!172 f2fs: extent cache: support unaligned extent

Merge Pull Request from: @gewus 
 
    Compressed inode may suffer read performance issue due to it can not
    use extent cache, so I propose to add this unaligned extent support
    to improve it.

    Currently, it only works in readonly format f2fs image.

    Unaligned extent: in one compressed cluster, physical block number
    will be less than logical block number, so we add an extra physical
    block length in extent info in order to indicate such extent status.

    The idea is if one whole cluster blocks are contiguous physically,
    once its mapping info was readed at first time, we will cache an
    unaligned (or aligned) extent info entry in extent cache, it expects
    that the mapping info will be hitted when rereading cluster. 
 
Link:https://gitee.com/openeuler/kernel/pulls/172 
Reviewed-by: Zheng Zengkai <zhengzengkai@huawei.com> 
Signed-off-by: Zheng Zengkai <zhengzengkai@huawei.com> 

fs/f2fs/compress.c

@@ -1623,6 +1623,30 @@ void f2fs_decompress_end_io(struct page **rpages,
 	}
 }
 
+/*
+ * check whether cluster blocks are contiguous, and add extent cache entry
+ * only if cluster blocks are logically and physically contiguous.
+ */
+unsigned int f2fs_cluster_blocks_are_contiguous(struct dnode_of_data *dn)
+{
+	bool compressed = f2fs_data_blkaddr(dn) == COMPRESS_ADDR;
+	int i = compressed ? 1 : 0;
+	block_t first_blkaddr = data_blkaddr(dn->inode, dn->node_page,
+						dn->ofs_in_node + i);
+
+	for (i += 1; i < F2FS_I(dn->inode)->i_cluster_size; i++) {
+		block_t blkaddr = data_blkaddr(dn->inode, dn->node_page,
+						dn->ofs_in_node + i);
+
+		if (!__is_valid_data_blkaddr(blkaddr))
+			break;
+		if (first_blkaddr + i - (compressed ? 1 : 0) != blkaddr)
+			return 0;
+	}
+
+	return compressed ? i - 1 : i;
+}
+
 int f2fs_init_page_array_cache(struct f2fs_sb_info *sbi)
 {
 	dev_t dev = sbi->sb->s_bdev->bd_dev;

fs/f2fs/data.c

@@ -1195,7 +1195,7 @@ int f2fs_reserve_block(struct dnode_of_data *dn, pgoff_t index)
 
 int f2fs_get_block(struct dnode_of_data *dn, pgoff_t index)
 {
-	struct extent_info ei = {0, 0, 0};
+	struct extent_info ei = {0, };
 	struct inode *inode = dn->inode;
 
 	if (f2fs_lookup_extent_cache(inode, index, &ei)) {
@@ -1212,7 +1212,7 @@ struct page *f2fs_get_read_data_page(struct inode *inode, pgoff_t index,
 	struct address_space *mapping = inode->i_mapping;
 	struct dnode_of_data dn;
 	struct page *page;
-	struct extent_info ei = {0,0,0};
+	struct extent_info ei = {0, };
 	int err;
 
 	page = f2fs_grab_cache_page(mapping, index, for_write);
@@ -1508,7 +1508,7 @@ int f2fs_map_blocks(struct inode *inode, struct f2fs_map_blocks *map,
 	int err = 0, ofs = 1;
 	unsigned int ofs_in_node, last_ofs_in_node;
 	blkcnt_t prealloc;
-	struct extent_info ei = {0,0,0};
+	struct extent_info ei = {0, };
 	block_t blkaddr;
 	unsigned int start_pgofs;
 
@@ -2195,6 +2195,8 @@ int f2fs_read_multi_pages(struct compress_ctx *cc, struct bio **bio_ret,
 	const unsigned blkbits = inode->i_blkbits;
 	const unsigned blocksize = 1 << blkbits;
 	struct decompress_io_ctx *dic = NULL;
+	struct extent_info ei = {0, };
+	bool from_dnode = true;
 	struct bio_post_read_ctx *ctx;
 	bool for_verity = false;
 	int i;
@@ -2229,6 +2231,12 @@ int f2fs_read_multi_pages(struct compress_ctx *cc, struct bio **bio_ret,
 	if (f2fs_cluster_is_empty(cc))
 		goto out;
 
+	if (f2fs_lookup_extent_cache(inode, start_idx, &ei))
+		from_dnode = false;
+
+	if (!from_dnode)
+		goto skip_reading_dnode;
+
 	set_new_dnode(&dn, inode, NULL, NULL, 0);
 	ret = f2fs_get_dnode_of_data(&dn, start_idx, LOOKUP_NODE);
 	if (ret)
@@ -2236,11 +2244,13 @@ int f2fs_read_multi_pages(struct compress_ctx *cc, struct bio **bio_ret,
 
 	f2fs_bug_on(sbi, dn.data_blkaddr != COMPRESS_ADDR);
 
+skip_reading_dnode:
 	for (i = 1; i < cc->cluster_size; i++) {
 		block_t blkaddr;
 
-		blkaddr = data_blkaddr(dn.inode, dn.node_page,
-						dn.ofs_in_node + i);
+		blkaddr = from_dnode ? data_blkaddr(dn.inode, dn.node_page,
+					dn.ofs_in_node + i) :
+					ei.blk + i - 1;
 
 		if (!__is_valid_data_blkaddr(blkaddr))
 			break;
@@ -2250,6 +2260,9 @@ int f2fs_read_multi_pages(struct compress_ctx *cc, struct bio **bio_ret,
 			goto out_put_dnode;
 		}
 		cc->nr_cpages++;
+
+		if (!from_dnode && i >= ei.c_len)
+			break;
 	}
 
 	/* nothing to decompress */
@@ -2350,13 +2363,15 @@ int f2fs_read_multi_pages(struct compress_ctx *cc, struct bio **bio_ret,
 		*last_block_in_bio = blkaddr;
 	}
 
-	f2fs_put_dnode(&dn);
+	if (from_dnode)
+		f2fs_put_dnode(&dn);
 
 	*bio_ret = bio;
 	return 0;
 
 out_put_dnode:
-	f2fs_put_dnode(&dn);
+	if (from_dnode)
+		f2fs_put_dnode(&dn);
 out:
 	f2fs_decompress_end_io(cc->rpages, cc->cluster_size, true, false);
 	*bio_ret = bio;
@@ -2659,7 +2674,7 @@ int f2fs_do_write_data_page(struct f2fs_io_info *fio)
 	struct page *page = fio->page;
 	struct inode *inode = page->mapping->host;
 	struct dnode_of_data dn;
-	struct extent_info ei = {0,0,0};
+	struct extent_info ei = {0, };
 	struct node_info ni;
 	bool ipu_force = false;
 	int err = 0;
@@ -3339,7 +3354,7 @@ static int prepare_write_begin(struct f2fs_sb_info *sbi,
 	struct dnode_of_data dn;
 	struct page *ipage;
 	bool locked = false;
-	struct extent_info ei = {0,0,0};
+	struct extent_info ei = {0, };
 	int err = 0;
 	int flag;
 

fs/f2fs/extent_cache.c

@@ -661,6 +661,47 @@ static void f2fs_update_extent_tree_range(struct inode *inode,
 		f2fs_mark_inode_dirty_sync(inode, true);
 }
 
+#ifdef CONFIG_F2FS_FS_COMPRESSION
+void f2fs_update_extent_tree_range_compressed(struct inode *inode,
+				pgoff_t fofs, block_t blkaddr, unsigned int llen,
+				unsigned int c_len)
+{
+	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
+	struct extent_tree *et = F2FS_I(inode)->extent_tree;
+	struct extent_node *en = NULL;
+	struct extent_node *prev_en = NULL, *next_en = NULL;
+	struct extent_info ei;
+	struct rb_node **insert_p = NULL, *insert_parent = NULL;
+	bool leftmost = false;
+
+	trace_f2fs_update_extent_tree_range(inode, fofs, blkaddr, llen);
+
+	/* it is safe here to check FI_NO_EXTENT w/o et->lock in ro image */
+	if (is_inode_flag_set(inode, FI_NO_EXTENT))
+		return;
+
+	write_lock(&et->lock);
+
+	en = (struct extent_node *)f2fs_lookup_rb_tree_ret(&et->root,
+				(struct rb_entry *)et->cached_en, fofs,
+				(struct rb_entry **)&prev_en,
+				(struct rb_entry **)&next_en,
+				&insert_p, &insert_parent, false,
+				&leftmost);
+	if (en)
+		goto unlock_out;
+
+	set_extent_info(&ei, fofs, blkaddr, llen);
+	ei.c_len = c_len;
+
+	if (!__try_merge_extent_node(sbi, et, &ei, prev_en, next_en))
+		__insert_extent_tree(sbi, et, &ei,
+				insert_p, insert_parent, leftmost);
+unlock_out:
+	write_unlock(&et->lock);
+}
+#endif
+
 unsigned int f2fs_shrink_extent_tree(struct f2fs_sb_info *sbi, int nr_shrink)
 {
 	struct extent_tree *et, *next;

fs/f2fs/f2fs.h

@@ -168,6 +168,7 @@ struct f2fs_mount_info {
 #define F2FS_FEATURE_SB_CHKSUM		0x0800
 #define F2FS_FEATURE_CASEFOLD		0x1000
 #define F2FS_FEATURE_COMPRESSION	0x2000
+#define F2FS_FEATURE_RO			0x4000
 
 #define __F2FS_HAS_FEATURE(raw_super, mask)				\
 	((raw_super->feature & cpu_to_le32(mask)) != 0)
@@ -549,6 +550,9 @@ struct extent_info {
 	unsigned int fofs;		/* start offset in a file */
 	unsigned int len;		/* length of the extent */
 	u32 blk;			/* start block address of the extent */
+#ifdef CONFIG_F2FS_FS_COMPRESSION
+	unsigned int c_len;		/* physical extent length of compressed blocks */
+#endif
 };
 
 struct extent_node {
@@ -761,6 +765,9 @@ static inline void set_extent_info(struct extent_info *ei, unsigned int fofs,
 	ei->fofs = fofs;
 	ei->blk = blk;
 	ei->len = len;
+#ifdef CONFIG_F2FS_FS_COMPRESSION
+	ei->c_len = 0;
+#endif
 }
 
 static inline bool __is_discard_mergeable(struct discard_info *back,
@@ -785,6 +792,12 @@ static inline bool __is_discard_front_mergeable(struct discard_info *cur,
 static inline bool __is_extent_mergeable(struct extent_info *back,
 						struct extent_info *front)
 {
+#ifdef CONFIG_F2FS_FS_COMPRESSION
+	if (back->c_len && back->len != back->c_len)
+		return false;
+	if (front->c_len && front->len != front->c_len)
+		return false;
+#endif
 	return (back->fofs + back->len == front->fofs &&
 			back->blk + back->len == front->blk);
 }
@@ -2955,25 +2968,6 @@ static inline bool is_dot_dotdot(const u8 *name, size_t len)
 	return false;
 }
 
-static inline bool f2fs_may_extent_tree(struct inode *inode)
-{
-	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
-
-	if (!test_opt(sbi, EXTENT_CACHE) ||
-			is_inode_flag_set(inode, FI_NO_EXTENT) ||
-			is_inode_flag_set(inode, FI_COMPRESSED_FILE))
-		return false;
-
-	/*
-	 * for recovered files during mount do not create extents
-	 * if shrinker is not registered.
-	 */
-	if (list_empty(&sbi->s_list))
-		return false;
-
-	return S_ISREG(inode->i_mode);
-}
-
 static inline void *f2fs_kmalloc(struct f2fs_sb_info *sbi,
 					size_t size, gfp_t flags)
 {
@@ -3874,6 +3868,9 @@ int f2fs_write_multi_pages(struct compress_ctx *cc,
 						struct writeback_control *wbc,
 						enum iostat_type io_type);
 int f2fs_is_compressed_cluster(struct inode *inode, pgoff_t index);
+void f2fs_update_extent_tree_range_compressed(struct inode *inode,
+				pgoff_t fofs, block_t blkaddr, unsigned int llen,
+				unsigned int c_len);
 int f2fs_read_multi_pages(struct compress_ctx *cc, struct bio **bio_ret,
 				unsigned nr_pages, sector_t *last_block_in_bio,
 				bool is_readahead, bool for_write);
@@ -3881,6 +3878,7 @@ struct decompress_io_ctx *f2fs_alloc_dic(struct compress_ctx *cc);
 void f2fs_free_dic(struct decompress_io_ctx *dic);
 void f2fs_decompress_end_io(struct page **rpages,
 			unsigned int cluster_size, bool err, bool verity);
+unsigned int f2fs_cluster_blocks_are_contiguous(struct dnode_of_data *dn);
 int f2fs_init_compress_ctx(struct compress_ctx *cc);
 void f2fs_destroy_compress_ctx(struct compress_ctx *cc, bool reuse);
 void f2fs_init_compress_info(struct f2fs_sb_info *sbi);
@@ -3902,12 +3900,16 @@ static inline struct page *f2fs_compress_control_page(struct page *page)
 	WARN_ON_ONCE(1);
 	return ERR_PTR(-EINVAL);
 }
+static inline unsigned int f2fs_cluster_blocks_are_contiguous(struct dnode_of_data *dn) { return 0; }
 static inline int f2fs_init_compress_mempool(void) { return 0; }
 static inline void f2fs_destroy_compress_mempool(void) { }
 static inline int f2fs_init_page_array_cache(struct f2fs_sb_info *sbi) { return 0; }
 static inline void f2fs_destroy_page_array_cache(struct f2fs_sb_info *sbi) { }
 static inline int __init f2fs_init_compress_cache(void) { return 0; }
 static inline void f2fs_destroy_compress_cache(void) { }
+static inline void f2fs_update_extent_tree_range_compressed(struct inode *inode,
+				pgoff_t fofs, block_t blkaddr, unsigned int llen,
+				unsigned int c_len) { }
 #endif
 
 static inline void set_compress_context(struct inode *inode)
@@ -3969,6 +3971,27 @@ F2FS_FEATURE_FUNCS(verity, VERITY);
 F2FS_FEATURE_FUNCS(sb_chksum, SB_CHKSUM);
 F2FS_FEATURE_FUNCS(casefold, CASEFOLD);
 F2FS_FEATURE_FUNCS(compression, COMPRESSION);
+F2FS_FEATURE_FUNCS(readonly, RO);
+
+static inline bool f2fs_may_extent_tree(struct inode *inode)
+{
+	struct f2fs_sb_info *sbi = F2FS_I_SB(inode);
+
+	if (!test_opt(sbi, EXTENT_CACHE) ||
+			is_inode_flag_set(inode, FI_NO_EXTENT) ||
+			(is_inode_flag_set(inode, FI_COMPRESSED_FILE) &&
+			 !f2fs_sb_has_readonly(sbi)))
+		return false;
+
+	/*
+	 * for recovered files during mount do not create extents
+	 * if shrinker is not registered.
+	 */
+	if (list_empty(&sbi->s_list))
+		return false;
+
+	return S_ISREG(inode->i_mode);
+}
 
 #ifdef CONFIG_BLK_DEV_ZONED
 static inline bool f2fs_blkz_is_seq(struct f2fs_sb_info *sbi, int devi,

fs/f2fs/node.c

@@ -810,6 +810,26 @@ int f2fs_get_dnode_of_data(struct dnode_of_data *dn, pgoff_t index, int mode)
 	dn->ofs_in_node = offset[level];
 	dn->node_page = npage[level];
 	dn->data_blkaddr = f2fs_data_blkaddr(dn);
+
+	if (is_inode_flag_set(dn->inode, FI_COMPRESSED_FILE) &&
+					f2fs_sb_has_readonly(sbi)) {
+		unsigned int c_len = f2fs_cluster_blocks_are_contiguous(dn);
+		block_t blkaddr;
+
+		if (!c_len)
+			goto out;
+
+		blkaddr = f2fs_data_blkaddr(dn);
+		if (blkaddr == COMPRESS_ADDR)
+			blkaddr = data_blkaddr(dn->inode, dn->node_page,
+						dn->ofs_in_node + 1);
+
+		f2fs_update_extent_tree_range_compressed(dn->inode,
+					index, blkaddr,
+					F2FS_I(dn->inode)->i_cluster_size,
+					c_len);
+	}
+out:
 	return 0;
 
 release_pages: