f2fs: add block_age-based extent cache

This patch introduces a runtime hot/cold data separation method
for f2fs, in order to improve the accuracy for data temperature
classification, reduce the garbage collection overhead after
long-term data updates.

Enhanced hot/cold data separation can record data block update
frequency as "age" of the extent per inode, and take use of the age
info to indicate better temperature type for data block allocation:
 - It records total data blocks allocated since mount;
 - When file extent has been updated, it calculate the count of data
blocks allocated since last update as the age of the extent;
 - Before the data block allocated, it searches for the age info and
chooses the suitable segment for allocation.

Test and result:
 - Prepare: create about 30000 files
  * 3% for cold files (with cold file extension like .apk, from 3M to 10M)
  * 50% for warm files (with random file extension like .FcDxq, from 1K
to 4M)
  * 47% for hot files (with hot file extension like .db, from 1K to 256K)
 - create(5%)/random update(90%)/delete(5%) the files
  * total write amount is about 70G
  * fsync will be called for .db files, and buffered write will be used
for other files

The storage of test device is large enough(128G) so that it will not
switch to SSR mode during the test.

Benefit: dirty segment count increment reduce about 14%
 - before: Dirty +21110
 - after:  Dirty +18286

Signed-off-by: qixiaoyu1 <qixiaoyu1@xiaomi.com>
Signed-off-by: xiongping1 <xiongping1@xiaomi.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>

Documentation/ABI/testing/sysfs-fs-f2fs

@@ -655,3 +655,17 @@ Description:	When space utilization exceeds this, do background DISCARD aggressi
 		Does DISCARD forcibly in a period of given min_discard_issue_time when the number
 		of discards is not 0 and set discard granularity to 1.
 		Default: 80
+
+What:		/sys/fs/f2fs/<disk>/hot_data_age_threshold
+Date:		November 2022
+Contact:	"Ping Xiong" <xiongping1@xiaomi.com>
+Description:	When DATA SEPARATION is on, it controls the age threshold to indicate
+		the data blocks as hot. By default it was initialized as 262144 blocks
+		(equals to 1GB).
+
+What:		/sys/fs/f2fs/<disk>/warm_data_age_threshold
+Date:		November 2022
+Contact:	"Ping Xiong" <xiongping1@xiaomi.com>
+Description:	When DATA SEPARATION is on, it controls the age threshold to indicate
+		the data blocks as warm. By default it was initialized as 2621440 blocks
+		(equals to 10GB).

Documentation/filesystems/f2fs.rst

@@ -347,6 +347,10 @@ memory=%s		 Control memory mode. This supports "normal" and "low" modes.
 			 Because of the nature of low memory devices, in this mode, f2fs
 			 will try to save memory sometimes by sacrificing performance.
 			 "normal" mode is the default mode and same as before.
+age_extent_cache	 Enable an age extent cache based on rb-tree. It records
+			 data block update frequency of the extent per inode, in
+			 order to provide better temperature hints for data block
+			 allocation.
 ======================== ============================================================
 
 Debugfs Entries

fs/f2fs/debug.c

@@ -88,6 +88,9 @@ static void update_general_status(struct f2fs_sb_info *sbi)
 	si->hit_largest = atomic64_read(&sbi->read_hit_largest);
 	si->hit_total[EX_READ] += si->hit_largest;
 
+	/* block age extent_cache only */
+	si->allocated_data_blocks = atomic64_read(&sbi->allocated_data_blocks);
+
 	/* validation check of the segment numbers */
 	si->ndirty_node = get_pages(sbi, F2FS_DIRTY_NODES);
 	si->ndirty_dent = get_pages(sbi, F2FS_DIRTY_DENTS);
@@ -516,6 +519,22 @@ static int stat_show(struct seq_file *s, void *v)
 		seq_printf(s, "  - Inner Struct Count: tree: %d(%d), node: %d\n",
 				si->ext_tree[EX_READ], si->zombie_tree[EX_READ],
 				si->ext_node[EX_READ]);
+		seq_puts(s, "\nExtent Cache (Block Age):\n");
+		seq_printf(s, "  - Allocated Data Blocks: %llu\n",
+				si->allocated_data_blocks);
+		seq_printf(s, "  - Hit Count: L1:%llu L2:%llu\n",
+				si->hit_cached[EX_BLOCK_AGE],
+				si->hit_rbtree[EX_BLOCK_AGE]);
+		seq_printf(s, "  - Hit Ratio: %llu%% (%llu / %llu)\n",
+				!si->total_ext[EX_BLOCK_AGE] ? 0 :
+				div64_u64(si->hit_total[EX_BLOCK_AGE] * 100,
+				si->total_ext[EX_BLOCK_AGE]),
+				si->hit_total[EX_BLOCK_AGE],
+				si->total_ext[EX_BLOCK_AGE]);
+		seq_printf(s, "  - Inner Struct Count: tree: %d(%d), node: %d\n",
+				si->ext_tree[EX_BLOCK_AGE],
+				si->zombie_tree[EX_BLOCK_AGE],
+				si->ext_node[EX_BLOCK_AGE]);
 		seq_puts(s, "\nBalancing F2FS Async:\n");
 		seq_printf(s, "  - DIO (R: %4d, W: %4d)\n",
 			   si->nr_dio_read, si->nr_dio_write);
@@ -586,6 +605,8 @@ static int stat_show(struct seq_file *s, void *v)
 				si->cache_mem >> 10);
 		seq_printf(s, "  - read extent cache: %llu KB\n",
 				si->ext_mem[EX_READ] >> 10);
+		seq_printf(s, "  - block age extent cache: %llu KB\n",
+				si->ext_mem[EX_BLOCK_AGE] >> 10);
 		seq_printf(s, "  - paged : %llu KB\n",
 				si->page_mem >> 10);
 	}