f2fs: introduce fragment allocation mode mount option

Added two options into "mode=" mount option to make it possible for
developers to simulate filesystem fragmentation/after-GC situation
itself. The developers use these modes to understand filesystem
fragmentation/after-GC condition well, and eventually get some
insights to handle them better.

"fragment:segment": f2fs allocates a new segment in ramdom position.
		With this, we can simulate the after-GC condition.
"fragment:block" : We can scatter block allocation with
		"max_fragment_chunk" and "max_fragment_hole" sysfs
		nodes. f2fs will allocate 1..<max_fragment_chunk>
		blocks in a chunk and make a hole in the length of
		1..<max_fragment_hole> by turns	in a newly allocated
		free segment. Plus, this mode implicitly enables
		"fragment:segment" option for more randomness.
Reviewed-by: Chao Yu <chao@kernel.org>
Signed-off-by: Daeho Jeong <daehojeong@google.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>

Documentation/ABI/testing/sysfs-fs-f2fs

@@ -512,3 +512,19 @@ Date:		July 2021
 Contact:	"Daeho Jeong" <daehojeong@google.com>
 Description:	You can	control the multiplier value of	bdi device readahead window size
 		between 2 (default) and 256 for POSIX_FADV_SEQUENTIAL advise option.
+
+What:		/sys/fs/f2fs/<disk>/max_fragment_chunk
+Date:		August 2021
+Contact:	"Daeho Jeong" <daehojeong@google.com>
+Description:	With "mode=fragment:block" mount options, we can scatter block allocation.
+		f2fs will allocate 1..<max_fragment_chunk> blocks in a chunk and make a hole
+		in the length of 1..<max_fragment_hole> by turns. This value can be set
+		between 1..512 and the default value is 4.
+
+What:		/sys/fs/f2fs/<disk>/max_fragment_hole
+Date:		August 2021
+Contact:	"Daeho Jeong" <daehojeong@google.com>
+Description:	With "mode=fragment:block" mount options, we can scatter block allocation.
+		f2fs will allocate 1..<max_fragment_chunk> blocks in a chunk and make a hole
+		in the length of 1..<max_fragment_hole> by turns. This value can be set
+		between 1..512 and the default value is 4.

Documentation/filesystems/f2fs.rst

@@ -201,6 +201,24 @@ fault_type=%d		 Support configuring fault injection type, should be
 mode=%s			 Control block allocation mode which supports "adaptive"
 			 and "lfs". In "lfs" mode, there should be no random
 			 writes towards main area.
+			 "fragment:segment" and "fragment:block" are newly added here.
+			 These are developer options for experiments to simulate filesystem
+			 fragmentation/after-GC situation itself. The developers use these
+			 modes to understand filesystem fragmentation/after-GC condition well,
+			 and eventually get some insights to handle them better.
+			 In "fragment:segment", f2fs allocates a new segment in ramdom
+			 position. With this, we can simulate the after-GC condition.
+			 In "fragment:block", we can scatter block allocation with
+			 "max_fragment_chunk" and "max_fragment_hole" sysfs nodes.
+			 We added some randomness to both chunk and hole size to make
+			 it close to realistic IO pattern. So, in this mode, f2fs will allocate
+			 1..<max_fragment_chunk> blocks in a chunk and make a hole in the
+			 length of 1..<max_fragment_hole> by turns. With this, the newly
+			 allocated blocks will be scattered throughout the whole partition.
+			 Note that "fragment:block" implicitly enables "fragment:segment"
+			 option for more randomness.
+			 Please, use these options for your experiments and we strongly
+			 recommend to re-format the filesystem after using these options.
 io_bits=%u		 Set the bit size of write IO requests. It should be set
 			 with "mode=lfs".
 usrquota		 Enable plain user disk quota accounting.

fs/f2fs/f2fs.h

@@ -1289,6 +1289,8 @@ enum {
 enum {
 	FS_MODE_ADAPTIVE,		/* use both lfs/ssr allocation */
 	FS_MODE_LFS,			/* use lfs allocation only */
+	FS_MODE_FRAGMENT_SEG,		/* segment fragmentation mode */
+	FS_MODE_FRAGMENT_BLK,		/* block fragmentation mode */
 };
 
 enum {
@@ -1759,6 +1761,9 @@ struct f2fs_sb_info {
 
 	unsigned long seq_file_ra_mul;		/* multiplier for ra_pages of seq. files in fadvise */
 
+	int max_fragment_chunk;			/* max chunk size for block fragmentation mode */
+	int max_fragment_hole;			/* max hole size for block fragmentation mode */
+
 #ifdef CONFIG_F2FS_FS_COMPRESSION
 	struct kmem_cache *page_array_slab;	/* page array entry */
 	unsigned int page_array_slab_size;	/* default page array slab size */
@@ -3519,6 +3524,16 @@ unsigned int f2fs_usable_segs_in_sec(struct f2fs_sb_info *sbi,
 unsigned int f2fs_usable_blks_in_seg(struct f2fs_sb_info *sbi,
 			unsigned int segno);
 
+#define DEF_FRAGMENT_SIZE	4
+#define MIN_FRAGMENT_SIZE	1
+#define MAX_FRAGMENT_SIZE	512
+
+static inline bool f2fs_need_rand_seg(struct f2fs_sb_info *sbi)
+{
+	return F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_SEG ||
+		F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK;
+}
+
 /*
  * checkpoint.c
  */

fs/f2fs/gc.c

@@ -14,6 +14,7 @@
 #include <linux/delay.h>
 #include <linux/freezer.h>
 #include <linux/sched/signal.h>
+#include <linux/random.h>
 
 #include "f2fs.h"
 #include "node.h"
@@ -257,7 +258,9 @@ static void select_policy(struct f2fs_sb_info *sbi, int gc_type,
 		p->max_search = sbi->max_victim_search;
 
 	/* let's select beginning hot/small space first in no_heap mode*/
-	if (test_opt(sbi, NOHEAP) &&
+	if (f2fs_need_rand_seg(sbi))
+		p->offset = prandom_u32() % (MAIN_SECS(sbi) * sbi->segs_per_sec);
+	else if (test_opt(sbi, NOHEAP) &&
 		(type == CURSEG_HOT_DATA || IS_NODESEG(type)))
 		p->offset = 0;
 	else

fs/f2fs/segment.c

@@ -15,6 +15,7 @@
 #include <linux/timer.h>
 #include <linux/freezer.h>
 #include <linux/sched/signal.h>
+#include <linux/random.h>
 
 #include "f2fs.h"
 #include "segment.h"
@@ -2649,6 +2650,8 @@ static unsigned int __get_next_segno(struct f2fs_sb_info *sbi, int type)
 	unsigned short seg_type = curseg->seg_type;
 
 	sanity_check_seg_type(sbi, seg_type);
+	if (f2fs_need_rand_seg(sbi))
+		return prandom_u32() % (MAIN_SECS(sbi) * sbi->segs_per_sec);
 
 	/* if segs_per_sec is large than 1, we need to keep original policy. */
 	if (__is_large_section(sbi))
@@ -2700,6 +2703,9 @@ static void new_curseg(struct f2fs_sb_info *sbi, int type, bool new_sec)
 	curseg->next_segno = segno;
 	reset_curseg(sbi, type, 1);
 	curseg->alloc_type = LFS;
+	if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK)
+		curseg->fragment_remained_chunk =
+				prandom_u32() % sbi->max_fragment_chunk + 1;
 }
 
 static int __next_free_blkoff(struct f2fs_sb_info *sbi,
@@ -2726,12 +2732,22 @@ static int __next_free_blkoff(struct f2fs_sb_info *sbi,
 static void __refresh_next_blkoff(struct f2fs_sb_info *sbi,
 				struct curseg_info *seg)
 {
-	if (seg->alloc_type == SSR)
+	if (seg->alloc_type == SSR) {
 		seg->next_blkoff =
 			__next_free_blkoff(sbi, seg->segno,
 						seg->next_blkoff + 1);
-	else
+	} else {
 		seg->next_blkoff++;
+		if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK) {
+			/* To allocate block chunks in different sizes, use random number */
+			if (--seg->fragment_remained_chunk <= 0) {
+				seg->fragment_remained_chunk =
+				   prandom_u32() % sbi->max_fragment_chunk + 1;
+				seg->next_blkoff +=
+				   prandom_u32() % sbi->max_fragment_hole + 1;
+			}
+		}
+	}
 }
 
 bool f2fs_segment_has_free_slot(struct f2fs_sb_info *sbi, int segno)

fs/f2fs/segment.h

@@ -314,6 +314,7 @@ struct curseg_info {
 	unsigned short next_blkoff;		/* next block offset to write */
 	unsigned int zone;			/* current zone number */
 	unsigned int next_segno;		/* preallocated segment */
+	int fragment_remained_chunk;		/* remained block size in a chunk for block fragmentation mode */
 	bool inited;				/* indicate inmem log is inited */
 };
 

fs/f2fs/super.c

@@ -817,6 +817,10 @@ static int parse_options(struct super_block *sb, char *options, bool is_remount)
 				F2FS_OPTION(sbi).fs_mode = FS_MODE_ADAPTIVE;
 			} else if (!strcmp(name, "lfs")) {
 				F2FS_OPTION(sbi).fs_mode = FS_MODE_LFS;
+			} else if (!strcmp(name, "fragment:segment")) {
+				F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_SEG;
+			} else if (!strcmp(name, "fragment:block")) {
+				F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_BLK;
 			} else {
 				kfree(name);
 				return -EINVAL;
@@ -1896,6 +1900,10 @@ static int f2fs_show_options(struct seq_file *seq, struct dentry *root)
 		seq_puts(seq, "adaptive");
 	else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_LFS)
 		seq_puts(seq, "lfs");
+	else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_SEG)
+		seq_puts(seq, "fragment:segment");
+	else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_BLK)
+		seq_puts(seq, "fragment:block");
 	seq_printf(seq, ",active_logs=%u", F2FS_OPTION(sbi).active_logs);
 	if (test_opt(sbi, RESERVE_ROOT))
 		seq_printf(seq, ",reserve_root=%u,resuid=%u,resgid=%u",
@@ -3523,6 +3531,8 @@ static void init_sb_info(struct f2fs_sb_info *sbi)
 	sbi->max_victim_search = DEF_MAX_VICTIM_SEARCH;
 	sbi->migration_granularity = sbi->segs_per_sec;
 	sbi->seq_file_ra_mul = MIN_RA_MUL;
+	sbi->max_fragment_chunk = DEF_FRAGMENT_SIZE;
+	sbi->max_fragment_hole = DEF_FRAGMENT_SIZE;
 
 	sbi->dir_level = DEF_DIR_LEVEL;
 	sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;

fs/f2fs/sysfs.c

@@ -551,6 +551,22 @@ static ssize_t __sbi_store(struct f2fs_attr *a,
 		return count;
 	}
 
+	if (!strcmp(a->attr.name, "max_fragment_chunk")) {
+		if (t >= MIN_FRAGMENT_SIZE && t <= MAX_FRAGMENT_SIZE)
+			sbi->max_fragment_chunk = t;
+		else
+			return -EINVAL;
+		return count;
+	}
+
+	if (!strcmp(a->attr.name, "max_fragment_hole")) {
+		if (t >= MIN_FRAGMENT_SIZE && t <= MAX_FRAGMENT_SIZE)
+			sbi->max_fragment_hole = t;
+		else
+			return -EINVAL;
+		return count;
+	}
+
 	*ui = (unsigned int)t;
 
 	return count;
@@ -781,6 +797,8 @@ F2FS_RW_ATTR(ATGC_INFO, atgc_management, atgc_age_threshold, age_threshold);
 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, seq_file_ra_mul, seq_file_ra_mul);
 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, gc_segment_mode, gc_segment_mode);
 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, gc_reclaimed_segments, gc_reclaimed_segs);
+F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_fragment_chunk, max_fragment_chunk);
+F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_fragment_hole, max_fragment_hole);
 
 #define ATTR_LIST(name) (&f2fs_attr_##name.attr)
 static struct attribute *f2fs_attrs[] = {
@@ -859,6 +877,8 @@ static struct attribute *f2fs_attrs[] = {
 	ATTR_LIST(seq_file_ra_mul),
 	ATTR_LIST(gc_segment_mode),
 	ATTR_LIST(gc_reclaimed_segments),
+	ATTR_LIST(max_fragment_chunk),
+	ATTR_LIST(max_fragment_hole),
 	NULL,
 };
 ATTRIBUTE_GROUPS(f2fs);