[deliverable/linux.git] / fs / ext4 / mballoc.h

/*
 *  fs/ext4/mballoc.h
 *
 *  Written by: Alex Tomas <alex@clusterfs.com>
 *
 */
#ifndef _EXT4_MBALLOC_H
#define _EXT4_MBALLOC_H

#include <linux/time.h>
#include <linux/fs.h>
#include <linux/namei.h>
#include <linux/quotaops.h>
#include <linux/buffer_head.h>
#include <linux/module.h>
#include <linux/swap.h>
#include <linux/proc_fs.h>
#include <linux/pagemap.h>
#include <linux/seq_file.h>
#include <linux/version.h>
#include "ext4_jbd2.h"
#include "ext4.h"
#include "group.h"

/*
 * with AGGRESSIVE_CHECK allocator runs consistency checks over
 * structures. these checks slow things down a lot
 */
#define AGGRESSIVE_CHECK__

/*
 * with DOUBLE_CHECK defined mballoc creates persistent in-core
 * bitmaps, maintains and uses them to check for double allocations
 */
#define DOUBLE_CHECK__

/*
 */
#define MB_DEBUG__
#ifdef MB_DEBUG
#define mb_debug(fmt, a...)	printk(fmt, ##a)
#else
#define mb_debug(fmt, a...)
#endif

/*
 * with EXT4_MB_HISTORY mballoc stores last N allocations in memory
 * and you can monitor it in /proc/fs/ext4/<dev>/mb_history
 */
#define EXT4_MB_HISTORY
#define EXT4_MB_HISTORY_ALLOC		1	/* allocation */
#define EXT4_MB_HISTORY_PREALLOC	2	/* preallocated blocks used */
#define EXT4_MB_HISTORY_DISCARD		4	/* preallocation discarded */
#define EXT4_MB_HISTORY_FREE		8	/* free */

#define EXT4_MB_HISTORY_DEFAULT		(EXT4_MB_HISTORY_ALLOC | \
					 EXT4_MB_HISTORY_PREALLOC)

/*
 * How long mballoc can look for a best extent (in found extents)
 */
#define MB_DEFAULT_MAX_TO_SCAN		200

/*
 * How long mballoc must look for a best extent
 */
#define MB_DEFAULT_MIN_TO_SCAN		10

/*
 * How many groups mballoc will scan looking for the best chunk
 */
#define MB_DEFAULT_MAX_GROUPS_TO_SCAN	5

/*
 * with 'ext4_mb_stats' allocator will collect stats that will be
 * shown at umount. The collecting costs though!
 */
#define MB_DEFAULT_STATS		1

/*
 * files smaller than MB_DEFAULT_STREAM_THRESHOLD are served
 * by the stream allocator, which purpose is to pack requests
 * as close each to other as possible to produce smooth I/O traffic
 * We use locality group prealloc space for stream request.
 * We can tune the same via /proc/fs/ext4/<parition>/stream_req
 */
#define MB_DEFAULT_STREAM_THRESHOLD	16	/* 64K */

/*
 * for which requests use 2^N search using buddies
 */
#define MB_DEFAULT_ORDER2_REQS		2

/*
 * default group prealloc size 512 blocks
 */
#define MB_DEFAULT_GROUP_PREALLOC	512

static struct kmem_cache *ext4_pspace_cachep;
static struct kmem_cache *ext4_ac_cachep;

#ifdef EXT4_BB_MAX_BLOCKS
#undef EXT4_BB_MAX_BLOCKS
#endif
#define EXT4_BB_MAX_BLOCKS	30

struct ext4_free_metadata {
	ext4_group_t group;
	unsigned short num;
	ext4_grpblk_t  blocks[EXT4_BB_MAX_BLOCKS];
	struct list_head list;
};

struct ext4_group_info {
	unsigned long	bb_state;
	unsigned long	bb_tid;
	struct ext4_free_metadata *bb_md_cur;
	unsigned short	bb_first_free;
	unsigned short	bb_free;
	unsigned short	bb_fragments;
	struct		list_head bb_prealloc_list;
#ifdef DOUBLE_CHECK
	void		*bb_bitmap;
#endif
	unsigned short	bb_counters[];
};

#define EXT4_GROUP_INFO_NEED_INIT_BIT	0
#define EXT4_GROUP_INFO_LOCKED_BIT	1

#define EXT4_MB_GRP_NEED_INIT(grp)	\
	(test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))


struct ext4_prealloc_space {
	struct list_head	pa_inode_list;
	struct list_head	pa_group_list;
	union {
		struct list_head pa_tmp_list;
		struct rcu_head	pa_rcu;
	} u;
	spinlock_t		pa_lock;
	atomic_t		pa_count;
	unsigned		pa_deleted;
	ext4_fsblk_t		pa_pstart;	/* phys. block */
	ext4_lblk_t		pa_lstart;	/* log. block */
	unsigned short		pa_len;		/* len of preallocated chunk */
	unsigned short		pa_free;	/* how many blocks are free */
	unsigned short		pa_linear;	/* consumed in one direction
						 * strictly, for grp prealloc */
	spinlock_t		*pa_obj_lock;
	struct inode		*pa_inode;	/* hack, for history only */
};


struct ext4_free_extent {
	ext4_lblk_t fe_logical;
	ext4_grpblk_t fe_start;
	ext4_group_t fe_group;
	int fe_len;
};

/*
 * Locality group:
 *   we try to group all related changes together
 *   so that writeback can flush/allocate them together as well
 *   Size of lg_prealloc_list hash is determined by MB_DEFAULT_GROUP_PREALLOC
 *   (512). We store prealloc space into the hash based on the pa_free blocks
 *   order value.ie, fls(pa_free)-1;
 */
#define PREALLOC_TB_SIZE 10
struct ext4_locality_group {
	/* for allocator */
	/* to serialize allocates */
	struct mutex		lg_mutex;
	/* list of preallocations */
	struct list_head	lg_prealloc_list[PREALLOC_TB_SIZE];
	spinlock_t		lg_prealloc_lock;
};

struct ext4_allocation_context {
	struct inode *ac_inode;
	struct super_block *ac_sb;

	/* original request */
	struct ext4_free_extent ac_o_ex;

	/* goal request (after normalization) */
	struct ext4_free_extent ac_g_ex;

	/* the best found extent */
	struct ext4_free_extent ac_b_ex;

	/* copy of the bext found extent taken before preallocation efforts */
	struct ext4_free_extent ac_f_ex;

	/* number of iterations done. we have to track to limit searching */
	unsigned long ac_ex_scanned;
	__u16 ac_groups_scanned;
	__u16 ac_found;
	__u16 ac_tail;
	__u16 ac_buddy;
	__u16 ac_flags;		/* allocation hints */
	__u8 ac_status;
	__u8 ac_criteria;
	__u8 ac_repeats;
	__u8 ac_2order;		/* if request is to allocate 2^N blocks and
				 * N > 0, the field stores N, otherwise 0 */
	__u8 ac_op;		/* operation, for history only */
	struct page *ac_bitmap_page;
	struct page *ac_buddy_page;
	struct ext4_prealloc_space *ac_pa;
	struct ext4_locality_group *ac_lg;
};

#define AC_STATUS_CONTINUE	1
#define AC_STATUS_FOUND		2
#define AC_STATUS_BREAK		3

struct ext4_mb_history {
	struct ext4_free_extent orig;	/* orig allocation */
	struct ext4_free_extent goal;	/* goal allocation */
	struct ext4_free_extent result;	/* result allocation */
	unsigned pid;
	unsigned ino;
	__u16 found;	/* how many extents have been found */
	__u16 groups;	/* how many groups have been scanned */
	__u16 tail;	/* what tail broke some buddy */
	__u16 buddy;	/* buddy the tail ^^^ broke */
	__u16 flags;
	__u8 cr:3;	/* which phase the result extent was found at */
	__u8 op:4;
	__u8 merged:1;
};

struct ext4_buddy {
	struct page *bd_buddy_page;
	void *bd_buddy;
	struct page *bd_bitmap_page;
	void *bd_bitmap;
	struct ext4_group_info *bd_info;
	struct super_block *bd_sb;
	__u16 bd_blkbits;
	ext4_group_t bd_group;
};
#define EXT4_MB_BITMAP(e4b)	((e4b)->bd_bitmap)
#define EXT4_MB_BUDDY(e4b)	((e4b)->bd_buddy)

#ifndef EXT4_MB_HISTORY
static inline void ext4_mb_store_history(struct ext4_allocation_context *ac)
{
	return;
}
#else
static void ext4_mb_store_history(struct ext4_allocation_context *ac);
#endif

#define in_range(b, first, len)	((b) >= (first) && (b) <= (first) + (len) - 1)

struct buffer_head *read_block_bitmap(struct super_block *, ext4_group_t);

static void ext4_mb_generate_from_pa(struct super_block *sb, void *bitmap,
					ext4_group_t group);
static void ext4_mb_poll_new_transaction(struct super_block *, handle_t *);
static void ext4_mb_free_committed_blocks(struct super_block *);
static void ext4_mb_return_to_preallocation(struct inode *inode,
					struct ext4_buddy *e4b, sector_t block,
					int count);
static void ext4_mb_put_pa(struct ext4_allocation_context *,
			struct super_block *, struct ext4_prealloc_space *pa);
static int ext4_mb_init_per_dev_proc(struct super_block *sb);
static int ext4_mb_destroy_per_dev_proc(struct super_block *sb);


static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
{
	struct ext4_group_info *grinfo = ext4_get_group_info(sb, group);

	bit_spin_lock(EXT4_GROUP_INFO_LOCKED_BIT, &(grinfo->bb_state));
}

static inline void ext4_unlock_group(struct super_block *sb,
					ext4_group_t group)
{
	struct ext4_group_info *grinfo = ext4_get_group_info(sb, group);

	bit_spin_unlock(EXT4_GROUP_INFO_LOCKED_BIT, &(grinfo->bb_state));
}

static inline int ext4_is_group_locked(struct super_block *sb,
					ext4_group_t group)
{
	struct ext4_group_info *grinfo = ext4_get_group_info(sb, group);

	return bit_spin_is_locked(EXT4_GROUP_INFO_LOCKED_BIT,
						&(grinfo->bb_state));
}

static ext4_fsblk_t ext4_grp_offs_to_block(struct super_block *sb,
					struct ext4_free_extent *fex)
{
	ext4_fsblk_t block;

	block = (ext4_fsblk_t) fex->fe_group * EXT4_BLOCKS_PER_GROUP(sb)
			+ fex->fe_start
			+ le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
	return block;
}
#endif
Commit	Line	Data
8f6e39a7 MC	1	/*
	2	* fs/ext4/mballoc.h
	3	*
	4	* Written by: Alex Tomas <alex@clusterfs.com>
	5	*
	6	*/
	7	#ifndef _EXT4_MBALLOC_H
	8	#define _EXT4_MBALLOC_H
	9
	10	#include <linux/time.h>
	11	#include <linux/fs.h>
	12	#include <linux/namei.h>
	13	#include <linux/quotaops.h>
	14	#include <linux/buffer_head.h>
	15	#include <linux/module.h>
	16	#include <linux/swap.h>
	17	#include <linux/proc_fs.h>
	18	#include <linux/pagemap.h>
	19	#include <linux/seq_file.h>
	20	#include <linux/version.h>
	21	#include "ext4_jbd2.h"
	22	#include "ext4.h"
	23	#include "group.h"
	24
	25	/*
	26	* with AGGRESSIVE_CHECK allocator runs consistency checks over
	27	* structures. these checks slow things down a lot
	28	*/
	29	#define AGGRESSIVE_CHECK__
	30
	31	/*
	32	* with DOUBLE_CHECK defined mballoc creates persistent in-core
	33	* bitmaps, maintains and uses them to check for double allocations
	34	*/
	35	#define DOUBLE_CHECK__
	36
	37	/*
	38	*/
	39	#define MB_DEBUG__
	40	#ifdef MB_DEBUG
	41	#define mb_debug(fmt, a...) printk(fmt, ##a)
	42	#else
	43	#define mb_debug(fmt, a...)
	44	#endif
	45
	46	/*
	47	* with EXT4_MB_HISTORY mballoc stores last N allocations in memory
	48	* and you can monitor it in /proc/fs/ext4/<dev>/mb_history
	49	*/
	50	#define EXT4_MB_HISTORY
	51	#define EXT4_MB_HISTORY_ALLOC 1 /* allocation */
	52	#define EXT4_MB_HISTORY_PREALLOC 2 /* preallocated blocks used */
	53	#define EXT4_MB_HISTORY_DISCARD 4 /* preallocation discarded */
	54	#define EXT4_MB_HISTORY_FREE 8 /* free */
	55
	56	#define EXT4_MB_HISTORY_DEFAULT (EXT4_MB_HISTORY_ALLOC \| \
	57	EXT4_MB_HISTORY_PREALLOC)
	58
	59	/*
	60	* How long mballoc can look for a best extent (in found extents)
	61	*/
	62	#define MB_DEFAULT_MAX_TO_SCAN 200
	63
	64	/*
65	* How long mballoc must look for a best extent
66	*/
67	#define MB_DEFAULT_MIN_TO_SCAN 10
68
69	/*
70	* How many groups mballoc will scan looking for the best chunk
71	*/
72	#define MB_DEFAULT_MAX_GROUPS_TO_SCAN 5
73
74	/*
75	* with 'ext4_mb_stats' allocator will collect stats that will be
76	* shown at umount. The collecting costs though!
77	*/
78	#define MB_DEFAULT_STATS 1
79
80	/*
81	* files smaller than MB_DEFAULT_STREAM_THRESHOLD are served
82	* by the stream allocator, which purpose is to pack requests
83	* as close each to other as possible to produce smooth I/O traffic
84	* We use locality group prealloc space for stream request.
85	* We can tune the same via /proc/fs/ext4/<parition>/stream_req
86	*/
87	#define MB_DEFAULT_STREAM_THRESHOLD 16 /* 64K */
88
89	/*
90	* for which requests use 2^N search using buddies
91	*/
92	#define MB_DEFAULT_ORDER2_REQS 2
93
94	/*
95	* default group prealloc size 512 blocks
96	*/
97	#define MB_DEFAULT_GROUP_PREALLOC 512
98
99	static struct kmem_cache *ext4_pspace_cachep;
100	static struct kmem_cache *ext4_ac_cachep;
101
102	#ifdef EXT4_BB_MAX_BLOCKS
103	#undef EXT4_BB_MAX_BLOCKS
104	#endif
105	#define EXT4_BB_MAX_BLOCKS 30
106
107	struct ext4_free_metadata {
108	ext4_group_t group;
109	unsigned short num;
110	ext4_grpblk_t blocks[EXT4_BB_MAX_BLOCKS];
111	struct list_head list;
112	};
113
114	struct ext4_group_info {
115	unsigned long bb_state;
116	unsigned long bb_tid;
117	struct ext4_free_metadata *bb_md_cur;
118	unsigned short bb_first_free;
119	unsigned short bb_free;
120	unsigned short bb_fragments;
121	struct list_head bb_prealloc_list;
122	#ifdef DOUBLE_CHECK
123	void *bb_bitmap;
124	#endif
125	unsigned short bb_counters[];
126	};
127
128	#define EXT4_GROUP_INFO_NEED_INIT_BIT 0
129	#define EXT4_GROUP_INFO_LOCKED_BIT 1
130
131	#define EXT4_MB_GRP_NEED_INIT(grp) \
132	(test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
133
134
135	struct ext4_prealloc_space {
136	struct list_head pa_inode_list;
137	struct list_head pa_group_list;
138	union {
139	struct list_head pa_tmp_list;
140	struct rcu_head pa_rcu;
141	} u;
142	spinlock_t pa_lock;
143	atomic_t pa_count;
144	unsigned pa_deleted;
145	ext4_fsblk_t pa_pstart; /* phys. block */
146	ext4_lblk_t pa_lstart; /* log. block */
147	unsigned short pa_len; /* len of preallocated chunk */
148	unsigned short pa_free; /* how many blocks are free */
149	unsigned short pa_linear; /* consumed in one direction
150	* strictly, for grp prealloc */
151	spinlock_t *pa_obj_lock;
152	struct inode pa_inode; / hack, for history only */
153	};
154
155
156	struct ext4_free_extent {
157	ext4_lblk_t fe_logical;
158	ext4_grpblk_t fe_start;
159	ext4_group_t fe_group;
160	int fe_len;
161	};
162
163	/*
164	* Locality group:
165	* we try to group all related changes together
166	* so that writeback can flush/allocate them together as well
6be2ded1 AK	167	* Size of lg_prealloc_list hash is determined by MB_DEFAULT_GROUP_PREALLOC
	168	* (512). We store prealloc space into the hash based on the pa_free blocks
	169	* order value.ie, fls(pa_free)-1;
8f6e39a7	170	*/
6be2ded1	171	#define PREALLOC_TB_SIZE 10
8f6e39a7 MC	172	struct ext4_locality_group {
8f6e39a7 MC	173	/* for allocator */
6be2ded1 AK	174	/* to serialize allocates */
	175	struct mutex lg_mutex;
	176	/* list of preallocations */
	177	struct list_head lg_prealloc_list[PREALLOC_TB_SIZE];
8f6e39a7 MC	178	spinlock_t lg_prealloc_lock;
	179	};
	180
	181	struct ext4_allocation_context {
	182	struct inode *ac_inode;
	183	struct super_block *ac_sb;
	184
	185	/* original request */
	186	struct ext4_free_extent ac_o_ex;
	187
	188	/* goal request (after normalization) */
	189	struct ext4_free_extent ac_g_ex;
	190
	191	/* the best found extent */
	192	struct ext4_free_extent ac_b_ex;
	193
	194	/* copy of the bext found extent taken before preallocation efforts */
	195	struct ext4_free_extent ac_f_ex;
	196
	197	/* number of iterations done. we have to track to limit searching */
	198	unsigned long ac_ex_scanned;
	199	__u16 ac_groups_scanned;
	200	__u16 ac_found;
	201	__u16 ac_tail;
	202	__u16 ac_buddy;
	203	__u16 ac_flags; /* allocation hints */
	204	__u8 ac_status;
	205	__u8 ac_criteria;
	206	__u8 ac_repeats;
	207	__u8 ac_2order; /* if request is to allocate 2^N blocks and
	208	* N > 0, the field stores N, otherwise 0 */
	209	__u8 ac_op; /* operation, for history only */
	210	struct page *ac_bitmap_page;
	211	struct page *ac_buddy_page;
	212	struct ext4_prealloc_space *ac_pa;
	213	struct ext4_locality_group *ac_lg;
	214	};
	215
	216	#define AC_STATUS_CONTINUE 1
	217	#define AC_STATUS_FOUND 2
	218	#define AC_STATUS_BREAK 3
	219
	220	struct ext4_mb_history {
	221	struct ext4_free_extent orig; /* orig allocation */
	222	struct ext4_free_extent goal; /* goal allocation */
	223	struct ext4_free_extent result; /* result allocation */
	224	unsigned pid;
	225	unsigned ino;
	226	__u16 found; /* how many extents have been found */
	227	__u16 groups; /* how many groups have been scanned */
	228	__u16 tail; /* what tail broke some buddy */
	229	__u16 buddy; /* buddy the tail ^^^ broke */
	230	__u16 flags;
	231	__u8 cr:3; /* which phase the result extent was found at */
	232	__u8 op:4;
	233	__u8 merged:1;
	234	};
	235
	236	struct ext4_buddy {
	237	struct page *bd_buddy_page;
	238	void *bd_buddy;
	239	struct page *bd_bitmap_page;
	240	void *bd_bitmap;
	241	struct ext4_group_info *bd_info;
242	struct super_block *bd_sb;
243	__u16 bd_blkbits;
244	ext4_group_t bd_group;
245	};
246	#define EXT4_MB_BITMAP(e4b) ((e4b)->bd_bitmap)
247	#define EXT4_MB_BUDDY(e4b) ((e4b)->bd_buddy)
248
249	#ifndef EXT4_MB_HISTORY
250	static inline void ext4_mb_store_history(struct ext4_allocation_context *ac)
251	{
252	return;
253	}
254	#else
255	static void ext4_mb_store_history(struct ext4_allocation_context *ac);
256	#endif
257
258	#define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
259
8f6e39a7 MC	260	struct buffer_head read_block_bitmap(struct super_block , ext4_group_t);
	261
	262	static void ext4_mb_generate_from_pa(struct super_block sb, void bitmap,
	263	ext4_group_t group);
	264	static void ext4_mb_poll_new_transaction(struct super_block , handle_t );
	265	static void ext4_mb_free_committed_blocks(struct super_block *);
	266	static void ext4_mb_return_to_preallocation(struct inode *inode,
	267	struct ext4_buddy *e4b, sector_t block,
	268	int count);
	269	static void ext4_mb_put_pa(struct ext4_allocation_context *,
	270	struct super_block , struct ext4_prealloc_space pa);
	271	static int ext4_mb_init_per_dev_proc(struct super_block *sb);
	272	static int ext4_mb_destroy_per_dev_proc(struct super_block *sb);
	273
	274
	275	static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
	276	{
	277	struct ext4_group_info *grinfo = ext4_get_group_info(sb, group);
	278
	279	bit_spin_lock(EXT4_GROUP_INFO_LOCKED_BIT, &(grinfo->bb_state));
	280	}
	281
	282	static inline void ext4_unlock_group(struct super_block *sb,
	283	ext4_group_t group)
	284	{
	285	struct ext4_group_info *grinfo = ext4_get_group_info(sb, group);
	286
	287	bit_spin_unlock(EXT4_GROUP_INFO_LOCKED_BIT, &(grinfo->bb_state));
	288	}
	289
	290	static inline int ext4_is_group_locked(struct super_block *sb,
	291	ext4_group_t group)
	292	{
	293	struct ext4_group_info *grinfo = ext4_get_group_info(sb, group);
	294
	295	return bit_spin_is_locked(EXT4_GROUP_INFO_LOCKED_BIT,
	296	&(grinfo->bb_state));
	297	}
	298
	299	static ext4_fsblk_t ext4_grp_offs_to_block(struct super_block *sb,
	300	struct ext4_free_extent *fex)
	301	{
	302	ext4_fsblk_t block;
	303
	304	block = (ext4_fsblk_t) fex->fe_group * EXT4_BLOCKS_PER_GROUP(sb)
	305	+ fex->fe_start
	306	+ le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
	307	return block;
	308	}
	309	#endif