Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Fri, 10 Sep 2021 16:13:09 -0700
From:   Jaegeuk Kim <jaegeuk@kernel.org>
To:     Daeho Jeong <daeho43@gmail.com>
Cc:     linux-kernel@vger.kernel.org,
        linux-f2fs-devel@lists.sourceforge.net, kernel-team@android.com,
        Daeho Jeong <daehojeong@google.com>
Subject: Re: [PATCH v4] f2fs: introduce fragment allocation mode mount option
Message-ID: <YTvmhVhLlBPeASHT@google.com>
References: <20210902172404.3517626-1-daeho43@gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20210902172404.3517626-1-daeho43@gmail.com>
Precedence: bulk

On 09/02, Daeho Jeong wrote:
> From: Daeho Jeong <daehojeong@google.com>
> 
> Added three options into "mode=" mount option to make it possible for
> developers to make the filesystem fragmented or 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:fixed_block" : We can scatter block allocation with
> 		"fragment_chunk_size" and "fragment_hole_size" sysfs
> 		nodes. f2fs will allocate <fragment_chunk_size> blocks
> 		in a chunk and make a hole in the length of
> 		<fragment_hole_size> by turns in a newly allocated free
> 		segment.

Wait. Why do we need to add so many options here? I was expecting to see
performance difference when getting random segments or random blocks as
an extreme case. I don't get the point why we need the middle of those cases.

> "fragment:rand_block" : Working like "fragment:fixed_block" mode, but
> 		added some randomness to both chunk and hole size. So,
> 		f2fs will allocate 1..<fragment_chunk_size> blocks in a
> 		chunk and make a hole in the nodes. f2fs will allocate
> 		1..<fragment_chunk_size> blocks in a chunk and make a
> 		hole in the length of 1..<fragment_hole_size> by turns
> 		in a newly allocated free segment.
> 		Plus, f2fs implicitly enables "fragment:segment" option
> 		for more randomness in allocation in "fragment:rand_block".
> 
> Signed-off-by: Daeho Jeong <daehojeong@google.com>
> ---
> v4: implicitly enabled "fragment:segment" option only in
>     "fragment:rand_block".
> v3: divided "fragment:block" mode and fixed a race condition related to
>     making chunks.
> v2: changed mode name and added sysfs nodes to control the fragmentation
>     pattern.
> ---
>  Documentation/ABI/testing/sysfs-fs-f2fs | 24 ++++++++++++++++++++
>  Documentation/filesystems/f2fs.rst      | 22 +++++++++++++++++++
>  fs/f2fs/f2fs.h                          | 20 +++++++++++++++--
>  fs/f2fs/gc.c                            |  5 ++++-
>  fs/f2fs/segment.c                       | 29 +++++++++++++++++++++++--
>  fs/f2fs/segment.h                       |  1 +
>  fs/f2fs/super.c                         | 14 ++++++++++++
>  fs/f2fs/sysfs.c                         | 20 +++++++++++++++++
>  8 files changed, 130 insertions(+), 5 deletions(-)
> 
> diff --git a/Documentation/ABI/testing/sysfs-fs-f2fs b/Documentation/ABI/testing/sysfs-fs-f2fs
> index f627e705e663..d56ecfd16abf 100644
> --- a/Documentation/ABI/testing/sysfs-fs-f2fs
> +++ b/Documentation/ABI/testing/sysfs-fs-f2fs
> @@ -512,3 +512,27 @@ 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>/fragment_chunk_size
> +Date:		August 2021
> +Contact:	"Daeho Jeong" <daehojeong@google.com>
> +Description:	With "mode=fragment:fixed_block" and "mode=fragment:rand_block" mount options,
> +		we can scatter block allocation. Using this node, in "fragment:fixed_block"
> +		mode, f2fs will allocate <fragment_chunk_size> blocks in a chunk and make
> +		a hole in the length of	<fragment_hole_size> by turns in a newly allocated
> +		free segment. Plus, in "fragment:rand_block" mode, f2fs will allocate
> +		1..<fragment_chunk_size> blocks in a chunk and make a hole in the length of
> +		1..<fragment_hole_size> by turns. This value can be set between 1..512 and
> +		the default value is 4.
> +
> +What:		/sys/fs/f2fs/<disk>/fragment_hole_size
> +Date:		August 2021
> +Contact:	"Daeho Jeong" <daehojeong@google.com>
> +Description:	With "mode=fragment:fixed_block" and "mode=fragment:rand_block" mount options,
> +		we can scatter block allocation. Using this node, in "fragment:fixed_block"
> +		mode, f2fs will allocate <fragment_chunk_size> blocks in a chunk and make
> +		a hole in the length of	<fragment_hole_size> by turns in a newly allocated
> +		free segment. Plus, in "fragment:rand_block" mode, f2fs will allocate
> +		1..<fragment_chunk_size> blocks in a chunk and make a hole in the length of
> +		1..<fragment_hole_size> by turns. This value can be set between 1..512 and
> +		the default value is 4.
> diff --git a/Documentation/filesystems/f2fs.rst b/Documentation/filesystems/f2fs.rst
> index 09de6ebbbdfa..04ddae8754cc 100644
> --- a/Documentation/filesystems/f2fs.rst
> +++ b/Documentation/filesystems/f2fs.rst
> @@ -201,6 +201,28 @@ 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", "fragment:fixed_block" and "fragment:rand_block"
> +			 are newly added here. These are developer options for experiments
> +			 to make the filesystem fragmented or 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:fixed_block" and "fragment:rand_block", we can scatter
> +			 block allocation with "fragment_chunk_size" and "fragment_hole_size"
> +			 sysfs nodes. In "fragment:fixed_block" mode, f2fs will allocate
> +			 <fragment_chunk_size> blocks in a chunk and make a hole in the length
> +			 of <fragment_hole_size> by turns in a newly allocated free segment.
> +			 But, in "fragment:rand_block" mode, f2fs adds some randomness to
> +			 both chunk and hole size. So, f2fs will allocate
> +			 1..<fragment_chunk_size> blocks in a chunk and make a hole in the
> +			 length of 1..<fragment_hole_size> by turns. With these, the newly
> +			 allocated blocks will be scattered throughout the whole partition.
> +			 Plus, f2fs implicitly enables "fragment:segment" option for more
> +			 randomness in allocation in "fragment:rand_block".
> +			 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.
> diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
> index c24f03e054cb..f9aa00b92222 100644
> --- a/fs/f2fs/f2fs.h
> +++ b/fs/f2fs/f2fs.h
> @@ -1285,8 +1285,11 @@ enum {
>  };
>  
>  enum {
> -	FS_MODE_ADAPTIVE,	/* use both lfs/ssr allocation */
> -	FS_MODE_LFS,		/* use lfs allocation only */
> +	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_FIXED_BLK,	/* fixed block fragmentation mode */
> +	FS_MODE_FRAGMENT_RAND_BLK,	/* randomized block fragmentation mode */
>  };
>  
>  enum {
> @@ -1757,6 +1760,9 @@ struct f2fs_sb_info {
>  
>  	unsigned long seq_file_ra_mul;		/* multiplier for ra_pages of seq. files in fadvise */
>  
> +	int fragment_chunk_size;		/* the chunk size for block fragmentation mode */
> +	int fragment_hole_size;			/* the 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 */
> @@ -3517,6 +3523,16 @@ 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_RAND_BLK;
> +}
> +
>  /*
>   * checkpoint.c
>   */
> diff --git a/fs/f2fs/gc.c b/fs/f2fs/gc.c
> index 2c18443972b6..72cfad99bbbe 100644
> --- a/fs/f2fs/gc.c
> +++ b/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
> diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
> index a135d2247415..954c06728b94 100644
> --- a/fs/f2fs/segment.c
> +++ b/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"
> @@ -2630,6 +2631,8 @@ 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))
> @@ -2681,6 +2684,11 @@ 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_FIXED_BLK)
> +		curseg->fragment_remained_chunk = sbi->fragment_chunk_size;
> +	else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_RAND_BLK)
> +		curseg->fragment_remained_chunk =
> +				prandom_u32() % sbi->fragment_chunk_size + 1;
>  }
>  
>  static int __next_free_blkoff(struct f2fs_sb_info *sbi,
> @@ -2707,12 +2715,29 @@ 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_FIXED_BLK) {
> +			if (--seg->fragment_remained_chunk <= 0) {
> +				seg->fragment_remained_chunk =
> +				   sbi->fragment_chunk_size;
> +				seg->next_blkoff +=
> +				   sbi->fragment_hole_size;
> +			}
> +		} else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_RAND_BLK) {
> +			/* To allocate block chunks in different sizes, use random number */
> +			if (--seg->fragment_remained_chunk <= 0) {
> +				seg->fragment_remained_chunk =
> +				   prandom_u32() % sbi->fragment_chunk_size + 1;
> +				seg->next_blkoff +=
> +				   prandom_u32() % sbi->fragment_hole_size + 1;
> +			}
> +		}
> +	}
>  }
>  
>  bool f2fs_segment_has_free_slot(struct f2fs_sb_info *sbi, int segno)
> diff --git a/fs/f2fs/segment.h b/fs/f2fs/segment.h
> index 89fff258727d..46fde9f3f28e 100644
> --- a/fs/f2fs/segment.h
> +++ b/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 */
>  };
>  
> diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
> index 49e153fd8183..60067b6d9fea 100644
> --- a/fs/f2fs/super.c
> +++ b/fs/f2fs/super.c
> @@ -817,6 +817,12 @@ 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:fixed_block")) {
> +				F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_FIXED_BLK;
> +			} else if (!strcmp(name, "fragment:rand_block")) {
> +				F2FS_OPTION(sbi).fs_mode = FS_MODE_FRAGMENT_RAND_BLK;
>  			} else {
>  				kfree(name);
>  				return -EINVAL;
> @@ -1897,6 +1903,12 @@ 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_FIXED_BLK)
> +		seq_puts(seq, "fragment:fixed_block");
> +	else if (F2FS_OPTION(sbi).fs_mode == FS_MODE_FRAGMENT_RAND_BLK)
> +		seq_puts(seq, "fragment:rand_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",
> @@ -3515,6 +3527,8 @@ 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->fragment_chunk_size = DEF_FRAGMENT_SIZE;
> +	sbi->fragment_hole_size = DEF_FRAGMENT_SIZE;
>  
>  	sbi->dir_level = DEF_DIR_LEVEL;
>  	sbi->interval_time[CP_TIME] = DEF_CP_INTERVAL;
> diff --git a/fs/f2fs/sysfs.c b/fs/f2fs/sysfs.c
> index a1a3e0f6d658..ab34b3c2e09d 100644
> --- a/fs/f2fs/sysfs.c
> +++ b/fs/f2fs/sysfs.c
> @@ -551,6 +551,22 @@ static ssize_t __sbi_store(struct f2fs_attr *a,
>  		return count;
>  	}
>  
> +	if (!strcmp(a->attr.name, "fragment_chunk_size")) {
> +		if (t >= MIN_FRAGMENT_SIZE && t <= MAX_FRAGMENT_SIZE)
> +			sbi->fragment_chunk_size = t;
> +		else
> +			return -EINVAL;
> +		return count;
> +	}
> +
> +	if (!strcmp(a->attr.name, "fragment_hole_size")) {
> +		if (t >= MIN_FRAGMENT_SIZE && t <= MAX_FRAGMENT_SIZE)
> +			sbi->fragment_hole_size = 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, fragment_chunk_size, fragment_chunk_size);
> +F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, fragment_hole_size, fragment_hole_size);
>  
>  #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(fragment_chunk_size),
> +	ATTR_LIST(fragment_hole_size),
>  	NULL,
>  };
>  ATTRIBUTE_GROUPS(f2fs);
> -- 
> 2.33.0.153.gba50c8fa24-goog