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 /* 22 * Copyright 2010 Sun Microsystems, Inc. All rights reserved. 23 * Use is subject to license terms. 24 */ 25 26 #ifndef _SYS_DNODE_H 27 #define _SYS_DNODE_H 28 29 #include <sys/zfs_context.h> 30 #include <sys/avl.h> 31 #include <sys/spa.h> 32 #include <sys/txg.h> 33 #include <sys/zio.h> 34 #include <sys/refcount.h> 35 #include <sys/dmu_zfetch.h> 36 37 #ifdef __cplusplus 38 extern "C" { 39 #endif 40 41 /* 42 * dnode_hold() flags. 43 */ 44 #define DNODE_MUST_BE_ALLOCATED 1 45 #define DNODE_MUST_BE_FREE 2 46 47 /* 48 * dnode_next_offset() flags. 49 */ 50 #define DNODE_FIND_HOLE 1 51 #define DNODE_FIND_BACKWARDS 2 52 #define DNODE_FIND_HAVELOCK 4 53 54 /* 55 * Fixed constants. 56 */ 57 #define DNODE_SHIFT 9 /* 512 bytes */ 58 #define DN_MIN_INDBLKSHIFT 10 /* 1k */ 59 #define DN_MAX_INDBLKSHIFT 14 /* 16k */ 60 #define DNODE_BLOCK_SHIFT 14 /* 16k */ 61 #define DNODE_CORE_SIZE 64 /* 64 bytes for dnode sans blkptrs */ 62 #define DN_MAX_OBJECT_SHIFT 48 /* 256 trillion (zfs_fid_t limit) */ 63 #define DN_MAX_OFFSET_SHIFT 64 /* 2^64 bytes in a dnode */ 64 65 /* 66 * Derived constants. 67 */ 68 #define DNODE_SIZE (1 << DNODE_SHIFT) 69 #define DN_MAX_NBLKPTR ((DNODE_SIZE - DNODE_CORE_SIZE) >> SPA_BLKPTRSHIFT) 70 #define DN_MAX_BONUSLEN (DNODE_SIZE - DNODE_CORE_SIZE - (1 << SPA_BLKPTRSHIFT)) 71 #define DN_MAX_OBJECT (1ULL << DN_MAX_OBJECT_SHIFT) 72 #define DN_ZERO_BONUSLEN (DN_MAX_BONUSLEN + 1) 73 74 #define DNODES_PER_BLOCK_SHIFT (DNODE_BLOCK_SHIFT - DNODE_SHIFT) 75 #define DNODES_PER_BLOCK (1ULL << DNODES_PER_BLOCK_SHIFT) 76 #define DNODES_PER_LEVEL_SHIFT (DN_MAX_INDBLKSHIFT - SPA_BLKPTRSHIFT) 77 #define DNODES_PER_LEVEL (1ULL << DNODES_PER_LEVEL_SHIFT) 78 79 /* The +2 here is a cheesy way to round up */ 80 #define DN_MAX_LEVELS (2 + ((DN_MAX_OFFSET_SHIFT - SPA_MINBLOCKSHIFT) / \ 81 (DN_MIN_INDBLKSHIFT - SPA_BLKPTRSHIFT))) 82 83 #define DN_BONUS(dnp) ((void*)((dnp)->dn_bonus + \ 84 (((dnp)->dn_nblkptr - 1) * sizeof (blkptr_t)))) 85 86 #define DN_USED_BYTES(dnp) (((dnp)->dn_flags & DNODE_FLAG_USED_BYTES) ? \ 87 (dnp)->dn_used : (dnp)->dn_used << SPA_MINBLOCKSHIFT) 88 89 #define EPB(blkshift, typeshift) (1 << (blkshift - typeshift)) 90 91 struct dmu_buf_impl; 92 struct objset; 93 struct zio; 94 95 enum dnode_dirtycontext { 96 DN_UNDIRTIED, 97 DN_DIRTY_OPEN, 98 DN_DIRTY_SYNC 99 }; 100 101 /* Is dn_used in bytes? if not, it's in multiples of SPA_MINBLOCKSIZE */ 102 #define DNODE_FLAG_USED_BYTES (1<<0) 103 #define DNODE_FLAG_USERUSED_ACCOUNTED (1<<1) 104 105 typedef struct dnode_phys { 106 uint8_t dn_type; /* dmu_object_type_t */ 107 uint8_t dn_indblkshift; /* ln2(indirect block size) */ 108 uint8_t dn_nlevels; /* 1=dn_blkptr->data blocks */ 109 uint8_t dn_nblkptr; /* length of dn_blkptr */ 110 uint8_t dn_bonustype; /* type of data in bonus buffer */ 111 uint8_t dn_checksum; /* ZIO_CHECKSUM type */ 112 uint8_t dn_compress; /* ZIO_COMPRESS type */ 113 uint8_t dn_flags; /* DNODE_FLAG_* */ 114 uint16_t dn_datablkszsec; /* data block size in 512b sectors */ 115 uint16_t dn_bonuslen; /* length of dn_bonus */ 116 uint8_t dn_pad2[4]; 117 118 /* accounting is protected by dn_dirty_mtx */ 119 uint64_t dn_maxblkid; /* largest allocated block ID */ 120 uint64_t dn_used; /* bytes (or sectors) of disk space */ 121 122 uint64_t dn_pad3[4]; 123 124 blkptr_t dn_blkptr[1]; 125 uint8_t dn_bonus[DN_MAX_BONUSLEN]; 126 } dnode_phys_t; 127 128 typedef struct dnode { 129 /* 130 * dn_struct_rwlock protects the structure of the dnode, 131 * including the number of levels of indirection (dn_nlevels), 132 * dn_maxblkid, and dn_next_* 133 */ 134 krwlock_t dn_struct_rwlock; 135 136 /* Our link on dn_objset->os_dnodes list; protected by os_lock. */ 137 list_node_t dn_link; 138 139 /* immutable: */ 140 struct objset *dn_objset; 141 uint64_t dn_object; 142 struct dmu_buf_impl *dn_dbuf; 143 dnode_phys_t *dn_phys; /* pointer into dn->dn_dbuf->db.db_data */ 144 145 /* 146 * Copies of stuff in dn_phys. They're valid in the open 147 * context (eg. even before the dnode is first synced). 148 * Where necessary, these are protected by dn_struct_rwlock. 149 */ 150 dmu_object_type_t dn_type; /* object type */ 151 uint16_t dn_bonuslen; /* bonus length */ 152 uint8_t dn_bonustype; /* bonus type */ 153 uint8_t dn_nblkptr; /* number of blkptrs (immutable) */ 154 uint8_t dn_checksum; /* ZIO_CHECKSUM type */ 155 uint8_t dn_compress; /* ZIO_COMPRESS type */ 156 uint8_t dn_nlevels; 157 uint8_t dn_indblkshift; 158 uint8_t dn_datablkshift; /* zero if blksz not power of 2! */ 159 uint16_t dn_datablkszsec; /* in 512b sectors */ 160 uint32_t dn_datablksz; /* in bytes */ 161 uint64_t dn_maxblkid; 162 uint8_t dn_next_nblkptr[TXG_SIZE]; 163 uint8_t dn_next_nlevels[TXG_SIZE]; 164 uint8_t dn_next_indblkshift[TXG_SIZE]; 165 uint16_t dn_next_bonuslen[TXG_SIZE]; 166 uint32_t dn_next_blksz[TXG_SIZE]; /* next block size in bytes */ 167 168 /* protected by os_lock: */ 169 list_node_t dn_dirty_link[TXG_SIZE]; /* next on dataset's dirty */ 170 171 /* protected by dn_mtx: */ 172 kmutex_t dn_mtx; 173 list_t dn_dirty_records[TXG_SIZE]; 174 avl_tree_t dn_ranges[TXG_SIZE]; 175 uint64_t dn_allocated_txg; 176 uint64_t dn_free_txg; 177 uint64_t dn_assigned_txg; 178 kcondvar_t dn_notxholds; 179 enum dnode_dirtycontext dn_dirtyctx; 180 uint8_t *dn_dirtyctx_firstset; /* dbg: contents meaningless */ 181 182 /* protected by own devices */ 183 refcount_t dn_tx_holds; 184 refcount_t dn_holds; 185 186 kmutex_t dn_dbufs_mtx; 187 list_t dn_dbufs; /* linked list of descendent dbuf_t's */ 188 struct dmu_buf_impl *dn_bonus; /* bonus buffer dbuf */ 189 190 /* parent IO for current sync write */ 191 zio_t *dn_zio; 192 193 /* used in syncing context */ 194 dnode_phys_t *dn_oldphys; 195 196 /* holds prefetch structure */ 197 struct zfetch dn_zfetch; 198 } dnode_t; 199 200 typedef struct free_range { 201 avl_node_t fr_node; 202 uint64_t fr_blkid; 203 uint64_t fr_nblks; 204 } free_range_t; 205 206 dnode_t *dnode_special_open(struct objset *dd, dnode_phys_t *dnp, 207 uint64_t object); 208 void dnode_special_close(dnode_t *dn); 209 210 void dnode_setbonuslen(dnode_t *dn, int newsize, dmu_tx_t *tx); 211 int dnode_hold(struct objset *dd, uint64_t object, 212 void *ref, dnode_t **dnp); 213 int dnode_hold_impl(struct objset *dd, uint64_t object, int flag, 214 void *ref, dnode_t **dnp); 215 boolean_t dnode_add_ref(dnode_t *dn, void *ref); 216 void dnode_rele(dnode_t *dn, void *ref); 217 void dnode_setdirty(dnode_t *dn, dmu_tx_t *tx); 218 void dnode_sync(dnode_t *dn, dmu_tx_t *tx); 219 void dnode_allocate(dnode_t *dn, dmu_object_type_t ot, int blocksize, int ibs, 220 dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx); 221 void dnode_reallocate(dnode_t *dn, dmu_object_type_t ot, int blocksize, 222 dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx); 223 void dnode_free(dnode_t *dn, dmu_tx_t *tx); 224 void dnode_byteswap(dnode_phys_t *dnp); 225 void dnode_buf_byteswap(void *buf, size_t size); 226 void dnode_verify(dnode_t *dn); 227 int dnode_set_blksz(dnode_t *dn, uint64_t size, int ibs, dmu_tx_t *tx); 228 uint64_t dnode_current_max_length(dnode_t *dn); 229 void dnode_free_range(dnode_t *dn, uint64_t off, uint64_t len, dmu_tx_t *tx); 230 void dnode_clear_range(dnode_t *dn, uint64_t blkid, 231 uint64_t nblks, dmu_tx_t *tx); 232 void dnode_diduse_space(dnode_t *dn, int64_t space); 233 void dnode_willuse_space(dnode_t *dn, int64_t space, dmu_tx_t *tx); 234 void dnode_new_blkid(dnode_t *dn, uint64_t blkid, dmu_tx_t *tx, boolean_t); 235 uint64_t dnode_block_freed(dnode_t *dn, uint64_t blkid); 236 void dnode_init(void); 237 void dnode_fini(void); 238 int dnode_next_offset(dnode_t *dn, int flags, uint64_t *off, 239 int minlvl, uint64_t blkfill, uint64_t txg); 240 void dnode_evict_dbufs(dnode_t *dn); 241 242 #ifdef ZFS_DEBUG 243 244 /* 245 * There should be a ## between the string literal and fmt, to make it 246 * clear that we're joining two strings together, but that piece of shit 247 * gcc doesn't support that preprocessor token. 248 */ 249 #define dprintf_dnode(dn, fmt, ...) do { \ 250 if (zfs_flags & ZFS_DEBUG_DPRINTF) { \ 251 char __db_buf[32]; \ 252 uint64_t __db_obj = (dn)->dn_object; \ 253 if (__db_obj == DMU_META_DNODE_OBJECT) \ 254 (void) strcpy(__db_buf, "mdn"); \ 255 else \ 256 (void) snprintf(__db_buf, sizeof (__db_buf), "%lld", \ 257 (u_longlong_t)__db_obj);\ 258 dprintf_ds((dn)->dn_objset->os_dsl_dataset, "obj=%s " fmt, \ 259 __db_buf, __VA_ARGS__); \ 260 } \ 261 _NOTE(CONSTCOND) } while (0) 262 263 #define DNODE_VERIFY(dn) dnode_verify(dn) 264 #define FREE_VERIFY(db, start, end, tx) free_verify(db, start, end, tx) 265 266 #else 267 268 #define dprintf_dnode(db, fmt, ...) 269 #define DNODE_VERIFY(dn) 270 #define FREE_VERIFY(db, start, end, tx) 271 272 #endif 273 274 #ifdef __cplusplus 275 } 276 #endif 277 278 #endif /* _SYS_DNODE_H */ 279