +++ b/fs/f2fs/file.c @@ -437,6 +437,9 @@ int truncate_data_blocks_range(struct dnode_of_data *dn, int count) dn->data_blkaddr = NULL_ADDR; f2fs_update_extent_cache(dn); invalidate_blocks(sbi, blkaddr); + if (dn->ofs_in_node == 0 && IS_INODE(dn->node_page)) + clear_inode_flag(F2FS_I(dn->inode), + FI_FIRST_BLOCK_WRITTEN); nr_free++; } if (nr_free) { @@ -1005,6 +1008,9 @@ static int f2fs_ioc_release_volatile_write(struct file *filp) if (!f2fs_is_volatile_file(inode)) return 0; + if (!f2fs_is_first_block_written(inode)) + return truncate_partial_data_page(inode, 0, true); + punch_hole(inode, 0, F2FS_BLKSIZE); return 0; } diff --git a/fs/f2fs/inode.c b/fs/f2fs/inode.c index b5087443be41..bb5800969bcc 100644 --- a/fs/f2fs/inode.c +++ b/fs/f2fs/inode.c @@ -51,6 +51,13 @@ static void __get_inode_rdev(struct inode *inode, struct f2fs_inode *ri) } } +static bool __written_first_block(struct f2fs_inode *ri) +{ + if (ri->i_addr[0] != NEW_ADDR && ri->i_addr[0] != NULL_ADDR) + return true; + return false; +} + static void __set_inode_rdev(struct inode *inode, struct f2fs_inode *ri) { if (S_ISCHR(inode->i_mode) || S_ISBLK(inode->i_mode)) { @@ -143,6 +150,9 @@ static int do_read_inode(struct inode *inode) /* get rdev by using inline_info */ __get_inode_rdev(inode, ri); + if (__written_first_block(ri)) + set_inode_flag(F2FS_I(inode), FI_FIRST_BLOCK_WRITTEN); + f2fs_put_page(node_page, 1); stat_inc_inline_inode(inode); -- cgit v1.2.3 From 028a41e89383e1208dff1afe3e260b8cb6d3431c Mon Sep 17 00:00:00 2001 From: Chao Yu Date: Thu, 19 Mar 2015 19:26:02 +0800 Subject: f2fs: initialize extent tree with on-disk extent info of inode With normal extent info cache, we records largest extent mapping between logical block and physical block into extent info, and we persist extent info in on-disk inode. When we enable extent tree cache, if extent info of on-disk inode is exist, and the extent is not a small fragmented mapping extent. We'd better to load the extent info into extent tree cache when inode is loaded. By this way we can have more chance to hit extent tree cache rather than taking more time to read dnode page for block address. Signed-off-by: Chao Yu Signed-off-by: Jaegeuk Kim --- fs/f2fs/data.c | 43 +++++++++++++++++++++++++++++++++++++++++++ fs/f2fs/f2fs.h | 1 + fs/f2fs/inode.c | 4 +--- 3 files changed, 45 insertions(+), 3 deletions(-) (limited to 'fs/f2fs/inode.c') diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c index 67b3a99e03d5..e3be4be3a6d8 100644 --- a/fs/f2fs/data.c +++ b/fs/f2fs/data.c @@ -572,6 +572,39 @@ static unsigned int __free_extent_tree(struct f2fs_sb_info *sbi, return count - et->count; } +static void f2fs_init_extent_tree(struct inode *inode, + struct f2fs_extent *i_ext) +{ + struct f2fs_sb_info *sbi = F2FS_I_SB(inode); + struct extent_tree *et; + struct extent_node *en; + struct extent_info ei; + + if (le32_to_cpu(i_ext->len) < F2FS_MIN_EXTENT_LEN) + return; + + et = __grab_extent_tree(inode); + + write_lock(&et->lock); + if (et->count) + goto out; + + set_extent_info(&ei, le32_to_cpu(i_ext->fofs), + le32_to_cpu(i_ext->blk), le32_to_cpu(i_ext->len)); + + en = __insert_extent_tree(sbi, et, &ei, NULL); + if (en) { + et->cached_en = en; + + spin_lock(&sbi->extent_lock); + list_add_tail(&en->list, &sbi->extent_list); + spin_unlock(&sbi->extent_lock); + } +out: + write_unlock(&et->lock); + atomic_dec(&et->refcount); +} + static bool f2fs_lookup_extent_tree(struct inode *inode, pgoff_t pgofs, struct extent_info *ei) { @@ -782,6 +815,16 @@ out: return; } +void f2fs_init_extent_cache(struct inode *inode, struct f2fs_extent *i_ext) +{ + if (test_opt(F2FS_I_SB(inode), EXTENT_CACHE)) + f2fs_init_extent_tree(inode, i_ext); + + write_lock(&F2FS_I(inode)->ext_lock); + get_extent_info(&F2FS_I(inode)->ext, *i_ext); + write_unlock(&F2FS_I(inode)->ext_lock); +} + static bool f2fs_lookup_extent_cache(struct inode *inode, pgoff_t pgofs, struct extent_info *ei) { diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h index 83223d8b32bb..431b4af693e5 100644 --- a/fs/f2fs/f2fs.h +++ b/fs/f2fs/f2fs.h @@ -1596,6 +1596,7 @@ int reserve_new_block(struct dnode_of_data *); int f2fs_reserve_block(struct dnode_of_data *, pgoff_t); void f2fs_shrink_extent_tree(struct f2fs_sb_info *, int); void f2fs_destroy_extent_tree(struct inode *); +void f2fs_init_extent_cache(struct inode *, struct f2fs_extent *); void f2fs_update_extent_cache(struct dnode_of_data *); struct page *find_data_page(struct inode *, pgoff_t, bool); struct page *get_lock_data_page(struct inode *, pgoff_t); diff --git a/fs/f2fs/inode.c b/fs/f2fs/inode.c index bb5800969bcc..9e64d3efb2d2 100644 --- a/fs/f2fs/inode.c +++ b/fs/f2fs/inode.c @@ -137,9 +137,7 @@ static int do_read_inode(struct inode *inode) fi->i_pino = le32_to_cpu(ri->i_pino); fi->i_dir_level = ri->i_dir_level; - write_lock(&fi->ext_lock); - get_extent_info(&fi->ext, ri->i_ext); - write_unlock(&fi->ext_lock); + f2fs_init_extent_cache(inode, &ri->i_ext); get_inline_info(fi, ri); -- cgit v1.2.3 From 0bdee482509fe8c3cf0e66231ed37b8e70954093 Mon Sep 17 00:00:00 2001 From: Chao Yu Date: Thu, 19 Mar 2015 19:27:51 +0800 Subject: f2fs: preserve extent info for extent cache This patch tries to preserve last extent info in extent tree cache into on-disk inode, so this can help us to reuse the last extent info next time for performance. Signed-off-by: Chao Yu Signed-off-by: Jaegeuk Kim --- fs/f2fs/data.c | 49 +++++++++++++++++++++++++++++++++++++++++++++++++ fs/f2fs/f2fs.h | 8 ++++++++ fs/f2fs/inode.c | 5 +++++ 3 files changed, 62 insertions(+) (limited to 'fs/f2fs/inode.c') diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c index e3be4be3a6d8..8a940e518be8 100644 --- a/fs/f2fs/data.c +++ b/fs/f2fs/data.c @@ -719,6 +719,55 @@ update_extent: atomic_dec(&et->refcount); } +void f2fs_preserve_extent_tree(struct inode *inode) +{ + struct extent_tree *et; + struct extent_info *ext = &F2FS_I(inode)->ext; + bool sync = false; + + if (!test_opt(F2FS_I_SB(inode), EXTENT_CACHE)) + return; + + et = __find_extent_tree(F2FS_I_SB(inode), inode->i_ino); + if (!et) { + if (ext->len) { + ext->len = 0; + update_inode_page(inode); + } + return; + } + + read_lock(&et->lock); + if (et->count) { + struct extent_node *en; + + if (et->cached_en) { + en = et->cached_en; + } else { + struct rb_node *node = rb_first(&et->root); + + if (!node) + node = rb_last(&et->root); + en = rb_entry(node, struct extent_node, rb_node); + } + + if (__is_extent_same(ext, &en->ei)) + goto out; + + *ext = en->ei; + sync = true; + } else if (ext->len) { + ext->len = 0; + sync = true; + } +out: + read_unlock(&et->lock); + atomic_dec(&et->refcount); + + if (sync) + update_inode_page(inode); +} + void f2fs_shrink_extent_tree(struct f2fs_sb_info *sbi, int nr_shrink) { struct extent_tree *treevec[EXT_TREE_VEC_SIZE]; diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h index 431b4af693e5..0be198339c3b 100644 --- a/fs/f2fs/f2fs.h +++ b/fs/f2fs/f2fs.h @@ -371,6 +371,13 @@ static inline void set_extent_info(struct extent_info *ei, unsigned int fofs, ei->len = len; } +static inline bool __is_extent_same(struct extent_info *ei1, + struct extent_info *ei2) +{ + return (ei1->fofs == ei2->fofs && ei1->blk == ei2->blk && + ei1->len == ei2->len); +} + static inline bool __is_extent_mergeable(struct extent_info *back, struct extent_info *front) { @@ -1598,6 +1605,7 @@ void f2fs_shrink_extent_tree(struct f2fs_sb_info *, int); void f2fs_destroy_extent_tree(struct inode *); void f2fs_init_extent_cache(struct inode *, struct f2fs_extent *); void f2fs_update_extent_cache(struct dnode_of_data *); +void f2fs_preserve_extent_tree(struct inode *); struct page *find_data_page(struct inode *, pgoff_t, bool); struct page *get_lock_data_page(struct inode *, pgoff_t); struct page *get_new_data_page(struct inode *, struct page *, pgoff_t, bool); diff --git a/fs/f2fs/inode.c b/fs/f2fs/inode.c index 9e64d3efb2d2..07237ac2935e 100644 --- a/fs/f2fs/inode.c +++ b/fs/f2fs/inode.c @@ -343,7 +343,12 @@ void f2fs_evict_inode(struct inode *inode) no_delete: stat_dec_inline_dir(inode); stat_dec_inline_inode(inode); + + /* update extent info in inode */ + if (inode->i_nlink) + f2fs_preserve_extent_tree(inode); f2fs_destroy_extent_tree(inode); + invalidate_mapping_pages(NODE_MAPPING(sbi), inode->i_ino, inode->i_ino); if (xnid) invalidate_mapping_pages(NODE_MAPPING(sbi), xnid, xnid); -- cgit v1.2.3 From adad81ed42bbc537f37192dcdd9a83e34bb61987 Mon Sep 17 00:00:00 2001 From: Jaegeuk Kim Date: Tue, 24 Mar 2015 12:04:20 -0700 Subject: f2fs: fix sparse warnings This patch fixes the below warning. sparse warnings: (new ones prefixed by >>) >> fs/f2fs/inode.c:56:23: sparse: restricted __le32 degrades to integer >> fs/f2fs/inode.c:56:52: sparse: restricted __le32 degrades to integer Reported-by: kbuild test robot Signed-off-by: Jaegeuk Kim --- fs/f2fs/inode.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'fs/f2fs/inode.c') diff --git a/fs/f2fs/inode.c b/fs/f2fs/inode.c index 07237ac2935e..e622ec95409e 100644 --- a/fs/f2fs/inode.c +++ b/fs/f2fs/inode.c @@ -53,7 +53,9 @@ static void __get_inode_rdev(struct inode *inode, struct f2fs_inode *ri) static bool __written_first_block(struct f2fs_inode *ri) { - if (ri->i_addr[0] != NEW_ADDR && ri->i_addr[0] != NULL_ADDR) + block_t addr = le32_to_cpu(ri->i_addr[0]); + + if (addr != NEW_ADDR && addr != NULL_ADDR) return true; return false; } -- cgit v1.2.3