[mirror_zfs.git] / include / sys / dnode.h

/*
 * CDDL HEADER START
 *
 * The contents of this file are subject to the terms of the
 * Common Development and Distribution License (the "License").
 * You may not use this file except in compliance with the License.
 *
 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
 * or http://www.opensolaris.org/os/licensing.
 * See the License for the specific language governing permissions
 * and limitations under the License.
 *
 * When distributing Covered Code, include this CDDL HEADER in each
 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
 * If applicable, add the following below this CDDL HEADER, with the
 * fields enclosed by brackets "[]" replaced with your own identifying
 * information: Portions Copyright [yyyy] [name of copyright owner]
 *
 * CDDL HEADER END
 */
/*
 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
 * Copyright (c) 2012, 2016 by Delphix. All rights reserved.
 * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
 */

#ifndef	_SYS_DNODE_H
#define	_SYS_DNODE_H

#include <sys/zfs_context.h>
#include <sys/avl.h>
#include <sys/spa.h>
#include <sys/txg.h>
#include <sys/zio.h>
#include <sys/refcount.h>
#include <sys/dmu_zfetch.h>
#include <sys/zrlock.h>

#ifdef	__cplusplus
extern "C" {
#endif

/*
 * dnode_hold() flags.
 */
#define	DNODE_MUST_BE_ALLOCATED	1
#define	DNODE_MUST_BE_FREE	2

/*
 * dnode_next_offset() flags.
 */
#define	DNODE_FIND_HOLE		1
#define	DNODE_FIND_BACKWARDS	2
#define	DNODE_FIND_HAVELOCK	4

/*
 * Fixed constants.
 */
#define	DNODE_SHIFT		9	/* 512 bytes */
#define	DN_MIN_INDBLKSHIFT	12	/* 4k */
/*
 * If we ever increase this value beyond 20, we need to revisit all logic that
 * does x << level * ebps to handle overflow.  With a 1M indirect block size,
 * 4 levels of indirect blocks would not be able to guarantee addressing an
 * entire object, so 5 levels will be used, but 5 * (20 - 7) = 65.
 */
#define	DN_MAX_INDBLKSHIFT	14	/* 16k */
#define	DNODE_BLOCK_SHIFT	14	/* 16k */
#define	DNODE_CORE_SIZE		64	/* 64 bytes for dnode sans blkptrs */
#define	DN_MAX_OBJECT_SHIFT	48	/* 256 trillion (zfs_fid_t limit) */
#define	DN_MAX_OFFSET_SHIFT	64	/* 2^64 bytes in a dnode */

/*
 * dnode id flags
 *
 * Note: a file will never ever have its
 * ids moved from bonus->spill
 * and only in a crypto environment would it be on spill
 */
#define	DN_ID_CHKED_BONUS	0x1
#define	DN_ID_CHKED_SPILL	0x2
#define	DN_ID_OLD_EXIST		0x4
#define	DN_ID_NEW_EXIST		0x8

/*
 * Derived constants.
 */
#define	DNODE_MIN_SIZE		(1 << DNODE_SHIFT)
#define	DNODE_MAX_SIZE		(1 << DNODE_BLOCK_SHIFT)
#define	DNODE_BLOCK_SIZE	(1 << DNODE_BLOCK_SHIFT)
#define	DNODE_MIN_SLOTS		(DNODE_MIN_SIZE >> DNODE_SHIFT)
#define	DNODE_MAX_SLOTS		(DNODE_MAX_SIZE >> DNODE_SHIFT)
#define	DN_BONUS_SIZE(dnsize)	((dnsize) - DNODE_CORE_SIZE - \
	(1 << SPA_BLKPTRSHIFT))
#define	DN_SLOTS_TO_BONUSLEN(slots)	DN_BONUS_SIZE((slots) << DNODE_SHIFT)
#define	DN_OLD_MAX_BONUSLEN	(DN_BONUS_SIZE(DNODE_MIN_SIZE))
#define	DN_MAX_NBLKPTR	((DNODE_MIN_SIZE - DNODE_CORE_SIZE) >> SPA_BLKPTRSHIFT)
#define	DN_MAX_OBJECT	(1ULL << DN_MAX_OBJECT_SHIFT)
#define	DN_ZERO_BONUSLEN	(DN_BONUS_SIZE(DNODE_MAX_SIZE) + 1)
#define	DN_KILL_SPILLBLK (1)

#define	DNODES_PER_BLOCK_SHIFT	(DNODE_BLOCK_SHIFT - DNODE_SHIFT)
#define	DNODES_PER_BLOCK	(1ULL << DNODES_PER_BLOCK_SHIFT)
#define	DNODES_PER_LEVEL_SHIFT	(DN_MAX_INDBLKSHIFT - SPA_BLKPTRSHIFT)
#define	DNODES_PER_LEVEL	(1ULL << DNODES_PER_LEVEL_SHIFT)

#define	DN_MAX_LEVELS	(DIV_ROUND_UP(DN_MAX_OFFSET_SHIFT - SPA_MINBLOCKSHIFT, \
	DN_MIN_INDBLKSHIFT - SPA_BLKPTRSHIFT) + 1)

#define	DN_BONUS(dnp)	((void*)((dnp)->dn_bonus + \
	(((dnp)->dn_nblkptr - 1) * sizeof (blkptr_t))))

#define	DN_USED_BYTES(dnp) (((dnp)->dn_flags & DNODE_FLAG_USED_BYTES) ? \
	(dnp)->dn_used : (dnp)->dn_used << SPA_MINBLOCKSHIFT)

#define	EPB(blkshift, typeshift)	(1 << (blkshift - typeshift))

struct dmu_buf_impl;
struct objset;
struct zio;

enum dnode_dirtycontext {
	DN_UNDIRTIED,
	DN_DIRTY_OPEN,
	DN_DIRTY_SYNC
};

/* Is dn_used in bytes?  if not, it's in multiples of SPA_MINBLOCKSIZE */
#define	DNODE_FLAG_USED_BYTES		(1<<0)
#define	DNODE_FLAG_USERUSED_ACCOUNTED	(1<<1)

/* Does dnode have a SA spill blkptr in bonus? */
#define	DNODE_FLAG_SPILL_BLKPTR	(1<<2)

typedef struct dnode_phys {
	uint8_t dn_type;		/* dmu_object_type_t */
	uint8_t dn_indblkshift;		/* ln2(indirect block size) */
	uint8_t dn_nlevels;		/* 1=dn_blkptr->data blocks */
	uint8_t dn_nblkptr;		/* length of dn_blkptr */
	uint8_t dn_bonustype;		/* type of data in bonus buffer */
	uint8_t	dn_checksum;		/* ZIO_CHECKSUM type */
	uint8_t	dn_compress;		/* ZIO_COMPRESS type */
	uint8_t dn_flags;		/* DNODE_FLAG_* */
	uint16_t dn_datablkszsec;	/* data block size in 512b sectors */
	uint16_t dn_bonuslen;		/* length of dn_bonus */
	uint8_t dn_extra_slots;		/* # of subsequent slots consumed */
	uint8_t dn_pad2[3];

	/* accounting is protected by dn_dirty_mtx */
	uint64_t dn_maxblkid;		/* largest allocated block ID */
	uint64_t dn_used;		/* bytes (or sectors) of disk space */

	uint64_t dn_pad3[4];

	/*
	 * The tail region is 448 bytes for a 512 byte dnode, and
	 * correspondingly larger for larger dnode sizes. The spill
	 * block pointer, when present, is always at the end of the tail
	 * region. There are three ways this space may be used, using
	 * a 512 byte dnode for this diagram:
	 *
	 * 0       64      128     192     256     320     384     448 (offset)
	 * +---------------+---------------+---------------+-------+
	 * | dn_blkptr[0]  | dn_blkptr[1]  | dn_blkptr[2]  | /     |
	 * +---------------+---------------+---------------+-------+
	 * | dn_blkptr[0]  | dn_bonus[0..319]                      |
	 * +---------------+-----------------------+---------------+
	 * | dn_blkptr[0]  | dn_bonus[0..191]      | dn_spill      |
	 * +---------------+-----------------------+---------------+
	 */
	union {
		blkptr_t dn_blkptr[1+DN_OLD_MAX_BONUSLEN/sizeof (blkptr_t)];
		struct {
			blkptr_t __dn_ignore1;
			uint8_t dn_bonus[DN_OLD_MAX_BONUSLEN];
		};
		struct {
			blkptr_t __dn_ignore2;
			uint8_t __dn_ignore3[DN_OLD_MAX_BONUSLEN -
			    sizeof (blkptr_t)];
			blkptr_t dn_spill;
		};
	};
} dnode_phys_t;

#define	DN_SPILL_BLKPTR(dnp)	(blkptr_t *)((char *)(dnp) + \
	(((dnp)->dn_extra_slots + 1) << DNODE_SHIFT) - (1 << SPA_BLKPTRSHIFT))

struct dnode {
	/*
	 * Protects the structure of the dnode, including the number of levels
	 * of indirection (dn_nlevels), dn_maxblkid, and dn_next_*
	 */
	krwlock_t dn_struct_rwlock;

	/* Our link on dn_objset->os_dnodes list; protected by os_lock.  */
	list_node_t dn_link;

	/* immutable: */
	struct objset *dn_objset;
	uint64_t dn_object;
	struct dmu_buf_impl *dn_dbuf;
	struct dnode_handle *dn_handle;
	dnode_phys_t *dn_phys; /* pointer into dn->dn_dbuf->db.db_data */

	/*
	 * Copies of stuff in dn_phys.  They're valid in the open
	 * context (eg. even before the dnode is first synced).
	 * Where necessary, these are protected by dn_struct_rwlock.
	 */
	dmu_object_type_t dn_type;	/* object type */
	uint16_t dn_bonuslen;		/* bonus length */
	uint8_t dn_bonustype;		/* bonus type */
	uint8_t dn_nblkptr;		/* number of blkptrs (immutable) */
	uint8_t dn_checksum;		/* ZIO_CHECKSUM type */
	uint8_t dn_compress;		/* ZIO_COMPRESS type */
	uint8_t dn_nlevels;
	uint8_t dn_indblkshift;
	uint8_t dn_datablkshift;	/* zero if blksz not power of 2! */
	uint8_t dn_moved;		/* Has this dnode been moved? */
	uint16_t dn_datablkszsec;	/* in 512b sectors */
	uint32_t dn_datablksz;		/* in bytes */
	uint64_t dn_maxblkid;
	uint8_t dn_next_type[TXG_SIZE];
	uint8_t dn_num_slots;		/* metadnode slots consumed on disk */
	uint8_t dn_next_nblkptr[TXG_SIZE];
	uint8_t dn_next_nlevels[TXG_SIZE];
	uint8_t dn_next_indblkshift[TXG_SIZE];
	uint8_t dn_next_bonustype[TXG_SIZE];
	uint8_t dn_rm_spillblk[TXG_SIZE];	/* for removing spill blk */
	uint16_t dn_next_bonuslen[TXG_SIZE];
	uint32_t dn_next_blksz[TXG_SIZE];	/* next block size in bytes */

	/* protected by dn_dbufs_mtx; declared here to fill 32-bit hole */
	uint32_t dn_dbufs_count;	/* count of dn_dbufs */
	/* There are no level-0 blocks of this blkid or higher in dn_dbufs */
	uint64_t dn_unlisted_l0_blkid;

	/* protected by os_lock: */
	list_node_t dn_dirty_link[TXG_SIZE];	/* next on dataset's dirty */

	/* protected by dn_mtx: */
	kmutex_t dn_mtx;
	list_t dn_dirty_records[TXG_SIZE];
	struct range_tree *dn_free_ranges[TXG_SIZE];
	uint64_t dn_allocated_txg;
	uint64_t dn_free_txg;
	uint64_t dn_assigned_txg;
	kcondvar_t dn_notxholds;
	enum dnode_dirtycontext dn_dirtyctx;
	uint8_t *dn_dirtyctx_firstset;		/* dbg: contents meaningless */

	/* protected by own devices */
	refcount_t dn_tx_holds;
	refcount_t dn_holds;

	kmutex_t dn_dbufs_mtx;
	/*
	 * Descendent dbufs, ordered by dbuf_compare. Note that dn_dbufs
	 * can contain multiple dbufs of the same (level, blkid) when a
	 * dbuf is marked DB_EVICTING without being removed from
	 * dn_dbufs. To maintain the avl invariant that there cannot be
	 * duplicate entries, we order the dbufs by an arbitrary value -
	 * their address in memory. This means that dn_dbufs cannot be used to
	 * directly look up a dbuf. Instead, callers must use avl_walk, have
	 * a reference to the dbuf, or look up a non-existant node with
	 * db_state = DB_SEARCH (see dbuf_free_range for an example).
	 */
	avl_tree_t dn_dbufs;

	/* protected by dn_struct_rwlock */
	struct dmu_buf_impl *dn_bonus;	/* bonus buffer dbuf */

	boolean_t dn_have_spill;	/* have spill or are spilling */

	/* parent IO for current sync write */
	zio_t *dn_zio;

	/* used in syncing context */
	uint64_t dn_oldused;	/* old phys used bytes */
	uint64_t dn_oldflags;	/* old phys dn_flags */
	uint64_t dn_olduid, dn_oldgid;
	uint64_t dn_newuid, dn_newgid;
	int dn_id_flags;

	/* holds prefetch structure */
	struct zfetch	dn_zfetch;
};

/*
 * Adds a level of indirection between the dbuf and the dnode to avoid
 * iterating descendent dbufs in dnode_move(). Handles are not allocated
 * individually, but as an array of child dnodes in dnode_hold_impl().
 */
typedef struct dnode_handle {
	/* Protects dnh_dnode from modification by dnode_move(). */
	zrlock_t dnh_zrlock;
	dnode_t *dnh_dnode;
} dnode_handle_t;

typedef struct dnode_children {
	dmu_buf_user_t dnc_dbu;		/* User evict data */
	size_t dnc_count;		/* number of children */
	dnode_handle_t dnc_children[];	/* sized dynamically */
} dnode_children_t;

typedef struct free_range {
	avl_node_t fr_node;
	uint64_t fr_blkid;
	uint64_t fr_nblks;
} free_range_t;

void dnode_special_open(struct objset *dd, dnode_phys_t *dnp,
    uint64_t object, dnode_handle_t *dnh);
void dnode_special_close(dnode_handle_t *dnh);

void dnode_setbonuslen(dnode_t *dn, int newsize, dmu_tx_t *tx);
void dnode_setbonus_type(dnode_t *dn, dmu_object_type_t, dmu_tx_t *tx);
void dnode_rm_spill(dnode_t *dn, dmu_tx_t *tx);

int dnode_hold(struct objset *dd, uint64_t object,
    void *ref, dnode_t **dnp);
int dnode_hold_impl(struct objset *dd, uint64_t object, int flag, int dn_slots,
    void *ref, dnode_t **dnp);
boolean_t dnode_add_ref(dnode_t *dn, void *ref);
void dnode_rele(dnode_t *dn, void *ref);
void dnode_rele_and_unlock(dnode_t *dn, void *tag);
void dnode_setdirty(dnode_t *dn, dmu_tx_t *tx);
void dnode_sync(dnode_t *dn, dmu_tx_t *tx);
void dnode_allocate(dnode_t *dn, dmu_object_type_t ot, int blocksize, int ibs,
    dmu_object_type_t bonustype, int bonuslen, int dn_slots, dmu_tx_t *tx);
void dnode_reallocate(dnode_t *dn, dmu_object_type_t ot, int blocksize,
    dmu_object_type_t bonustype, int bonuslen, int dn_slots, dmu_tx_t *tx);
void dnode_free(dnode_t *dn, dmu_tx_t *tx);
void dnode_byteswap(dnode_phys_t *dnp);
void dnode_buf_byteswap(void *buf, size_t size);
void dnode_verify(dnode_t *dn);
int dnode_set_blksz(dnode_t *dn, uint64_t size, int ibs, dmu_tx_t *tx);
void dnode_free_range(dnode_t *dn, uint64_t off, uint64_t len, dmu_tx_t *tx);
void dnode_diduse_space(dnode_t *dn, int64_t space);
void dnode_willuse_space(dnode_t *dn, int64_t space, dmu_tx_t *tx);
void dnode_new_blkid(dnode_t *dn, uint64_t blkid, dmu_tx_t *tx, boolean_t);
uint64_t dnode_block_freed(dnode_t *dn, uint64_t blkid);
void dnode_init(void);
void dnode_fini(void);
int dnode_next_offset(dnode_t *dn, int flags, uint64_t *off,
    int minlvl, uint64_t blkfill, uint64_t txg);
void dnode_evict_dbufs(dnode_t *dn);
void dnode_evict_bonus(dnode_t *dn);

#define	DNODE_IS_CACHEABLE(_dn)						\
	((_dn)->dn_objset->os_primary_cache == ZFS_CACHE_ALL ||		\
	(DMU_OT_IS_METADATA((_dn)->dn_type) &&				\
	(_dn)->dn_objset->os_primary_cache == ZFS_CACHE_METADATA))

#define	DNODE_META_IS_CACHEABLE(_dn)					\
	((_dn)->dn_objset->os_primary_cache == ZFS_CACHE_ALL ||		\
	(_dn)->dn_objset->os_primary_cache == ZFS_CACHE_METADATA)

#ifdef ZFS_DEBUG

/*
 * There should be a ## between the string literal and fmt, to make it
 * clear that we're joining two strings together, but that piece of shit
 * gcc doesn't support that preprocessor token.
 */
#define	dprintf_dnode(dn, fmt, ...) do { \
	if (zfs_flags & ZFS_DEBUG_DPRINTF) { \
	char __db_buf[32]; \
	uint64_t __db_obj = (dn)->dn_object; \
	if (__db_obj == DMU_META_DNODE_OBJECT) \
		(void) strcpy(__db_buf, "mdn"); \
	else \
		(void) snprintf(__db_buf, sizeof (__db_buf), "%lld", \
		    (u_longlong_t)__db_obj);\
	dprintf_ds((dn)->dn_objset->os_dsl_dataset, "obj=%s " fmt, \
	    __db_buf, __VA_ARGS__); \
	} \
_NOTE(CONSTCOND) } while (0)

#define	DNODE_VERIFY(dn)		dnode_verify(dn)
#define	FREE_VERIFY(db, start, end, tx)	free_verify(db, start, end, tx)

#else

#define	dprintf_dnode(db, fmt, ...)
#define	DNODE_VERIFY(dn)
#define	FREE_VERIFY(db, start, end, tx)

#endif

#ifdef	__cplusplus
}
#endif

#endif	/* _SYS_DNODE_H */
Commit	Line	Data
34dc7c2f BB	1	/*
	2	* CDDL HEADER START
	3	*
	4	* The contents of this file are subject to the terms of the
	5	* Common Development and Distribution License (the "License").
	6	* You may not use this file except in compliance with the License.
	7	*
	8	* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
	9	* or http://www.opensolaris.org/os/licensing.
	10	* See the License for the specific language governing permissions
	11	* and limitations under the License.
	12	*
	13	* When distributing Covered Code, include this CDDL HEADER in each
	14	* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
	15	* If applicable, add the following below this CDDL HEADER, with the
	16	* fields enclosed by brackets "[]" replaced with your own identifying
	17	* information: Portions Copyright [yyyy] [name of copyright owner]
	18	*
	19	* CDDL HEADER END
	20	*/
	21	/*
428870ff	22	* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
2bce8049	23	* Copyright (c) 2012, 2016 by Delphix. All rights reserved.
0c66c32d	24	* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
34dc7c2f BB	25	*/
	26
	27	#ifndef _SYS_DNODE_H
	28	#define _SYS_DNODE_H
	29
34dc7c2f BB	30	#include <sys/zfs_context.h>
	31	#include <sys/avl.h>
	32	#include <sys/spa.h>
	33	#include <sys/txg.h>
	34	#include <sys/zio.h>
	35	#include <sys/refcount.h>
	36	#include <sys/dmu_zfetch.h>
572e2857	37	#include <sys/zrlock.h>
34dc7c2f BB	38
	39	#ifdef __cplusplus
	40	extern "C" {
	41	#endif
	42
	43	/*
b128c09f	44	* dnode_hold() flags.
34dc7c2f BB	45	*/
	46	#define DNODE_MUST_BE_ALLOCATED 1
	47	#define DNODE_MUST_BE_FREE 2
	48
b128c09f BB	49	/*
	50	* dnode_next_offset() flags.
	51	*/
	52	#define DNODE_FIND_HOLE 1
	53	#define DNODE_FIND_BACKWARDS 2
	54	#define DNODE_FIND_HAVELOCK 4
	55
34dc7c2f BB	56	/*
	57	* Fixed constants.
	58	*/
	59	#define DNODE_SHIFT 9 /* 512 bytes */
d25b4497	60	#define DN_MIN_INDBLKSHIFT 12 /* 4k */
32d41fb7 PD	61	/*
	62	* If we ever increase this value beyond 20, we need to revisit all logic that
	63	* does x << level * ebps to handle overflow. With a 1M indirect block size,
	64	* 4 levels of indirect blocks would not be able to guarantee addressing an
	65	* entire object, so 5 levels will be used, but 5 * (20 - 7) = 65.
	66	*/
34dc7c2f BB	67	#define DN_MAX_INDBLKSHIFT 14 /* 16k */
	68	#define DNODE_BLOCK_SHIFT 14 /* 16k */
	69	#define DNODE_CORE_SIZE 64 /* 64 bytes for dnode sans blkptrs */
	70	#define DN_MAX_OBJECT_SHIFT 48 /* 256 trillion (zfs_fid_t limit) */
	71	#define DN_MAX_OFFSET_SHIFT 64 /* 2^64 bytes in a dnode */
	72
428870ff BB	73	/*
	74	* dnode id flags
	75	*
	76	* Note: a file will never ever have its
	77	* ids moved from bonus->spill
	78	* and only in a crypto environment would it be on spill
	79	*/
	80	#define DN_ID_CHKED_BONUS 0x1
	81	#define DN_ID_CHKED_SPILL 0x2
	82	#define DN_ID_OLD_EXIST 0x4
	83	#define DN_ID_NEW_EXIST 0x8
	84
34dc7c2f BB	85	/*
	86	* Derived constants.
	87	*/
50c957f7 NB	88	#define DNODE_MIN_SIZE (1 << DNODE_SHIFT)
	89	#define DNODE_MAX_SIZE (1 << DNODE_BLOCK_SHIFT)
	90	#define DNODE_BLOCK_SIZE (1 << DNODE_BLOCK_SHIFT)
	91	#define DNODE_MIN_SLOTS (DNODE_MIN_SIZE >> DNODE_SHIFT)
	92	#define DNODE_MAX_SLOTS (DNODE_MAX_SIZE >> DNODE_SHIFT)
	93	#define DN_BONUS_SIZE(dnsize) ((dnsize) - DNODE_CORE_SIZE - \
	94	(1 << SPA_BLKPTRSHIFT))
	95	#define DN_SLOTS_TO_BONUSLEN(slots) DN_BONUS_SIZE((slots) << DNODE_SHIFT)
	96	#define DN_OLD_MAX_BONUSLEN (DN_BONUS_SIZE(DNODE_MIN_SIZE))
	97	#define DN_MAX_NBLKPTR ((DNODE_MIN_SIZE - DNODE_CORE_SIZE) >> SPA_BLKPTRSHIFT)
34dc7c2f	98	#define DN_MAX_OBJECT (1ULL << DN_MAX_OBJECT_SHIFT)
50c957f7	99	#define DN_ZERO_BONUSLEN (DN_BONUS_SIZE(DNODE_MAX_SIZE) + 1)
428870ff	100	#define DN_KILL_SPILLBLK (1)
34dc7c2f BB	101
	102	#define DNODES_PER_BLOCK_SHIFT (DNODE_BLOCK_SHIFT - DNODE_SHIFT)
	103	#define DNODES_PER_BLOCK (1ULL << DNODES_PER_BLOCK_SHIFT)
	104	#define DNODES_PER_LEVEL_SHIFT (DN_MAX_INDBLKSHIFT - SPA_BLKPTRSHIFT)
428870ff	105	#define DNODES_PER_LEVEL (1ULL << DNODES_PER_LEVEL_SHIFT)
34dc7c2f	106
031d7c2f GN	107	#define DN_MAX_LEVELS (DIV_ROUND_UP(DN_MAX_OFFSET_SHIFT - SPA_MINBLOCKSHIFT, \
031d7c2f GN	108	DN_MIN_INDBLKSHIFT - SPA_BLKPTRSHIFT) + 1)
34dc7c2f BB	109
	110	#define DN_BONUS(dnp) ((void*)((dnp)->dn_bonus + \
	111	(((dnp)->dn_nblkptr - 1) * sizeof (blkptr_t))))
	112
	113	#define DN_USED_BYTES(dnp) (((dnp)->dn_flags & DNODE_FLAG_USED_BYTES) ? \
	114	(dnp)->dn_used : (dnp)->dn_used << SPA_MINBLOCKSHIFT)
	115
	116	#define EPB(blkshift, typeshift) (1 << (blkshift - typeshift))
	117
	118	struct dmu_buf_impl;
428870ff	119	struct objset;
34dc7c2f BB	120	struct zio;
	121
	122	enum dnode_dirtycontext {
	123	DN_UNDIRTIED,
	124	DN_DIRTY_OPEN,
	125	DN_DIRTY_SYNC
	126	};
	127
	128	/* Is dn_used in bytes? if not, it's in multiples of SPA_MINBLOCKSIZE */
9babb374 BB	129	#define DNODE_FLAG_USED_BYTES (1<<0)
9babb374 BB	130	#define DNODE_FLAG_USERUSED_ACCOUNTED (1<<1)
34dc7c2f	131
428870ff BB	132	/* Does dnode have a SA spill blkptr in bonus? */
	133	#define DNODE_FLAG_SPILL_BLKPTR (1<<2)
	134
34dc7c2f BB	135	typedef struct dnode_phys {
	136	uint8_t dn_type; /* dmu_object_type_t */
	137	uint8_t dn_indblkshift; /* ln2(indirect block size) */
	138	uint8_t dn_nlevels; /* 1=dn_blkptr->data blocks */
	139	uint8_t dn_nblkptr; /* length of dn_blkptr */
	140	uint8_t dn_bonustype; /* type of data in bonus buffer */
	141	uint8_t dn_checksum; /* ZIO_CHECKSUM type */
	142	uint8_t dn_compress; /* ZIO_COMPRESS type */
	143	uint8_t dn_flags; /* DNODE_FLAG_* */
	144	uint16_t dn_datablkszsec; /* data block size in 512b sectors */
	145	uint16_t dn_bonuslen; /* length of dn_bonus */
50c957f7 NB	146	uint8_t dn_extra_slots; /* # of subsequent slots consumed */
50c957f7 NB	147	uint8_t dn_pad2[3];
34dc7c2f BB	148
	149	/* accounting is protected by dn_dirty_mtx */
	150	uint64_t dn_maxblkid; /* largest allocated block ID */
	151	uint64_t dn_used; /* bytes (or sectors) of disk space */
	152
	153	uint64_t dn_pad3[4];
	154
aca19e06	155	/*
50c957f7 NB	156	* The tail region is 448 bytes for a 512 byte dnode, and
	157	* correspondingly larger for larger dnode sizes. The spill
	158	* block pointer, when present, is always at the end of the tail
	159	* region. There are three ways this space may be used, using
	160	* a 512 byte dnode for this diagram:
aca19e06 JE	161	*
	162	* 0 64 128 192 256 320 384 448 (offset)
	163	* +---------------+---------------+---------------+-------+
	164	* \| dn_blkptr[0] \| dn_blkptr[1] \| dn_blkptr[2] \| / \|
	165	* +---------------+---------------+---------------+-------+
	166	* \| dn_blkptr[0] \| dn_bonus[0..319] \|
	167	* +---------------+-----------------------+---------------+
50c957f7	168	* \| dn_blkptr[0] \| dn_bonus[0..191] \| dn_spill \|
aca19e06 JE	169	* +---------------+-----------------------+---------------+
	170	*/
	171	union {
50c957f7	172	blkptr_t dn_blkptr[1+DN_OLD_MAX_BONUSLEN/sizeof (blkptr_t)];
aca19e06 JE	173	struct {
aca19e06 JE	174	blkptr_t __dn_ignore1;
50c957f7	175	uint8_t dn_bonus[DN_OLD_MAX_BONUSLEN];
aca19e06 JE	176	};
	177	struct {
	178	blkptr_t __dn_ignore2;
50c957f7 NB	179	uint8_t __dn_ignore3[DN_OLD_MAX_BONUSLEN -
50c957f7 NB	180	sizeof (blkptr_t)];
aca19e06 JE	181	blkptr_t dn_spill;
	182	};
	183	};
34dc7c2f BB	184	} dnode_phys_t;
34dc7c2f BB	185
50c957f7 NB	186	#define DN_SPILL_BLKPTR(dnp) (blkptr_t )((char )(dnp) + \
	187	(((dnp)->dn_extra_slots + 1) << DNODE_SHIFT) - (1 << SPA_BLKPTRSHIFT))
	188
2bce8049	189	struct dnode {
34dc7c2f	190	/*
d3cc8b15 WA	191	* Protects the structure of the dnode, including the number of levels
d3cc8b15 WA	192	* of indirection (dn_nlevels), dn_maxblkid, and dn_next_*
34dc7c2f BB	193	*/
	194	krwlock_t dn_struct_rwlock;
	195
9babb374	196	/* Our link on dn_objset->os_dnodes list; protected by os_lock. */
34dc7c2f BB	197	list_node_t dn_link;
	198
	199	/* immutable: */
428870ff	200	struct objset *dn_objset;
34dc7c2f BB	201	uint64_t dn_object;
34dc7c2f BB	202	struct dmu_buf_impl *dn_dbuf;
572e2857	203	struct dnode_handle *dn_handle;
34dc7c2f BB	204	dnode_phys_t dn_phys; / pointer into dn->dn_dbuf->db.db_data */
	205
	206	/*
	207	* Copies of stuff in dn_phys. They're valid in the open
	208	* context (eg. even before the dnode is first synced).
	209	* Where necessary, these are protected by dn_struct_rwlock.
	210	*/
	211	dmu_object_type_t dn_type; /* object type */
	212	uint16_t dn_bonuslen; /* bonus length */
	213	uint8_t dn_bonustype; /* bonus type */
	214	uint8_t dn_nblkptr; /* number of blkptrs (immutable) */
	215	uint8_t dn_checksum; /* ZIO_CHECKSUM type */
	216	uint8_t dn_compress; /* ZIO_COMPRESS type */
	217	uint8_t dn_nlevels;
	218	uint8_t dn_indblkshift;
	219	uint8_t dn_datablkshift; /* zero if blksz not power of 2! */
572e2857	220	uint8_t dn_moved; /* Has this dnode been moved? */
34dc7c2f BB	221	uint16_t dn_datablkszsec; /* in 512b sectors */
	222	uint32_t dn_datablksz; /* in bytes */
	223	uint64_t dn_maxblkid;
fa86b5db	224	uint8_t dn_next_type[TXG_SIZE];
50c957f7	225	uint8_t dn_num_slots; /* metadnode slots consumed on disk */
d164b209	226	uint8_t dn_next_nblkptr[TXG_SIZE];
34dc7c2f BB	227	uint8_t dn_next_nlevels[TXG_SIZE];
34dc7c2f BB	228	uint8_t dn_next_indblkshift[TXG_SIZE];
428870ff BB	229	uint8_t dn_next_bonustype[TXG_SIZE];
428870ff BB	230	uint8_t dn_rm_spillblk[TXG_SIZE]; /* for removing spill blk */
34dc7c2f BB	231	uint16_t dn_next_bonuslen[TXG_SIZE];
	232	uint32_t dn_next_blksz[TXG_SIZE]; /* next block size in bytes */
	233
572e2857 BB	234	/* protected by dn_dbufs_mtx; declared here to fill 32-bit hole */
572e2857 BB	235	uint32_t dn_dbufs_count; /* count of dn_dbufs */
b663a23d MA	236	/* There are no level-0 blocks of this blkid or higher in dn_dbufs */
b663a23d MA	237	uint64_t dn_unlisted_l0_blkid;
572e2857	238
34dc7c2f BB	239	/* protected by os_lock: */
	240	list_node_t dn_dirty_link[TXG_SIZE]; /* next on dataset's dirty */
	241
	242	/* protected by dn_mtx: */
	243	kmutex_t dn_mtx;
	244	list_t dn_dirty_records[TXG_SIZE];
9bd274dd	245	struct range_tree *dn_free_ranges[TXG_SIZE];
34dc7c2f BB	246	uint64_t dn_allocated_txg;
	247	uint64_t dn_free_txg;
	248	uint64_t dn_assigned_txg;
	249	kcondvar_t dn_notxholds;
	250	enum dnode_dirtycontext dn_dirtyctx;
	251	uint8_t dn_dirtyctx_firstset; / dbg: contents meaningless */
	252
	253	/* protected by own devices */
	254	refcount_t dn_tx_holds;
	255	refcount_t dn_holds;
	256
	257	kmutex_t dn_dbufs_mtx;
9925c28c AR	258	/*
	259	* Descendent dbufs, ordered by dbuf_compare. Note that dn_dbufs
	260	* can contain multiple dbufs of the same (level, blkid) when a
	261	* dbuf is marked DB_EVICTING without being removed from
	262	* dn_dbufs. To maintain the avl invariant that there cannot be
	263	* duplicate entries, we order the dbufs by an arbitrary value -
	264	* their address in memory. This means that dn_dbufs cannot be used to
	265	* directly look up a dbuf. Instead, callers must use avl_walk, have
	266	* a reference to the dbuf, or look up a non-existant node with
	267	* db_state = DB_SEARCH (see dbuf_free_range for an example).
	268	*/
	269	avl_tree_t dn_dbufs;
572e2857 BB	270
572e2857 BB	271	/* protected by dn_struct_rwlock */
34dc7c2f	272	struct dmu_buf_impl dn_bonus; / bonus buffer dbuf */
572e2857	273
428870ff	274	boolean_t dn_have_spill; /* have spill or are spilling */
34dc7c2f BB	275
	276	/* parent IO for current sync write */
	277	zio_t *dn_zio;
	278
9babb374	279	/* used in syncing context */
428870ff BB	280	uint64_t dn_oldused; /* old phys used bytes */
	281	uint64_t dn_oldflags; /* old phys dn_flags */
	282	uint64_t dn_olduid, dn_oldgid;
	283	uint64_t dn_newuid, dn_newgid;
	284	int dn_id_flags;
9babb374	285
34dc7c2f BB	286	/* holds prefetch structure */
34dc7c2f BB	287	struct zfetch dn_zfetch;
2bce8049	288	};
34dc7c2f	289
572e2857 BB	290	/*
	291	* Adds a level of indirection between the dbuf and the dnode to avoid
	292	* iterating descendent dbufs in dnode_move(). Handles are not allocated
	293	* individually, but as an array of child dnodes in dnode_hold_impl().
	294	*/
	295	typedef struct dnode_handle {
	296	/* Protects dnh_dnode from modification by dnode_move(). */
	297	zrlock_t dnh_zrlock;
	298	dnode_t *dnh_dnode;
	299	} dnode_handle_t;
	300
	301	typedef struct dnode_children {
0c66c32d	302	dmu_buf_user_t dnc_dbu; /* User evict data */
572e2857	303	size_t dnc_count; /* number of children */
5aea3644	304	dnode_handle_t dnc_children[]; /* sized dynamically */
572e2857 BB	305	} dnode_children_t;
572e2857 BB	306
34dc7c2f BB	307	typedef struct free_range {
	308	avl_node_t fr_node;
	309	uint64_t fr_blkid;
	310	uint64_t fr_nblks;
	311	} free_range_t;
	312
0c66c32d	313	void dnode_special_open(struct objset dd, dnode_phys_t dnp,
572e2857 BB	314	uint64_t object, dnode_handle_t *dnh);
572e2857 BB	315	void dnode_special_close(dnode_handle_t *dnh);
34dc7c2f BB	316
34dc7c2f BB	317	void dnode_setbonuslen(dnode_t dn, int newsize, dmu_tx_t tx);
428870ff BB	318	void dnode_setbonus_type(dnode_t dn, dmu_object_type_t, dmu_tx_t tx);
	319	void dnode_rm_spill(dnode_t dn, dmu_tx_t tx);
	320
	321	int dnode_hold(struct objset *dd, uint64_t object,
34dc7c2f	322	void ref, dnode_t *dnp);
50c957f7	323	int dnode_hold_impl(struct objset *dd, uint64_t object, int flag, int dn_slots,
34dc7c2f BB	324	void ref, dnode_t *dnp);
	325	boolean_t dnode_add_ref(dnode_t dn, void ref);
	326	void dnode_rele(dnode_t dn, void ref);
4c7b7eed	327	void dnode_rele_and_unlock(dnode_t dn, void tag);
34dc7c2f BB	328	void dnode_setdirty(dnode_t dn, dmu_tx_t tx);
	329	void dnode_sync(dnode_t dn, dmu_tx_t tx);
	330	void dnode_allocate(dnode_t *dn, dmu_object_type_t ot, int blocksize, int ibs,
50c957f7	331	dmu_object_type_t bonustype, int bonuslen, int dn_slots, dmu_tx_t *tx);
34dc7c2f	332	void dnode_reallocate(dnode_t *dn, dmu_object_type_t ot, int blocksize,
50c957f7	333	dmu_object_type_t bonustype, int bonuslen, int dn_slots, dmu_tx_t *tx);
34dc7c2f BB	334	void dnode_free(dnode_t dn, dmu_tx_t tx);
	335	void dnode_byteswap(dnode_phys_t *dnp);
	336	void dnode_buf_byteswap(void *buf, size_t size);
	337	void dnode_verify(dnode_t *dn);
	338	int dnode_set_blksz(dnode_t dn, uint64_t size, int ibs, dmu_tx_t tx);
34dc7c2f	339	void dnode_free_range(dnode_t dn, uint64_t off, uint64_t len, dmu_tx_t tx);
34dc7c2f BB	340	void dnode_diduse_space(dnode_t *dn, int64_t space);
34dc7c2f BB	341	void dnode_willuse_space(dnode_t dn, int64_t space, dmu_tx_t tx);
b128c09f	342	void dnode_new_blkid(dnode_t dn, uint64_t blkid, dmu_tx_t tx, boolean_t);
34dc7c2f BB	343	uint64_t dnode_block_freed(dnode_t *dn, uint64_t blkid);
	344	void dnode_init(void);
	345	void dnode_fini(void);
b128c09f BB	346	int dnode_next_offset(dnode_t dn, int flags, uint64_t off,
b128c09f BB	347	int minlvl, uint64_t blkfill, uint64_t txg);
34dc7c2f	348	void dnode_evict_dbufs(dnode_t *dn);
4c7b7eed	349	void dnode_evict_bonus(dnode_t *dn);
34dc7c2f	350
755065f3 AM	351	#define DNODE_IS_CACHEABLE(_dn) \
	352	((_dn)->dn_objset->os_primary_cache == ZFS_CACHE_ALL \|\| \
	353	(DMU_OT_IS_METADATA((_dn)->dn_type) && \
	354	(_dn)->dn_objset->os_primary_cache == ZFS_CACHE_METADATA))
	355
	356	#define DNODE_META_IS_CACHEABLE(_dn) \
	357	((_dn)->dn_objset->os_primary_cache == ZFS_CACHE_ALL \|\| \
	358	(_dn)->dn_objset->os_primary_cache == ZFS_CACHE_METADATA)
	359
34dc7c2f BB	360	#ifdef ZFS_DEBUG
	361
	362	/*
	363	* There should be a ## between the string literal and fmt, to make it
	364	* clear that we're joining two strings together, but that piece of shit
	365	* gcc doesn't support that preprocessor token.
	366	*/
	367	#define dprintf_dnode(dn, fmt, ...) do { \
	368	if (zfs_flags & ZFS_DEBUG_DPRINTF) { \
	369	char __db_buf[32]; \
	370	uint64_t __db_obj = (dn)->dn_object; \
	371	if (__db_obj == DMU_META_DNODE_OBJECT) \
	372	(void) strcpy(__db_buf, "mdn"); \
	373	else \
	374	(void) snprintf(__db_buf, sizeof (__db_buf), "%lld", \
	375	(u_longlong_t)__db_obj);\
	376	dprintf_ds((dn)->dn_objset->os_dsl_dataset, "obj=%s " fmt, \
	377	__db_buf, __VA_ARGS__); \
	378	} \
	379	_NOTE(CONSTCOND) } while (0)
	380
	381	#define DNODE_VERIFY(dn) dnode_verify(dn)
	382	#define FREE_VERIFY(db, start, end, tx) free_verify(db, start, end, tx)
	383
	384	#else
	385
	386	#define dprintf_dnode(db, fmt, ...)
	387	#define DNODE_VERIFY(dn)
	388	#define FREE_VERIFY(db, start, end, tx)
	389
	390	#endif
	391
	392	#ifdef __cplusplus
	393	}
	394	#endif
	395
	396	#endif /* _SYS_DNODE_H */