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 2007 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 #pragma ident "%Z%%M% %I% %E% SMI" 30 31 #include <sys/zfs_context.h> 32 #include <sys/avl.h> 33 #include <sys/spa.h> 34 #include <sys/txg.h> 35 #include <sys/zio.h> 36 #include <sys/refcount.h> 37 #include <sys/dmu_zfetch.h> 38 39 #ifdef __cplusplus 40 extern "C" { 41 #endif 42 43 /* 44 * Flags. 45 */ 46 #define DNODE_MUST_BE_ALLOCATED 1 47 #define DNODE_MUST_BE_FREE 2 48 49 /* 50 * Fixed constants. 51 */ 52 #define DNODE_SHIFT 9 /* 512 bytes */ 53 #define DN_MIN_INDBLKSHIFT 10 /* 1k */ 54 #define DN_MAX_INDBLKSHIFT 14 /* 16k */ 55 #define DNODE_BLOCK_SHIFT 14 /* 16k */ 56 #define DNODE_CORE_SIZE 64 /* 64 bytes for dnode sans blkptrs */ 57 #define DN_MAX_OBJECT_SHIFT 48 /* 256 trillion (zfs_fid_t limit) */ 58 #define DN_MAX_OFFSET_SHIFT 64 /* 2^64 bytes in a dnode */ 59 60 /* 61 * Derived constants. 62 */ 63 #define DNODE_SIZE (1 << DNODE_SHIFT) 64 #define DN_MAX_NBLKPTR ((DNODE_SIZE - DNODE_CORE_SIZE) >> SPA_BLKPTRSHIFT) 65 #define DN_MAX_BONUSLEN (DNODE_SIZE - DNODE_CORE_SIZE - (1 << SPA_BLKPTRSHIFT)) 66 #define DN_MAX_OBJECT (1ULL << DN_MAX_OBJECT_SHIFT) 67 68 #define DNODES_PER_BLOCK_SHIFT (DNODE_BLOCK_SHIFT - DNODE_SHIFT) 69 #define DNODES_PER_BLOCK (1ULL << DNODES_PER_BLOCK_SHIFT) 70 #define DNODES_PER_LEVEL_SHIFT (DN_MAX_INDBLKSHIFT - SPA_BLKPTRSHIFT) 71 72 /* The +2 here is a cheesy way to round up */ 73 #define DN_MAX_LEVELS (2 + ((DN_MAX_OFFSET_SHIFT - SPA_MINBLOCKSHIFT) / \ 74 (DN_MIN_INDBLKSHIFT - SPA_BLKPTRSHIFT))) 75 76 #define DN_BONUS(dnp) ((void*)((dnp)->dn_bonus + \ 77 (((dnp)->dn_nblkptr - 1) * sizeof (blkptr_t)))) 78 79 #define DN_USED_BYTES(dnp) (((dnp)->dn_flags & DNODE_FLAG_USED_BYTES) ? \ 80 (dnp)->dn_used : (dnp)->dn_used << SPA_MINBLOCKSHIFT) 81 82 #define EPB(blkshift, typeshift) (1 << (blkshift - typeshift)) 83 84 struct dmu_buf_impl; 85 struct objset_impl; 86 struct zio; 87 88 enum dnode_dirtycontext { 89 DN_UNDIRTIED, 90 DN_DIRTY_OPEN, 91 DN_DIRTY_SYNC 92 }; 93 94 /* Is dn_used in bytes? if not, it's in multiples of SPA_MINBLOCKSIZE */ 95 #define DNODE_FLAG_USED_BYTES (1<<0) 96 97 typedef struct dnode_phys { 98 uint8_t dn_type; /* dmu_object_type_t */ 99 uint8_t dn_indblkshift; /* ln2(indirect block size) */ 100 uint8_t dn_nlevels; /* 1=dn_blkptr->data blocks */ 101 uint8_t dn_nblkptr; /* length of dn_blkptr */ 102 uint8_t dn_bonustype; /* type of data in bonus buffer */ 103 uint8_t dn_checksum; /* ZIO_CHECKSUM type */ 104 uint8_t dn_compress; /* ZIO_COMPRESS type */ 105 uint8_t dn_flags; /* DNODE_FLAG_* */ 106 uint16_t dn_datablkszsec; /* data block size in 512b sectors */ 107 uint16_t dn_bonuslen; /* length of dn_bonus */ 108 uint8_t dn_pad2[4]; 109 110 /* accounting is protected by dn_dirty_mtx */ 111 uint64_t dn_maxblkid; /* largest allocated block ID */ 112 uint64_t dn_used; /* bytes (or sectors) of disk space */ 113 114 uint64_t dn_pad3[4]; 115 116 blkptr_t dn_blkptr[1]; 117 uint8_t dn_bonus[DN_MAX_BONUSLEN]; 118 } dnode_phys_t; 119 120 typedef struct dnode { 121 /* 122 * dn_struct_rwlock protects the structure of the dnode, 123 * including the number of levels of indirection (dn_nlevels), 124 * dn_maxblkid, and dn_next_* 125 */ 126 krwlock_t dn_struct_rwlock; 127 128 /* 129 * Our link on dataset's dd_dnodes list. 130 * Protected by dd_accounting_mtx. 131 */ 132 list_node_t dn_link; 133 134 /* immutable: */ 135 struct objset_impl *dn_objset; 136 uint64_t dn_object; 137 struct dmu_buf_impl *dn_dbuf; 138 dnode_phys_t *dn_phys; /* pointer into dn->dn_dbuf->db.db_data */ 139 140 /* 141 * Copies of stuff in dn_phys. They're valid in the open 142 * context (eg. even before the dnode is first synced). 143 * Where necessary, these are protected by dn_struct_rwlock. 144 */ 145 dmu_object_type_t dn_type; /* object type */ 146 uint16_t dn_bonuslen; /* bonus length */ 147 uint8_t dn_bonustype; /* bonus type */ 148 uint8_t dn_nblkptr; /* number of blkptrs (immutable) */ 149 uint8_t dn_checksum; /* ZIO_CHECKSUM type */ 150 uint8_t dn_compress; /* ZIO_COMPRESS type */ 151 uint8_t dn_nlevels; 152 uint8_t dn_indblkshift; 153 uint8_t dn_datablkshift; /* zero if blksz not power of 2! */ 154 uint16_t dn_datablkszsec; /* in 512b sectors */ 155 uint32_t dn_datablksz; /* in bytes */ 156 uint64_t dn_maxblkid; 157 uint8_t dn_next_nlevels[TXG_SIZE]; 158 uint8_t dn_next_indblkshift[TXG_SIZE]; 159 uint32_t dn_next_blksz[TXG_SIZE]; /* next block size in bytes */ 160 161 /* protected by os_lock: */ 162 list_node_t dn_dirty_link[TXG_SIZE]; /* next on dataset's dirty */ 163 164 /* protected by dn_mtx: */ 165 kmutex_t dn_mtx; 166 list_t dn_dirty_records[TXG_SIZE]; 167 avl_tree_t dn_ranges[TXG_SIZE]; 168 uint64_t dn_allocated_txg; 169 uint64_t dn_free_txg; 170 uint64_t dn_assigned_txg; 171 kcondvar_t dn_notxholds; 172 enum dnode_dirtycontext dn_dirtyctx; 173 uint8_t *dn_dirtyctx_firstset; /* dbg: contents meaningless */ 174 175 /* protected by own devices */ 176 refcount_t dn_tx_holds; 177 refcount_t dn_holds; 178 179 kmutex_t dn_dbufs_mtx; 180 list_t dn_dbufs; /* linked list of descendent dbuf_t's */ 181 struct dmu_buf_impl *dn_bonus; /* bonus buffer dbuf */ 182 183 /* parent IO for current sync write */ 184 zio_t *dn_zio; 185 186 /* holds prefetch structure */ 187 struct zfetch dn_zfetch; 188 } dnode_t; 189 190 typedef struct free_range { 191 avl_node_t fr_node; 192 uint64_t fr_blkid; 193 uint64_t fr_nblks; 194 } free_range_t; 195 196 dnode_t *dnode_special_open(struct objset_impl *dd, dnode_phys_t *dnp, 197 uint64_t object); 198 void dnode_special_close(dnode_t *dn); 199 200 int dnode_hold(struct objset_impl *dd, uint64_t object, 201 void *ref, dnode_t **dnp); 202 int dnode_hold_impl(struct objset_impl *dd, uint64_t object, int flag, 203 void *ref, dnode_t **dnp); 204 void dnode_add_ref(dnode_t *dn, void *ref); 205 void dnode_rele(dnode_t *dn, void *ref); 206 void dnode_setdirty(dnode_t *dn, dmu_tx_t *tx); 207 void dnode_sync(dnode_t *dn, dmu_tx_t *tx); 208 void dnode_allocate(dnode_t *dn, dmu_object_type_t ot, int blocksize, int ibs, 209 dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx); 210 void dnode_reallocate(dnode_t *dn, dmu_object_type_t ot, int blocksize, 211 dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx); 212 void dnode_free(dnode_t *dn, dmu_tx_t *tx); 213 void dnode_byteswap(dnode_phys_t *dnp); 214 void dnode_buf_byteswap(void *buf, size_t size); 215 void dnode_verify(dnode_t *dn); 216 int dnode_set_blksz(dnode_t *dn, uint64_t size, int ibs, dmu_tx_t *tx); 217 uint64_t dnode_current_max_length(dnode_t *dn); 218 void dnode_free_range(dnode_t *dn, uint64_t off, uint64_t len, dmu_tx_t *tx); 219 void dnode_clear_range(dnode_t *dn, uint64_t blkid, 220 uint64_t nblks, dmu_tx_t *tx); 221 void dnode_diduse_space(dnode_t *dn, int64_t space); 222 void dnode_willuse_space(dnode_t *dn, int64_t space, dmu_tx_t *tx); 223 void dnode_new_blkid(dnode_t *dn, uint64_t blkid, dmu_tx_t *tx); 224 uint64_t dnode_block_freed(dnode_t *dn, uint64_t blkid); 225 void dnode_init(void); 226 void dnode_fini(void); 227 int dnode_next_offset(dnode_t *dn, boolean_t hole, uint64_t *off, int minlvl, 228 uint64_t blkfill, uint64_t txg); 229 int dnode_evict_dbufs(dnode_t *dn, boolean_t try); 230 231 #ifdef ZFS_DEBUG 232 233 /* 234 * There should be a ## between the string literal and fmt, to make it 235 * clear that we're joining two strings together, but that piece of shit 236 * gcc doesn't support that preprocessor token. 237 */ 238 #define dprintf_dnode(dn, fmt, ...) do { \ 239 if (zfs_flags & ZFS_DEBUG_DPRINTF) { \ 240 char __db_buf[32]; \ 241 uint64_t __db_obj = (dn)->dn_object; \ 242 if (__db_obj == DMU_META_DNODE_OBJECT) \ 243 (void) strcpy(__db_buf, "mdn"); \ 244 else \ 245 (void) snprintf(__db_buf, sizeof (__db_buf), "%lld", \ 246 (u_longlong_t)__db_obj);\ 247 dprintf_ds((dn)->dn_objset->os_dsl_dataset, "obj=%s " fmt, \ 248 __db_buf, __VA_ARGS__); \ 249 } \ 250 _NOTE(CONSTCOND) } while (0) 251 252 #define DNODE_VERIFY(dn) dnode_verify(dn) 253 #define FREE_VERIFY(db, start, end, tx) free_verify(db, start, end, tx) 254 255 #else 256 257 #define dprintf_dnode(db, fmt, ...) 258 #define DNODE_VERIFY(dn) 259 #define FREE_VERIFY(db, start, end, tx) 260 261 #endif 262 263 #ifdef __cplusplus 264 } 265 #endif 266 267 #endif /* _SYS_DNODE_H */ 268