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 (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 23 * Copyright (c) 2012, 2017 by Delphix. All rights reserved. 24 * Copyright (c) 2013 by Saso Kiselkov. All rights reserved. 25 */ 26 27 #ifndef _SYS_ARC_H 28 #define _SYS_ARC_H 29 30 #include <sys/zfs_context.h> 31 32 #ifdef __cplusplus 33 extern "C" { 34 #endif 35 36 #include <sys/zio.h> 37 #include <sys/dmu.h> 38 #include <sys/spa.h> 39 40 /* 41 * Used by arc_flush() to inform arc_evict_state() that it should evict 42 * all available buffers from the arc state being passed in. 43 */ 44 #define ARC_EVICT_ALL -1ULL 45 46 #define HDR_SET_LSIZE(hdr, x) do { \ 47 ASSERT(IS_P2ALIGNED(x, 1U << SPA_MINBLOCKSHIFT)); \ 48 (hdr)->b_lsize = ((x) >> SPA_MINBLOCKSHIFT); \ 49 _NOTE(CONSTCOND) } while (0) 50 51 #define HDR_SET_PSIZE(hdr, x) do { \ 52 ASSERT(IS_P2ALIGNED((x), 1U << SPA_MINBLOCKSHIFT)); \ 53 (hdr)->b_psize = ((x) >> SPA_MINBLOCKSHIFT); \ 54 _NOTE(CONSTCOND) } while (0) 55 56 #define HDR_GET_LSIZE(hdr) ((hdr)->b_lsize << SPA_MINBLOCKSHIFT) 57 #define HDR_GET_PSIZE(hdr) ((hdr)->b_psize << SPA_MINBLOCKSHIFT) 58 59 typedef struct arc_buf_hdr arc_buf_hdr_t; 60 typedef struct arc_buf arc_buf_t; 61 typedef struct arc_prune arc_prune_t; 62 63 /* 64 * Because the ARC can store encrypted data, errors (not due to bugs) may arise 65 * while transforming data into its desired format - specifically, when 66 * decrypting, the key may not be present, or the HMAC may not be correct 67 * which signifies deliberate tampering with the on-disk state 68 * (assuming that the checksum was correct). If any error occurs, the "buf" 69 * parameter will be NULL. 70 */ 71 typedef void arc_read_done_func_t(zio_t *zio, const zbookmark_phys_t *zb, 72 const blkptr_t *bp, arc_buf_t *buf, void *priv); 73 typedef void arc_write_done_func_t(zio_t *zio, arc_buf_t *buf, void *priv); 74 typedef void arc_prune_func_t(int64_t bytes, void *priv); 75 76 /* Shared module parameters */ 77 extern uint64_t zfs_arc_average_blocksize; 78 79 /* generic arc_done_func_t's which you can use */ 80 arc_read_done_func_t arc_bcopy_func; 81 arc_read_done_func_t arc_getbuf_func; 82 83 /* generic arc_prune_func_t wrapper for callbacks */ 84 struct arc_prune { 85 arc_prune_func_t *p_pfunc; 86 void *p_private; 87 uint64_t p_adjust; 88 list_node_t p_node; 89 zfs_refcount_t p_refcnt; 90 }; 91 92 typedef enum arc_strategy { 93 ARC_STRATEGY_META_ONLY = 0, /* Evict only meta data buffers */ 94 ARC_STRATEGY_META_BALANCED = 1, /* Evict data buffers if needed */ 95 } arc_strategy_t; 96 97 typedef enum arc_flags 98 { 99 /* 100 * Public flags that can be passed into the ARC by external consumers. 101 */ 102 ARC_FLAG_WAIT = 1 << 0, /* perform sync I/O */ 103 ARC_FLAG_NOWAIT = 1 << 1, /* perform async I/O */ 104 ARC_FLAG_PREFETCH = 1 << 2, /* I/O is a prefetch */ 105 ARC_FLAG_CACHED = 1 << 3, /* I/O was in cache */ 106 ARC_FLAG_L2CACHE = 1 << 4, /* cache in L2ARC */ 107 ARC_FLAG_PREDICTIVE_PREFETCH = 1 << 5, /* I/O from zfetch */ 108 ARC_FLAG_PRESCIENT_PREFETCH = 1 << 6, /* long min lifespan */ 109 110 /* 111 * Private ARC flags. These flags are private ARC only flags that 112 * will show up in b_flags in the arc_hdr_buf_t. These flags should 113 * only be set by ARC code. 114 */ 115 ARC_FLAG_IN_HASH_TABLE = 1 << 7, /* buffer is hashed */ 116 ARC_FLAG_IO_IN_PROGRESS = 1 << 8, /* I/O in progress */ 117 ARC_FLAG_IO_ERROR = 1 << 9, /* I/O failed for buf */ 118 ARC_FLAG_INDIRECT = 1 << 10, /* indirect block */ 119 /* Indicates that block was read with ASYNC priority. */ 120 ARC_FLAG_PRIO_ASYNC_READ = 1 << 11, 121 ARC_FLAG_L2_WRITING = 1 << 12, /* write in progress */ 122 ARC_FLAG_L2_EVICTED = 1 << 13, /* evicted during I/O */ 123 ARC_FLAG_L2_WRITE_HEAD = 1 << 14, /* head of write list */ 124 /* indicates that the buffer contains metadata (otherwise, data) */ 125 ARC_FLAG_BUFC_METADATA = 1 << 15, 126 127 /* Flags specifying whether optional hdr struct fields are defined */ 128 ARC_FLAG_HAS_L1HDR = 1 << 16, 129 ARC_FLAG_HAS_L2HDR = 1 << 17, 130 131 /* 132 * Indicates the arc_buf_hdr_t's b_pdata matches the on-disk data. 133 * This allows the l2arc to use the blkptr's checksum to verify 134 * the data without having to store the checksum in the hdr. 135 */ 136 ARC_FLAG_COMPRESSED_ARC = 1 << 18, 137 ARC_FLAG_SHARED_DATA = 1 << 19, 138 139 /* 140 * The arc buffer's compression mode is stored in the top 7 bits of the 141 * flags field, so these dummy flags are included so that MDB can 142 * interpret the enum properly. 143 */ 144 ARC_FLAG_COMPRESS_0 = 1 << 24, 145 ARC_FLAG_COMPRESS_1 = 1 << 25, 146 ARC_FLAG_COMPRESS_2 = 1 << 26, 147 ARC_FLAG_COMPRESS_3 = 1 << 27, 148 ARC_FLAG_COMPRESS_4 = 1 << 28, 149 ARC_FLAG_COMPRESS_5 = 1 << 29, 150 ARC_FLAG_COMPRESS_6 = 1 << 30 151 152 } arc_flags_t; 153 154 typedef enum arc_buf_flags { 155 ARC_BUF_FLAG_SHARED = 1 << 0, 156 ARC_BUF_FLAG_COMPRESSED = 1 << 1 157 } arc_buf_flags_t; 158 159 struct arc_buf { 160 arc_buf_hdr_t *b_hdr; 161 arc_buf_t *b_next; 162 kmutex_t b_evict_lock; 163 void *b_data; 164 arc_buf_flags_t b_flags; 165 }; 166 167 typedef enum arc_buf_contents { 168 ARC_BUFC_INVALID, /* invalid type */ 169 ARC_BUFC_DATA, /* buffer contains data */ 170 ARC_BUFC_METADATA, /* buffer contains metadata */ 171 ARC_BUFC_NUMTYPES 172 } arc_buf_contents_t; 173 174 /* 175 * The following breakdows of arc_size exist for kstat only. 176 */ 177 typedef enum arc_space_type { 178 ARC_SPACE_DATA, 179 ARC_SPACE_META, 180 ARC_SPACE_HDRS, 181 ARC_SPACE_L2HDRS, 182 ARC_SPACE_DBUF, 183 ARC_SPACE_DNODE, 184 ARC_SPACE_BONUS, 185 ARC_SPACE_NUMTYPES 186 } arc_space_type_t; 187 188 typedef enum arc_state_type { 189 ARC_STATE_ANON, 190 ARC_STATE_MRU, 191 ARC_STATE_MRU_GHOST, 192 ARC_STATE_MFU, 193 ARC_STATE_MFU_GHOST, 194 ARC_STATE_L2C_ONLY, 195 ARC_STATE_NUMTYPES 196 } arc_state_type_t; 197 198 typedef struct arc_buf_info { 199 arc_state_type_t abi_state_type; 200 arc_buf_contents_t abi_state_contents; 201 uint64_t abi_state_index; 202 uint32_t abi_flags; 203 uint32_t abi_bufcnt; 204 uint64_t abi_size; 205 uint64_t abi_spa; 206 uint64_t abi_access; 207 uint32_t abi_mru_hits; 208 uint32_t abi_mru_ghost_hits; 209 uint32_t abi_mfu_hits; 210 uint32_t abi_mfu_ghost_hits; 211 uint32_t abi_l2arc_hits; 212 uint32_t abi_holds; 213 uint64_t abi_l2arc_dattr; 214 uint64_t abi_l2arc_asize; 215 enum zio_compress abi_l2arc_compress; 216 } arc_buf_info_t; 217 218 void arc_space_consume(uint64_t space, arc_space_type_t type); 219 void arc_space_return(uint64_t space, arc_space_type_t type); 220 boolean_t arc_is_metadata(arc_buf_t *buf); 221 enum zio_compress arc_get_compression(arc_buf_t *buf); 222 int arc_decompress(arc_buf_t *buf); 223 arc_buf_t *arc_alloc_buf(spa_t *spa, void *tag, arc_buf_contents_t type, 224 int32_t size); 225 arc_buf_t *arc_alloc_compressed_buf(spa_t *spa, void *tag, 226 uint64_t psize, uint64_t lsize, enum zio_compress compression_type); 227 arc_buf_t *arc_loan_buf(spa_t *spa, boolean_t is_metadata, int size); 228 arc_buf_t *arc_loan_compressed_buf(spa_t *spa, uint64_t psize, uint64_t lsize, 229 enum zio_compress compression_type); 230 void arc_return_buf(arc_buf_t *buf, void *tag); 231 void arc_loan_inuse_buf(arc_buf_t *buf, void *tag); 232 void arc_buf_destroy(arc_buf_t *buf, void *tag); 233 void arc_buf_info(arc_buf_t *buf, arc_buf_info_t *abi, int state_index); 234 int arc_buf_size(arc_buf_t *buf); 235 int arc_buf_lsize(arc_buf_t *buf); 236 void arc_buf_access(arc_buf_t *buf); 237 void arc_release(arc_buf_t *buf, void *tag); 238 int arc_released(arc_buf_t *buf); 239 void arc_buf_freeze(arc_buf_t *buf); 240 void arc_buf_thaw(arc_buf_t *buf); 241 #ifdef ZFS_DEBUG 242 int arc_referenced(arc_buf_t *buf); 243 #endif 244 245 int arc_read(zio_t *pio, spa_t *spa, const blkptr_t *bp, 246 arc_read_done_func_t *done, void *priv, zio_priority_t priority, 247 int flags, arc_flags_t *arc_flags, const zbookmark_phys_t *zb); 248 zio_t *arc_write(zio_t *pio, spa_t *spa, uint64_t txg, 249 blkptr_t *bp, arc_buf_t *buf, boolean_t l2arc, const zio_prop_t *zp, 250 arc_write_done_func_t *ready, arc_write_done_func_t *child_ready, 251 arc_write_done_func_t *physdone, arc_write_done_func_t *done, 252 void *priv, zio_priority_t priority, int zio_flags, 253 const zbookmark_phys_t *zb); 254 255 arc_prune_t *arc_add_prune_callback(arc_prune_func_t *func, void *priv); 256 void arc_remove_prune_callback(arc_prune_t *p); 257 void arc_freed(spa_t *spa, const blkptr_t *bp); 258 259 void arc_flush(spa_t *spa, boolean_t retry); 260 void arc_tempreserve_clear(uint64_t reserve); 261 int arc_tempreserve_space(spa_t *spa, uint64_t reserve, uint64_t txg); 262 263 uint64_t arc_max_bytes(void); 264 void arc_init(void); 265 void arc_fini(void); 266 267 /* 268 * Level 2 ARC 269 */ 270 271 void l2arc_add_vdev(spa_t *spa, vdev_t *vd); 272 void l2arc_remove_vdev(vdev_t *vd); 273 boolean_t l2arc_vdev_present(vdev_t *vd); 274 void l2arc_init(void); 275 void l2arc_fini(void); 276 void l2arc_start(void); 277 void l2arc_stop(void); 278 279 #ifdef illumos 280 #ifndef _KERNEL 281 extern boolean_t arc_watch; 282 extern int arc_procfd; 283 #endif 284 #endif /* illumos */ 285 286 #ifdef __cplusplus 287 } 288 #endif 289 290 #endif /* _SYS_ARC_H */ 291