f2fs: support age threshold based garbage collection

There are several issues in current background GC algorithm:
- valid blocks is one of key factors during cost overhead calculation,
so if segment has less valid block, however even its age is young or
it locates hot segment, CB algorithm will still choose the segment as
victim, it's not appropriate.
- GCed data/node will go to existing logs, no matter in-there datas'
update frequency is the same or not, it may mix hot and cold data
again.
- GC alloctor mainly use LFS type segment, it will cost free segment
more quickly.

This patch introduces a new algorithm named age threshold based
garbage collection to solve above issues, there are three steps
mainly:

1. select a source victim:
- set an age threshold, and select candidates beased threshold:
e.g.
 0 means youngest, 100 means oldest, if we set age threshold to 80
 then select dirty segments which has age in range of [80, 100] as
 candiddates;
- set candidate_ratio threshold, and select candidates based the
ratio, so that we can shrink candidates to those oldest segments;
- select target segment with fewest valid blocks in order to
migrate blocks with minimum cost;

2. select a target victim:
- select candidates beased age threshold;
- set candidate_radius threshold, search candidates whose age is
around source victims, searching radius should less than the
radius threshold.
- select target segment with most valid blocks in order to avoid
migrating current target segment.

3. merge valid blocks from source victim into target victim with
SSR alloctor.

Test steps:
- create 160 dirty segments:
 * half of them have 128 valid blocks per segment
 * left of them have 384 valid blocks per segment
- run background GC

Benefit: GC count and block movement count both decrease obviously:

- Before:
  - Valid: 86
  - Dirty: 1
  - Prefree: 11
  - Free: 6001 (6001)

GC calls: 162 (BG: 220)
  - data segments : 160 (160)
  - node segments : 2 (2)
Try to move 41454 blocks (BG: 41454)
  - data blocks : 40960 (40960)
  - node blocks : 494 (494)

IPU: 0 blocks
SSR: 0 blocks in 0 segments
LFS: 41364 blocks in 81 segments

- After:

  - Valid: 87
  - Dirty: 0
  - Prefree: 4
  - Free: 6008 (6008)

GC calls: 75 (BG: 76)
  - data segments : 74 (74)
  - node segments : 1 (1)
Try to move 12813 blocks (BG: 12813)
  - data blocks : 12544 (12544)
  - node blocks : 269 (269)

IPU: 0 blocks
SSR: 12032 blocks in 77 segments
LFS: 855 blocks in 2 segments
Signed-off-by: Chao Yu <yuchao0@huawei.com>
[Jaegeuk Kim: fix a bug along with pinfile in-mem segment & clean up]
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>

Documentation/ABI/testing/sysfs-fs-f2fs

@@ -22,7 +22,8 @@ Contact:	"Namjae Jeon" <namjae.jeon@samsung.com>
 Description:	Controls the victim selection policy for garbage collection.
 		Setting gc_idle = 0(default) will disable this option. Setting
 		gc_idle = 1 will select the Cost Benefit approach & setting
-		gc_idle = 2 will select the greedy approach.
+		gc_idle = 2 will select the greedy approach & setting
+		gc_idle = 3 will select the age-threshold based approach.
 
 What:		/sys/fs/f2fs/<disk>/reclaim_segments
 Date:		October 2013

Documentation/filesystems/f2fs.rst

@@ -266,6 +266,8 @@ inlinecrypt		 When possible, encrypt/decrypt the contents of encrypted
 			 inline encryption hardware. The on-disk format is
 			 unaffected. For more details, see
 			 Documentation/block/inline-encryption.rst.
+atgc			 Enable age-threshold garbage collection, it provides high
+			 effectiveness and efficiency on background GC.
 ======================== ============================================================
 
 Debugfs Entries

fs/f2fs/checkpoint.c

@@ -1620,7 +1620,7 @@ int f2fs_write_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc)
 	f2fs_flush_sit_entries(sbi, cpc);
 
 	/* save inmem log status */
-	f2fs_save_inmem_curseg(sbi, CURSEG_COLD_DATA_PINNED);
+	f2fs_save_inmem_curseg(sbi);
 
 	err = do_checkpoint(sbi, cpc);
 	if (err)
@@ -1628,7 +1628,7 @@ int f2fs_write_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc)
 	else
 		f2fs_clear_prefree_segments(sbi, cpc);
 
-	f2fs_restore_inmem_curseg(sbi, CURSEG_COLD_DATA_PINNED);
+	f2fs_restore_inmem_curseg(sbi);
 stop:
 	unblock_operations(sbi);
 	stat_inc_cp_count(sbi->stat_info);

fs/f2fs/data.c

@@ -1416,7 +1416,7 @@ static int __allocate_data_block(struct dnode_of_data *dn, int seg_type)
 	set_summary(&sum, dn->nid, dn->ofs_in_node, ni.version);
 	old_blkaddr = dn->data_blkaddr;
 	f2fs_allocate_data_block(sbi, NULL, old_blkaddr, &dn->data_blkaddr,
-				&sum, seg_type, NULL, false);
+				&sum, seg_type, NULL);
 	if (GET_SEGNO(sbi, old_blkaddr) != NULL_SEGNO)
 		invalidate_mapping_pages(META_MAPPING(sbi),
 					old_blkaddr, old_blkaddr);

fs/f2fs/debug.c

@@ -397,6 +397,10 @@ static int stat_show(struct seq_file *s, void *v)
 			   si->curseg[CURSEG_COLD_DATA_PINNED],
 			   si->cursec[CURSEG_COLD_DATA_PINNED],
 			   si->curzone[CURSEG_COLD_DATA_PINNED]);
+		seq_printf(s, "  - ATGC   data: %8d %8d %8d\n",
+			   si->curseg[CURSEG_ALL_DATA_ATGC],
+			   si->cursec[CURSEG_ALL_DATA_ATGC],
+			   si->curzone[CURSEG_ALL_DATA_ATGC]);
 		seq_printf(s, "\n  - Valid: %d\n  - Dirty: %d\n",
 			   si->main_area_segs - si->dirty_count -
 			   si->prefree_count - si->free_segs,

fs/f2fs/f2fs.h

@@ -98,6 +98,7 @@ extern const char *f2fs_fault_name[FAULT_MAX];
 #define F2FS_MOUNT_RESERVE_ROOT		0x01000000
 #define F2FS_MOUNT_DISABLE_CHECKPOINT	0x02000000
 #define F2FS_MOUNT_NORECOVERY		0x04000000
+#define F2FS_MOUNT_ATGC			0x08000000
 
 #define F2FS_OPTION(sbi)	((sbi)->mount_opt)
 #define clear_opt(sbi, option)	(F2FS_OPTION(sbi).opt &= ~F2FS_MOUNT_##option)
@@ -978,7 +979,7 @@ static inline void set_new_dnode(struct dnode_of_data *dn, struct inode *inode,
  */
 #define	NR_CURSEG_DATA_TYPE	(3)
 #define NR_CURSEG_NODE_TYPE	(3)
-#define NR_CURSEG_INMEM_TYPE	(1)
+#define NR_CURSEG_INMEM_TYPE	(2)
 #define NR_CURSEG_PERSIST_TYPE	(NR_CURSEG_DATA_TYPE + NR_CURSEG_NODE_TYPE)
 #define NR_CURSEG_TYPE		(NR_CURSEG_INMEM_TYPE + NR_CURSEG_PERSIST_TYPE)
 
@@ -992,6 +993,7 @@ enum {
 	NR_PERSISTENT_LOG,	/* number of persistent log */
 	CURSEG_COLD_DATA_PINNED = NR_PERSISTENT_LOG,
 				/* pinned file that needs consecutive block address */
+	CURSEG_ALL_DATA_ATGC,	/* SSR alloctor in hot/warm/cold data area */
 	NO_CHECK_TYPE,		/* number of persistent & inmem log */
 };
 
@@ -1238,6 +1240,18 @@ struct inode_management {
 	unsigned long ino_num;			/* number of entries */
 };
 
+/* for GC_AT */
+struct atgc_management {
+	bool atgc_enabled;			/* ATGC is enabled or not */
+	struct rb_root_cached root;		/* root of victim rb-tree */
+	struct list_head victim_list;		/* linked with all victim entries */
+	unsigned int victim_count;		/* victim count in rb-tree */
+	unsigned int candidate_ratio;		/* candidate ratio */
+	unsigned int max_candidate_count;	/* max candidate count */
+	unsigned int age_weight;		/* age weight, vblock_weight = 100 - age_weight */
+	unsigned long long age_threshold;	/* age threshold */
+};
+
 /* For s_flag in struct f2fs_sb_info */
 enum {
 	SBI_IS_DIRTY,				/* dirty flag for checkpoint */
@@ -1270,6 +1284,7 @@ enum {
 	GC_NORMAL,
 	GC_IDLE_CB,
 	GC_IDLE_GREEDY,
+	GC_IDLE_AT,
 	GC_URGENT_HIGH,
 	GC_URGENT_LOW,
 };
@@ -1521,6 +1536,7 @@ struct f2fs_sb_info {
 						 * race between GC and GC or CP
 						 */
 	struct f2fs_gc_kthread	*gc_thread;	/* GC thread */
+	struct atgc_management am;		/* atgc management */
 	unsigned int cur_victim_sec;		/* current victim section num */
 	unsigned int gc_mode;			/* current GC state */
 	unsigned int next_victim_seg[2];	/* next segment in victim section */
@@ -3338,8 +3354,11 @@ block_t f2fs_get_unusable_blocks(struct f2fs_sb_info *sbi);
 int f2fs_disable_cp_again(struct f2fs_sb_info *sbi, block_t unusable);
 void f2fs_release_discard_addrs(struct f2fs_sb_info *sbi);
 int f2fs_npages_for_summary_flush(struct f2fs_sb_info *sbi, bool for_ra);
-void f2fs_save_inmem_curseg(struct f2fs_sb_info *sbi, int type);
-void f2fs_restore_inmem_curseg(struct f2fs_sb_info *sbi, int type);
+void f2fs_init_inmem_curseg(struct f2fs_sb_info *sbi);
+void f2fs_save_inmem_curseg(struct f2fs_sb_info *sbi);
+void f2fs_restore_inmem_curseg(struct f2fs_sb_info *sbi);
+void f2fs_get_new_segment(struct f2fs_sb_info *sbi,
+			unsigned int *newseg, bool new_sec, int dir);
 void f2fs_allocate_segment_for_resize(struct f2fs_sb_info *sbi, int type,
 					unsigned int start, unsigned int end);
 void f2fs_allocate_new_segment(struct f2fs_sb_info *sbi, int type);
@@ -3367,7 +3386,7 @@ void f2fs_replace_block(struct f2fs_sb_info *sbi, struct dnode_of_data *dn,
 void f2fs_allocate_data_block(struct f2fs_sb_info *sbi, struct page *page,
 			block_t old_blkaddr, block_t *new_blkaddr,
 			struct f2fs_summary *sum, int type,
-			struct f2fs_io_info *fio, bool from_gc);
+			struct f2fs_io_info *fio);
 void f2fs_wait_on_page_writeback(struct page *page,
 			enum page_type type, bool ordered, bool locked);
 void f2fs_wait_on_block_writeback(struct inode *inode, block_t blkaddr);
@@ -3506,6 +3525,8 @@ int f2fs_gc(struct f2fs_sb_info *sbi, bool sync, bool background,
 			unsigned int segno);
 void f2fs_build_gc_manager(struct f2fs_sb_info *sbi);
 int f2fs_resize_fs(struct f2fs_sb_info *sbi, __u64 block_count);
+int __init f2fs_create_garbage_collection_cache(void);
+void f2fs_destroy_garbage_collection_cache(void);
 
 /*
  * recovery.c

fs/f2fs/gc.h

@@ -14,6 +14,14 @@
 #define DEF_GC_THREAD_MIN_SLEEP_TIME	30000	/* milliseconds */
 #define DEF_GC_THREAD_MAX_SLEEP_TIME	60000
 #define DEF_GC_THREAD_NOGC_SLEEP_TIME	300000	/* wait 5 min */
+
+/* choose candidates from sections which has age of more than 7 days */
+#define DEF_GC_THREAD_AGE_THRESHOLD		(60 * 60 * 24 * 7)
+#define DEF_GC_THREAD_CANDIDATE_RATIO		20	/* select 20% oldest sections as candidates */
+#define DEF_GC_THREAD_MAX_CANDIDATE_COUNT	10	/* select at most 10 sections as candidates */
+#define DEF_GC_THREAD_AGE_WEIGHT		60	/* age weight */
+#define DEFAULT_ACCURACY_CLASS			10000	/* accuracy class */
+
 #define LIMIT_INVALID_BLOCK	40 /* percentage over total user space */
 #define LIMIT_FREE_BLOCK	40 /* percentage over invalid + free space */
 
@@ -41,6 +49,23 @@ struct gc_inode_list {
 	struct radix_tree_root iroot;
 };
 
+struct victim_info {
+	unsigned long long mtime;	/* mtime of section */
+	unsigned int segno;		/* section No. */
+};
+
+struct victim_entry {
+	struct rb_node rb_node;		/* rb node located in rb-tree */
+	union {
+		struct {
+			unsigned long long mtime;	/* mtime of section */
+			unsigned int segno;		/* segment No. */
+		};
+		struct victim_info vi;	/* victim info */
+	};
+	struct list_head list;
+};
+
 /*
  * inline functions
  */

fs/f2fs/segment.h

@@ -24,6 +24,12 @@
 #define IS_DATASEG(t)	((t) <= CURSEG_COLD_DATA)
 #define IS_NODESEG(t)	((t) >= CURSEG_HOT_NODE && (t) <= CURSEG_COLD_NODE)
 
+static inline void sanity_check_seg_type(struct f2fs_sb_info *sbi,
+						unsigned short seg_type)
+{
+	f2fs_bug_on(sbi, seg_type >= NR_PERSISTENT_LOG);
+}
+
 #define IS_HOT(t)	((t) == CURSEG_HOT_NODE || (t) == CURSEG_HOT_DATA)
 #define IS_WARM(t)	((t) == CURSEG_WARM_NODE || (t) == CURSEG_WARM_DATA)
 #define IS_COLD(t)	((t) == CURSEG_COLD_NODE || (t) == CURSEG_COLD_DATA)
@@ -35,7 +41,8 @@
 	 ((seg) == CURSEG_I(sbi, CURSEG_HOT_NODE)->segno) ||	\
 	 ((seg) == CURSEG_I(sbi, CURSEG_WARM_NODE)->segno) ||	\
 	 ((seg) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno) ||	\
-	 ((seg) == CURSEG_I(sbi, CURSEG_COLD_DATA_PINNED)->segno))
+	 ((seg) == CURSEG_I(sbi, CURSEG_COLD_DATA_PINNED)->segno) ||	\
+	 ((seg) == CURSEG_I(sbi, CURSEG_ALL_DATA_ATGC)->segno))
 
 #define IS_CURSEC(sbi, secno)						\
 	(((secno) == CURSEG_I(sbi, CURSEG_HOT_DATA)->segno /		\
@@ -51,6 +58,8 @@
 	 ((secno) == CURSEG_I(sbi, CURSEG_COLD_NODE)->segno /		\
 	  (sbi)->segs_per_sec) ||	\
 	 ((secno) == CURSEG_I(sbi, CURSEG_COLD_DATA_PINNED)->segno /	\
+	  (sbi)->segs_per_sec) ||	\
+	 ((secno) == CURSEG_I(sbi, CURSEG_ALL_DATA_ATGC)->segno /	\
 	  (sbi)->segs_per_sec))
 
 #define MAIN_BLKADDR(sbi)						\
@@ -135,20 +144,25 @@ enum {
  * In the victim_sel_policy->alloc_mode, there are two block allocation modes.
  * LFS writes data sequentially with cleaning operations.
  * SSR (Slack Space Recycle) reuses obsolete space without cleaning operations.
+ * AT_SSR (Age Threshold based Slack Space Recycle) merges fragments into
+ * fragmented segment which has similar aging degree.
  */
 enum {
 	LFS = 0,
-	SSR
+	SSR,
+	AT_SSR,
 };
 
 /*
  * In the victim_sel_policy->gc_mode, there are two gc, aka cleaning, modes.
  * GC_CB is based on cost-benefit algorithm.
  * GC_GREEDY is based on greedy algorithm.
+ * GC_AT is based on age-threshold algorithm.
  */
 enum {
 	GC_CB = 0,
 	GC_GREEDY,
+	GC_AT,
 	ALLOC_NEXT,
 	FLUSH_DEVICE,
 	MAX_GC_POLICY,
@@ -177,7 +191,10 @@ struct victim_sel_policy {
 	unsigned int offset;		/* last scanned bitmap offset */
 	unsigned int ofs_unit;		/* bitmap search unit */
 	unsigned int min_cost;		/* minimum cost */
+	unsigned long long oldest_age;	/* oldest age of segments having the same min cost */
 	unsigned int min_segno;		/* segment # having min. cost */
+	unsigned long long age;		/* mtime of GCed section*/
+	unsigned long long age_threshold;/* age threshold */
 };
 
 struct seg_entry {
@@ -243,6 +260,8 @@ struct sit_info {
 	unsigned long long mounted_time;	/* mount time */
 	unsigned long long min_mtime;		/* min. modification time */
 	unsigned long long max_mtime;		/* max. modification time */
+	unsigned long long dirty_min_mtime;	/* rerange candidates in GC_AT */
+	unsigned long long dirty_max_mtime;	/* rerange candidates in GC_AT */
 
 	unsigned int last_victim[MAX_GC_POLICY]; /* last victim segment # */
 };
@@ -281,7 +300,7 @@ struct dirty_seglist_info {
 /* victim selection function for cleaning and SSR */
 struct victim_selection {
 	int (*get_victim)(struct f2fs_sb_info *, unsigned int *,
-							int, int, char);
+					int, int, char, unsigned long long);
 };
 
 /* for active log information */

fs/f2fs/super.c

@@ -146,6 +146,7 @@ enum {
 	Opt_compress_algorithm,
 	Opt_compress_log_size,
 	Opt_compress_extension,
+	Opt_atgc,
 	Opt_err,
 };
 
@@ -213,6 +214,7 @@ static match_table_t f2fs_tokens = {
 	{Opt_compress_algorithm, "compress_algorithm=%s"},
 	{Opt_compress_log_size, "compress_log_size=%u"},
 	{Opt_compress_extension, "compress_extension=%s"},
+	{Opt_atgc, "atgc"},
 	{Opt_err, NULL},
 };
 
@@ -939,6 +941,9 @@ static int parse_options(struct super_block *sb, char *options, bool is_remount)
 			f2fs_info(sbi, "compression options not supported");
 			break;
 #endif
+		case Opt_atgc:
+			set_opt(sbi, ATGC);
+			break;
 		default:
 			f2fs_err(sbi, "Unrecognized mount option \"%s\" or missing value",
 				 p);
@@ -1636,6 +1641,9 @@ static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
 #ifdef CONFIG_F2FS_FS_COMPRESSION
 	f2fs_show_compress_options(seq, sbi->sb);
 #endif
+
+	if (test_opt(sbi, ATGC))
+		seq_puts(seq, ",atgc");
 	return 0;
 }
 
@@ -1765,6 +1773,7 @@ static int f2fs_remount(struct super_block *sb, int *flags, char *data)
 	bool no_extent_cache = !test_opt(sbi, EXTENT_CACHE);
 	bool disable_checkpoint = test_opt(sbi, DISABLE_CHECKPOINT);
 	bool no_io_align = !F2FS_IO_ALIGNED(sbi);
+	bool no_atgc = !test_opt(sbi, ATGC);
 	bool checkpoint_changed;
 #ifdef CONFIG_QUOTA
 	int i, j;
@@ -1837,6 +1846,13 @@ static int f2fs_remount(struct super_block *sb, int *flags, char *data)
 		}
 	}
 #endif
+	/* disallow enable atgc dynamically */
+	if (no_atgc == !!test_opt(sbi, ATGC)) {
+		err = -EINVAL;
+		f2fs_warn(sbi, "switch atgc option is not allowed");
+		goto restore_opts;
+	}
+
 	/* disallow enable/disable extent_cache dynamically */
 	if (no_extent_cache == !!test_opt(sbi, EXTENT_CACHE)) {
 		err = -EINVAL;
@@ -3792,6 +3808,8 @@ static int f2fs_fill_super(struct super_block *sb, void *data, int silent)
 	}
 
 reset_checkpoint:
+	f2fs_init_inmem_curseg(sbi);
+
 	/* f2fs_recover_fsync_data() cleared this already */
 	clear_sbi_flag(sbi, SBI_POR_DOING);
 
@@ -3997,9 +4015,12 @@ static int __init init_f2fs_fs(void)
 	err = f2fs_create_extent_cache();
 	if (err)
 		goto free_checkpoint_caches;
-	err = f2fs_init_sysfs();
+	err = f2fs_create_garbage_collection_cache();
 	if (err)
 		goto free_extent_cache;
+	err = f2fs_init_sysfs();
+	if (err)
+		goto free_garbage_collection_cache;
 	err = register_shrinker(&f2fs_shrinker_info);
 	if (err)
 		goto free_sysfs;
@@ -4033,6 +4054,8 @@ static int __init init_f2fs_fs(void)
 	unregister_shrinker(&f2fs_shrinker_info);
 free_sysfs:
 	f2fs_exit_sysfs();
+free_garbage_collection_cache:
+	f2fs_destroy_garbage_collection_cache();
 free_extent_cache:
 	f2fs_destroy_extent_cache();
 free_checkpoint_caches:
@@ -4057,6 +4080,7 @@ static void __exit exit_f2fs_fs(void)
 	unregister_filesystem(&f2fs_fs_type);
 	unregister_shrinker(&f2fs_shrinker_info);
 	f2fs_exit_sysfs();
+	f2fs_destroy_garbage_collection_cache();
 	f2fs_destroy_extent_cache();
 	f2fs_destroy_checkpoint_caches();
 	f2fs_destroy_segment_manager_caches();

fs/f2fs/sysfs.c

@@ -377,12 +377,17 @@ static ssize_t __sbi_store(struct f2fs_attr *a,
 		return count;
 	}
 	if (!strcmp(a->attr.name, "gc_idle")) {
-		if (t == GC_IDLE_CB)
+		if (t == GC_IDLE_CB) {
 			sbi->gc_mode = GC_IDLE_CB;
-		else if (t == GC_IDLE_GREEDY)
+		} else if (t == GC_IDLE_GREEDY) {
 			sbi->gc_mode = GC_IDLE_GREEDY;
-		else
+		} else if (t == GC_IDLE_AT) {
+			if (!sbi->am.atgc_enabled)
+				return -EINVAL;
+			sbi->gc_mode = GC_AT;
+		} else {
 			sbi->gc_mode = GC_NORMAL;
+		}
 		return count;
 	}
 

include/trace/events/f2fs.h

@@ -111,13 +111,15 @@ TRACE_DEFINE_ENUM(CP_RESIZE);
 
 #define show_alloc_mode(type)						\
 	__print_symbolic(type,						\
-		{ LFS,	"LFS-mode" },					\
-		{ SSR,	"SSR-mode" })
+		{ LFS,		"LFS-mode" },				\
+		{ SSR,		"SSR-mode" },				\
+		{ AT_SSR,	"AT_SSR-mode" })
 
 #define show_victim_policy(type)					\
 	__print_symbolic(type,						\
 		{ GC_GREEDY,	"Greedy" },				\
-		{ GC_CB,	"Cost-Benefit" })
+		{ GC_CB,	"Cost-Benefit" },			\
+		{ GC_AT,	"Age-threshold" })
 
 #define show_cpreason(type)						\
 	__print_flags(type, "|",					\