[linux-block.git] / fs / ext4 / extents_status.h

// SPDX-License-Identifier: GPL-2.0
/*
 *  fs/ext4/extents_status.h
 *
 * Written by Yongqiang Yang <xiaoqiangnk@gmail.com>
 * Modified by
 *	Allison Henderson <achender@linux.vnet.ibm.com>
 *	Zheng Liu <wenqing.lz@taobao.com>
 *
 */

#ifndef _EXT4_EXTENTS_STATUS_H
#define _EXT4_EXTENTS_STATUS_H

/*
 * Turn on ES_DEBUG__ to get lots of info about extent status operations.
 */
#ifdef ES_DEBUG__
#define es_debug(fmt, ...)	printk(fmt, ##__VA_ARGS__)
#else
#define es_debug(fmt, ...)	no_printk(fmt, ##__VA_ARGS__)
#endif

/*
 * With ES_AGGRESSIVE_TEST defined, the result of es caching will be
 * checked with old map_block's result.
 */
#define ES_AGGRESSIVE_TEST__

/*
 * These flags live in the high bits of extent_status.es_pblk
 */
enum {
	ES_WRITTEN_B,
	ES_UNWRITTEN_B,
	ES_DELAYED_B,
	ES_HOLE_B,
	ES_REFERENCED_B,
	ES_FLAGS
};

#define ES_SHIFT (sizeof(ext4_fsblk_t)*8 - ES_FLAGS)
#define ES_MASK (~((ext4_fsblk_t)0) << ES_SHIFT)

#define EXTENT_STATUS_WRITTEN	(1 << ES_WRITTEN_B)
#define EXTENT_STATUS_UNWRITTEN (1 << ES_UNWRITTEN_B)
#define EXTENT_STATUS_DELAYED	(1 << ES_DELAYED_B)
#define EXTENT_STATUS_HOLE	(1 << ES_HOLE_B)
#define EXTENT_STATUS_REFERENCED	(1 << ES_REFERENCED_B)

#define ES_TYPE_MASK	((ext4_fsblk_t)(EXTENT_STATUS_WRITTEN | \
			  EXTENT_STATUS_UNWRITTEN | \
			  EXTENT_STATUS_DELAYED | \
			  EXTENT_STATUS_HOLE) << ES_SHIFT)

struct ext4_sb_info;
struct ext4_extent;

struct extent_status {
	struct rb_node rb_node;
	ext4_lblk_t es_lblk;	/* first logical block extent covers */
	ext4_lblk_t es_len;	/* length of extent in block */
	ext4_fsblk_t es_pblk;	/* first physical block */
};

struct ext4_es_tree {
	struct rb_root root;
	struct extent_status *cache_es;	/* recently accessed extent */
};

struct ext4_es_stats {
	unsigned long es_stats_shrunk;
	unsigned long es_stats_cache_hits;
	unsigned long es_stats_cache_misses;
	u64 es_stats_scan_time;
	u64 es_stats_max_scan_time;
	struct percpu_counter es_stats_all_cnt;
	struct percpu_counter es_stats_shk_cnt;
};

/*
 * Pending cluster reservations for bigalloc file systems
 *
 * A cluster with a pending reservation is a logical cluster shared by at
 * least one extent in the extents status tree with delayed and unwritten
 * status and at least one other written or unwritten extent.  The
 * reservation is said to be pending because a cluster reservation would
 * have to be taken in the event all blocks in the cluster shared with
 * written or unwritten extents were deleted while the delayed and
 * unwritten blocks remained.
 *
 * The set of pending cluster reservations is an auxiliary data structure
 * used with the extents status tree to implement reserved cluster/block
 * accounting for bigalloc file systems.  The set is kept in memory and
 * records all pending cluster reservations.
 *
 * Its primary function is to avoid the need to read extents from the
 * disk when invalidating pages as a result of a truncate, punch hole, or
 * collapse range operation.  Page invalidation requires a decrease in the
 * reserved cluster count if it results in the removal of all delayed
 * and unwritten extents (blocks) from a cluster that is not shared with a
 * written or unwritten extent, and no decrease otherwise.  Determining
 * whether the cluster is shared can be done by searching for a pending
 * reservation on it.
 *
 * Secondarily, it provides a potentially faster method for determining
 * whether the reserved cluster count should be increased when a physical
 * cluster is deallocated as a result of a truncate, punch hole, or
 * collapse range operation.  The necessary information is also present
 * in the extents status tree, but might be more rapidly accessed in
 * the pending reservation set in many cases due to smaller size.
 *
 * The pending cluster reservation set is implemented as a red-black tree
 * with the goal of minimizing per page search time overhead.
 */

struct pending_reservation {
	struct rb_node rb_node;
	ext4_lblk_t lclu;
};

struct ext4_pending_tree {
	struct rb_root root;
};

extern int __init ext4_init_es(void);
extern void ext4_exit_es(void);
extern void ext4_es_init_tree(struct ext4_es_tree *tree);

extern int ext4_es_insert_extent(struct inode *inode, ext4_lblk_t lblk,
				 ext4_lblk_t len, ext4_fsblk_t pblk,
				 unsigned int status);
extern void ext4_es_cache_extent(struct inode *inode, ext4_lblk_t lblk,
				 ext4_lblk_t len, ext4_fsblk_t pblk,
				 unsigned int status);
extern int ext4_es_remove_extent(struct inode *inode, ext4_lblk_t lblk,
				 ext4_lblk_t len);
extern void ext4_es_find_extent_range(struct inode *inode,
				      int (*match_fn)(struct extent_status *es),
				      ext4_lblk_t lblk, ext4_lblk_t end,
				      struct extent_status *es);
extern int ext4_es_lookup_extent(struct inode *inode, ext4_lblk_t lblk,
				 struct extent_status *es);
extern bool ext4_es_scan_range(struct inode *inode,
			       int (*matching_fn)(struct extent_status *es),
			       ext4_lblk_t lblk, ext4_lblk_t end);
extern bool ext4_es_scan_clu(struct inode *inode,
			     int (*matching_fn)(struct extent_status *es),
			     ext4_lblk_t lblk);

static inline unsigned int ext4_es_status(struct extent_status *es)
{
	return es->es_pblk >> ES_SHIFT;
}

static inline unsigned int ext4_es_type(struct extent_status *es)
{
	return (es->es_pblk & ES_TYPE_MASK) >> ES_SHIFT;
}

static inline int ext4_es_is_written(struct extent_status *es)
{
	return (ext4_es_type(es) & EXTENT_STATUS_WRITTEN) != 0;
}

static inline int ext4_es_is_unwritten(struct extent_status *es)
{
	return (ext4_es_type(es) & EXTENT_STATUS_UNWRITTEN) != 0;
}

static inline int ext4_es_is_delayed(struct extent_status *es)
{
	return (ext4_es_type(es) & EXTENT_STATUS_DELAYED) != 0;
}

static inline int ext4_es_is_hole(struct extent_status *es)
{
	return (ext4_es_type(es) & EXTENT_STATUS_HOLE) != 0;
}

static inline int ext4_es_is_mapped(struct extent_status *es)
{
	return (ext4_es_is_written(es) || ext4_es_is_unwritten(es));
}

static inline int ext4_es_is_delonly(struct extent_status *es)
{
	return (ext4_es_is_delayed(es) && !ext4_es_is_unwritten(es));
}

static inline void ext4_es_set_referenced(struct extent_status *es)
{
	es->es_pblk |= ((ext4_fsblk_t)EXTENT_STATUS_REFERENCED) << ES_SHIFT;
}

static inline void ext4_es_clear_referenced(struct extent_status *es)
{
	es->es_pblk &= ~(((ext4_fsblk_t)EXTENT_STATUS_REFERENCED) << ES_SHIFT);
}

static inline int ext4_es_is_referenced(struct extent_status *es)
{
	return (ext4_es_status(es) & EXTENT_STATUS_REFERENCED) != 0;
}

static inline ext4_fsblk_t ext4_es_pblock(struct extent_status *es)
{
	return es->es_pblk & ~ES_MASK;
}

static inline void ext4_es_store_pblock(struct extent_status *es,
					ext4_fsblk_t pb)
{
	ext4_fsblk_t block;

	block = (pb & ~ES_MASK) | (es->es_pblk & ES_MASK);
	es->es_pblk = block;
}

static inline void ext4_es_store_status(struct extent_status *es,
					unsigned int status)
{
	es->es_pblk = (((ext4_fsblk_t)status << ES_SHIFT) & ES_MASK) |
		      (es->es_pblk & ~ES_MASK);
}

static inline void ext4_es_store_pblock_status(struct extent_status *es,
					       ext4_fsblk_t pb,
					       unsigned int status)
{
	es->es_pblk = (((ext4_fsblk_t)status << ES_SHIFT) & ES_MASK) |
		      (pb & ~ES_MASK);
}

extern int ext4_es_register_shrinker(struct ext4_sb_info *sbi);
extern void ext4_es_unregister_shrinker(struct ext4_sb_info *sbi);

extern int ext4_seq_es_shrinker_info_show(struct seq_file *seq, void *v);

extern int __init ext4_init_pending(void);
extern void ext4_exit_pending(void);
extern void ext4_init_pending_tree(struct ext4_pending_tree *tree);
extern void ext4_remove_pending(struct inode *inode, ext4_lblk_t lblk);
extern bool ext4_is_pending(struct inode *inode, ext4_lblk_t lblk);
extern int ext4_es_insert_delayed_block(struct inode *inode, ext4_lblk_t lblk,
					bool allocated);
extern unsigned int ext4_es_delayed_clu(struct inode *inode, ext4_lblk_t lblk,
					ext4_lblk_t len);
extern void ext4_es_remove_blks(struct inode *inode, ext4_lblk_t lblk,
				ext4_lblk_t len);

#endif /* _EXT4_EXTENTS_STATUS_H */
Commit	Line	Data
f5166768	1	// SPDX-License-Identifier: GPL-2.0
c0677e6d ZL	2	/*
	3	* fs/ext4/extents_status.h
	4	*
	5	* Written by Yongqiang Yang <xiaoqiangnk@gmail.com>
	6	* Modified by
	7	* Allison Henderson <achender@linux.vnet.ibm.com>
	8	* Zheng Liu <wenqing.lz@taobao.com>
	9	*
	10	*/
	11
	12	#ifndef _EXT4_EXTENTS_STATUS_H
	13	#define _EXT4_EXTENTS_STATUS_H
	14
654598be ZL	15	/*
	16	* Turn on ES_DEBUG__ to get lots of info about extent status operations.
	17	*/
	18	#ifdef ES_DEBUG__
	19	#define es_debug(fmt, ...) printk(fmt, ##__VA_ARGS__)
	20	#else
	21	#define es_debug(fmt, ...) no_printk(fmt, ##__VA_ARGS__)
	22	#endif
	23
921f266b DM	24	/*
	25	* With ES_AGGRESSIVE_TEST defined, the result of es caching will be
	26	* checked with old map_block's result.
	27	*/
	28	#define ES_AGGRESSIVE_TEST__
	29
8e919d13 TT	30	/*
	31	* These flags live in the high bits of extent_status.es_pblk
	32	*/
624d0f1d JK	33	enum {
	34	ES_WRITTEN_B,
	35	ES_UNWRITTEN_B,
	36	ES_DELAYED_B,
	37	ES_HOLE_B,
2be12de9	38	ES_REFERENCED_B,
624d0f1d JK	39	ES_FLAGS
624d0f1d JK	40	};
fdc0212e	41
624d0f1d JK	42	#define ES_SHIFT (sizeof(ext4_fsblk_t)*8 - ES_FLAGS)
624d0f1d JK	43	#define ES_MASK (~((ext4_fsblk_t)0) << ES_SHIFT)
3be78c73	44
624d0f1d JK	45	#define EXTENT_STATUS_WRITTEN (1 << ES_WRITTEN_B)
	46	#define EXTENT_STATUS_UNWRITTEN (1 << ES_UNWRITTEN_B)
	47	#define EXTENT_STATUS_DELAYED (1 << ES_DELAYED_B)
	48	#define EXTENT_STATUS_HOLE (1 << ES_HOLE_B)
2be12de9 JK	49	#define EXTENT_STATUS_REFERENCED (1 << ES_REFERENCED_B)
	50
	51	#define ES_TYPE_MASK ((ext4_fsblk_t)(EXTENT_STATUS_WRITTEN \| \
	52	EXTENT_STATUS_UNWRITTEN \| \
	53	EXTENT_STATUS_DELAYED \| \
	54	EXTENT_STATUS_HOLE) << ES_SHIFT)
3be78c73	55
d3922a77	56	struct ext4_sb_info;
adb23551 ZL	57	struct ext4_extent;
adb23551 ZL	58
c0677e6d ZL	59	struct extent_status {
c0677e6d ZL	60	struct rb_node rb_node;
06b0c886 ZL	61	ext4_lblk_t es_lblk; /* first logical block extent covers */
06b0c886 ZL	62	ext4_lblk_t es_len; /* length of extent in block */
fdc0212e	63	ext4_fsblk_t es_pblk; /* first physical block */
c0677e6d ZL	64	};
	65
	66	struct ext4_es_tree {
	67	struct rb_root root;
	68	struct extent_status cache_es; / recently accessed extent */
	69	};
	70
eb68d0e2	71	struct ext4_es_stats {
eb68d0e2 ZL	72	unsigned long es_stats_shrunk;
	73	unsigned long es_stats_cache_hits;
	74	unsigned long es_stats_cache_misses;
	75	u64 es_stats_scan_time;
	76	u64 es_stats_max_scan_time;
	77	struct percpu_counter es_stats_all_cnt;
edaa53ca	78	struct percpu_counter es_stats_shk_cnt;
eb68d0e2 ZL	79	};
eb68d0e2 ZL	80
1dc0aa46 EW	81	/*
	82	* Pending cluster reservations for bigalloc file systems
	83	*
	84	* A cluster with a pending reservation is a logical cluster shared by at
	85	* least one extent in the extents status tree with delayed and unwritten
	86	* status and at least one other written or unwritten extent. The
	87	* reservation is said to be pending because a cluster reservation would
	88	* have to be taken in the event all blocks in the cluster shared with
	89	* written or unwritten extents were deleted while the delayed and
	90	* unwritten blocks remained.
	91	*
	92	* The set of pending cluster reservations is an auxiliary data structure
	93	* used with the extents status tree to implement reserved cluster/block
	94	* accounting for bigalloc file systems. The set is kept in memory and
	95	* records all pending cluster reservations.
	96	*
	97	* Its primary function is to avoid the need to read extents from the
	98	* disk when invalidating pages as a result of a truncate, punch hole, or
	99	* collapse range operation. Page invalidation requires a decrease in the
	100	* reserved cluster count if it results in the removal of all delayed
	101	* and unwritten extents (blocks) from a cluster that is not shared with a
	102	* written or unwritten extent, and no decrease otherwise. Determining
	103	* whether the cluster is shared can be done by searching for a pending
	104	* reservation on it.
	105	*
	106	* Secondarily, it provides a potentially faster method for determining
	107	* whether the reserved cluster count should be increased when a physical
	108	* cluster is deallocated as a result of a truncate, punch hole, or
	109	* collapse range operation. The necessary information is also present
	110	* in the extents status tree, but might be more rapidly accessed in
	111	* the pending reservation set in many cases due to smaller size.
	112	*
	113	* The pending cluster reservation set is implemented as a red-black tree
	114	* with the goal of minimizing per page search time overhead.
	115	*/
	116
	117	struct pending_reservation {
	118	struct rb_node rb_node;
	119	ext4_lblk_t lclu;
	120	};
	121
	122	struct ext4_pending_tree {
	123	struct rb_root root;
	124	};
	125
654598be ZL	126	extern int __init ext4_init_es(void);
	127	extern void ext4_exit_es(void);
	128	extern void ext4_es_init_tree(struct ext4_es_tree *tree);
	129
06b0c886	130	extern int ext4_es_insert_extent(struct inode *inode, ext4_lblk_t lblk,
fdc0212e	131	ext4_lblk_t len, ext4_fsblk_t pblk,
3be78c73	132	unsigned int status);
107a7bd3 TT	133	extern void ext4_es_cache_extent(struct inode *inode, ext4_lblk_t lblk,
	134	ext4_lblk_t len, ext4_fsblk_t pblk,
	135	unsigned int status);
06b0c886	136	extern int ext4_es_remove_extent(struct inode *inode, ext4_lblk_t lblk,
654598be	137	ext4_lblk_t len);
ad431025 EW	138	extern void ext4_es_find_extent_range(struct inode *inode,
	139	int (match_fn)(struct extent_status es),
	140	ext4_lblk_t lblk, ext4_lblk_t end,
	141	struct extent_status *es);
d100eef2 ZL	142	extern int ext4_es_lookup_extent(struct inode *inode, ext4_lblk_t lblk,
d100eef2 ZL	143	struct extent_status *es);
ad431025 EW	144	extern bool ext4_es_scan_range(struct inode *inode,
	145	int (matching_fn)(struct extent_status es),
	146	ext4_lblk_t lblk, ext4_lblk_t end);
	147	extern bool ext4_es_scan_clu(struct inode *inode,
	148	int (matching_fn)(struct extent_status es),
	149	ext4_lblk_t lblk);
654598be	150
624d0f1d JK	151	static inline unsigned int ext4_es_status(struct extent_status *es)
	152	{
	153	return es->es_pblk >> ES_SHIFT;
	154	}
	155
2be12de9 JK	156	static inline unsigned int ext4_es_type(struct extent_status *es)
	157	{
	158	return (es->es_pblk & ES_TYPE_MASK) >> ES_SHIFT;
	159	}
	160
fdc0212e ZL	161	static inline int ext4_es_is_written(struct extent_status *es)
fdc0212e ZL	162	{
2be12de9	163	return (ext4_es_type(es) & EXTENT_STATUS_WRITTEN) != 0;
fdc0212e ZL	164	}
	165
	166	static inline int ext4_es_is_unwritten(struct extent_status *es)
	167	{
2be12de9	168	return (ext4_es_type(es) & EXTENT_STATUS_UNWRITTEN) != 0;
fdc0212e ZL	169	}
	170
	171	static inline int ext4_es_is_delayed(struct extent_status *es)
	172	{
2be12de9	173	return (ext4_es_type(es) & EXTENT_STATUS_DELAYED) != 0;
fdc0212e ZL	174	}
	175
	176	static inline int ext4_es_is_hole(struct extent_status *es)
	177	{
2be12de9 JK	178	return (ext4_es_type(es) & EXTENT_STATUS_HOLE) != 0;
	179	}
	180
0b02f4c0 EW	181	static inline int ext4_es_is_mapped(struct extent_status *es)
	182	{
	183	return (ext4_es_is_written(es) \|\| ext4_es_is_unwritten(es));
	184	}
	185
	186	static inline int ext4_es_is_delonly(struct extent_status *es)
	187	{
	188	return (ext4_es_is_delayed(es) && !ext4_es_is_unwritten(es));
	189	}
	190
2be12de9 JK	191	static inline void ext4_es_set_referenced(struct extent_status *es)
	192	{
	193	es->es_pblk \|= ((ext4_fsblk_t)EXTENT_STATUS_REFERENCED) << ES_SHIFT;
	194	}
	195
	196	static inline void ext4_es_clear_referenced(struct extent_status *es)
	197	{
	198	es->es_pblk &= ~(((ext4_fsblk_t)EXTENT_STATUS_REFERENCED) << ES_SHIFT);
	199	}
	200
	201	static inline int ext4_es_is_referenced(struct extent_status *es)
	202	{
	203	return (ext4_es_status(es) & EXTENT_STATUS_REFERENCED) != 0;
fdc0212e ZL	204	}
	205
	206	static inline ext4_fsblk_t ext4_es_pblock(struct extent_status *es)
	207	{
3be78c73	208	return es->es_pblk & ~ES_MASK;
fdc0212e ZL	209	}
	210
	211	static inline void ext4_es_store_pblock(struct extent_status *es,
	212	ext4_fsblk_t pb)
	213	{
	214	ext4_fsblk_t block;
	215
3be78c73	216	block = (pb & ~ES_MASK) \| (es->es_pblk & ES_MASK);
fdc0212e ZL	217	es->es_pblk = block;
	218	}
	219
	220	static inline void ext4_es_store_status(struct extent_status *es,
3be78c73	221	unsigned int status)
fdc0212e	222	{
624d0f1d JK	223	es->es_pblk = (((ext4_fsblk_t)status << ES_SHIFT) & ES_MASK) \|
624d0f1d JK	224	(es->es_pblk & ~ES_MASK);
fdc0212e ZL	225	}
fdc0212e ZL	226
9a6633b1 TT	227	static inline void ext4_es_store_pblock_status(struct extent_status *es,
	228	ext4_fsblk_t pb,
	229	unsigned int status)
	230	{
624d0f1d JK	231	es->es_pblk = (((ext4_fsblk_t)status << ES_SHIFT) & ES_MASK) \|
624d0f1d JK	232	(pb & ~ES_MASK);
9a6633b1 TT	233	}
9a6633b1 TT	234
eb68d0e2	235	extern int ext4_es_register_shrinker(struct ext4_sb_info *sbi);
d3922a77	236	extern void ext4_es_unregister_shrinker(struct ext4_sb_info *sbi);
74cd15cd	237
ebd173be TT	238	extern int ext4_seq_es_shrinker_info_show(struct seq_file seq, void v);
ebd173be TT	239
1dc0aa46 EW	240	extern int __init ext4_init_pending(void);
	241	extern void ext4_exit_pending(void);
	242	extern void ext4_init_pending_tree(struct ext4_pending_tree *tree);
	243	extern void ext4_remove_pending(struct inode *inode, ext4_lblk_t lblk);
	244	extern bool ext4_is_pending(struct inode *inode, ext4_lblk_t lblk);
0b02f4c0 EW	245	extern int ext4_es_insert_delayed_block(struct inode *inode, ext4_lblk_t lblk,
0b02f4c0 EW	246	bool allocated);
b6bf9171 EW	247	extern unsigned int ext4_es_delayed_clu(struct inode *inode, ext4_lblk_t lblk,
	248	ext4_lblk_t len);
	249	extern void ext4_es_remove_blks(struct inode *inode, ext4_lblk_t lblk,
	250	ext4_lblk_t len);
1dc0aa46	251
c0677e6d	252	#endif /* _EXT4_EXTENTS_STATUS_H */