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) 2010, Oracle and/or its affiliates. All rights reserved.
| 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) 2010, Oracle and/or its affiliates. All rights reserved.
|
| 23 * Copyright (c) 2012 by Delphix. All rights reserved.
|
23 */ 24 25#ifndef _SYS_SA_IMPL_H 26#define _SYS_SA_IMPL_H 27 28#include <sys/dmu.h> 29#include <sys/refcount.h> 30#include <sys/list.h> 31 32/* 33 * Array of known attributes and their 34 * various characteristics. 35 */ 36typedef struct sa_attr_table { 37 sa_attr_type_t sa_attr; 38 uint8_t sa_registered; 39 uint16_t sa_length; 40 sa_bswap_type_t sa_byteswap; 41 char *sa_name; 42} sa_attr_table_t; 43 44/* 45 * Zap attribute format for attribute registration 46 * 47 * 64 56 48 40 32 24 16 8 0 48 * +-------+-------+-------+-------+-------+-------+-------+-------+ 49 * | unused | len | bswap | attr num | 50 * +-------+-------+-------+-------+-------+-------+-------+-------+ 51 * 52 * Zap attribute format for layout information. 53 * 54 * layout information is stored as an array of attribute numbers 55 * The name of the attribute is the layout number (0, 1, 2, ...) 56 * 57 * 16 0 58 * +---- ---+ 59 * | attr # | 60 * +--------+ 61 * | attr # | 62 * +--- ----+ 63 * ...... 64 * 65 */ 66 67#define ATTR_BSWAP(x) BF32_GET(x, 16, 8) 68#define ATTR_LENGTH(x) BF32_GET(x, 24, 16) 69#define ATTR_NUM(x) BF32_GET(x, 0, 16) 70#define ATTR_ENCODE(x, attr, length, bswap) \ 71{ \ 72 BF64_SET(x, 24, 16, length); \ 73 BF64_SET(x, 16, 8, bswap); \ 74 BF64_SET(x, 0, 16, attr); \ 75} 76 77#define TOC_OFF(x) BF32_GET(x, 0, 23) 78#define TOC_ATTR_PRESENT(x) BF32_GET(x, 31, 1) 79#define TOC_LEN_IDX(x) BF32_GET(x, 24, 4) 80#define TOC_ATTR_ENCODE(x, len_idx, offset) \ 81{ \ 82 BF32_SET(x, 31, 1, 1); \ 83 BF32_SET(x, 24, 7, len_idx); \ 84 BF32_SET(x, 0, 24, offset); \ 85} 86 87#define SA_LAYOUTS "LAYOUTS" 88#define SA_REGISTRY "REGISTRY" 89 90/* 91 * Each unique layout will have their own table 92 * sa_lot (layout_table) 93 */ 94typedef struct sa_lot { 95 avl_node_t lot_num_node; 96 avl_node_t lot_hash_node; 97 uint64_t lot_num; 98 uint64_t lot_hash; 99 sa_attr_type_t *lot_attrs; /* array of attr #'s */ 100 uint32_t lot_var_sizes; /* how many aren't fixed size */ 101 uint32_t lot_attr_count; /* total attr count */ 102 list_t lot_idx_tab; /* should be only a couple of entries */ 103 int lot_instance; /* used with lot_hash to identify entry */ 104} sa_lot_t; 105 106/* index table of offsets */ 107typedef struct sa_idx_tab { 108 list_node_t sa_next; 109 sa_lot_t *sa_layout; 110 uint16_t *sa_variable_lengths; 111 refcount_t sa_refcount; 112 uint32_t *sa_idx_tab; /* array of offsets */ 113} sa_idx_tab_t; 114 115/* 116 * Since the offset/index information into the actual data 117 * will usually be identical we can share that information with 118 * all handles that have the exact same offsets. 119 * 120 * You would typically only have a large number of different table of 121 * contents if you had a several variable sized attributes. 122 * 123 * Two AVL trees are used to track the attribute layout numbers. 124 * one is keyed by number and will be consulted when a DMU_OT_SA 125 * object is first read. The second tree is keyed by the hash signature 126 * of the attributes and will be consulted when an attribute is added 127 * to determine if we already have an instance of that layout. Both 128 * of these tree's are interconnected. The only difference is that 129 * when an entry is found in the "hash" tree the list of attributes will 130 * need to be compared against the list of attributes you have in hand. 131 * The assumption is that typically attributes will just be updated and 132 * adding a completely new attribute is a very rare operation. 133 */ 134struct sa_os { 135 kmutex_t sa_lock; 136 boolean_t sa_need_attr_registration; 137 boolean_t sa_force_spill; 138 uint64_t sa_master_obj; 139 uint64_t sa_reg_attr_obj; 140 uint64_t sa_layout_attr_obj; 141 int sa_num_attrs; 142 sa_attr_table_t *sa_attr_table; /* private attr table */ 143 sa_update_cb_t *sa_update_cb; 144 avl_tree_t sa_layout_num_tree; /* keyed by layout number */ 145 avl_tree_t sa_layout_hash_tree; /* keyed by layout hash value */ 146 int sa_user_table_sz; 147 sa_attr_type_t *sa_user_table; /* user name->attr mapping table */ 148}; 149 150/* 151 * header for all bonus and spill buffers. 152 * The header has a fixed portion with a variable number 153 * of "lengths" depending on the number of variable sized 154 * attribues which are determined by the "layout number" 155 */ 156 157#define SA_MAGIC 0x2F505A /* ZFS SA */ 158typedef struct sa_hdr_phys { 159 uint32_t sa_magic; 160 uint16_t sa_layout_info; /* Encoded with hdrsize and layout number */ 161 uint16_t sa_lengths[1]; /* optional sizes for variable length attrs */ 162 /* ... Data follows the lengths. */ 163} sa_hdr_phys_t; 164 165/* 166 * sa_hdr_phys -> sa_layout_info 167 * 168 * 16 10 0 169 * +--------+-------+ 170 * | hdrsz |layout | 171 * +--------+-------+ 172 * 173 * Bits 0-10 are the layout number 174 * Bits 11-16 are the size of the header. 175 * The hdrsize is the number * 8 176 * 177 * For example. 178 * hdrsz of 1 ==> 8 byte header 179 * 2 ==> 16 byte header 180 * 181 */ 182 183#define SA_HDR_LAYOUT_NUM(hdr) BF32_GET(hdr->sa_layout_info, 0, 10)
| 24 */ 25 26#ifndef _SYS_SA_IMPL_H 27#define _SYS_SA_IMPL_H 28 29#include <sys/dmu.h> 30#include <sys/refcount.h> 31#include <sys/list.h> 32 33/* 34 * Array of known attributes and their 35 * various characteristics. 36 */ 37typedef struct sa_attr_table { 38 sa_attr_type_t sa_attr; 39 uint8_t sa_registered; 40 uint16_t sa_length; 41 sa_bswap_type_t sa_byteswap; 42 char *sa_name; 43} sa_attr_table_t; 44 45/* 46 * Zap attribute format for attribute registration 47 * 48 * 64 56 48 40 32 24 16 8 0 49 * +-------+-------+-------+-------+-------+-------+-------+-------+ 50 * | unused | len | bswap | attr num | 51 * +-------+-------+-------+-------+-------+-------+-------+-------+ 52 * 53 * Zap attribute format for layout information. 54 * 55 * layout information is stored as an array of attribute numbers 56 * The name of the attribute is the layout number (0, 1, 2, ...) 57 * 58 * 16 0 59 * +---- ---+ 60 * | attr # | 61 * +--------+ 62 * | attr # | 63 * +--- ----+ 64 * ...... 65 * 66 */ 67 68#define ATTR_BSWAP(x) BF32_GET(x, 16, 8) 69#define ATTR_LENGTH(x) BF32_GET(x, 24, 16) 70#define ATTR_NUM(x) BF32_GET(x, 0, 16) 71#define ATTR_ENCODE(x, attr, length, bswap) \ 72{ \ 73 BF64_SET(x, 24, 16, length); \ 74 BF64_SET(x, 16, 8, bswap); \ 75 BF64_SET(x, 0, 16, attr); \ 76} 77 78#define TOC_OFF(x) BF32_GET(x, 0, 23) 79#define TOC_ATTR_PRESENT(x) BF32_GET(x, 31, 1) 80#define TOC_LEN_IDX(x) BF32_GET(x, 24, 4) 81#define TOC_ATTR_ENCODE(x, len_idx, offset) \ 82{ \ 83 BF32_SET(x, 31, 1, 1); \ 84 BF32_SET(x, 24, 7, len_idx); \ 85 BF32_SET(x, 0, 24, offset); \ 86} 87 88#define SA_LAYOUTS "LAYOUTS" 89#define SA_REGISTRY "REGISTRY" 90 91/* 92 * Each unique layout will have their own table 93 * sa_lot (layout_table) 94 */ 95typedef struct sa_lot { 96 avl_node_t lot_num_node; 97 avl_node_t lot_hash_node; 98 uint64_t lot_num; 99 uint64_t lot_hash; 100 sa_attr_type_t *lot_attrs; /* array of attr #'s */ 101 uint32_t lot_var_sizes; /* how many aren't fixed size */ 102 uint32_t lot_attr_count; /* total attr count */ 103 list_t lot_idx_tab; /* should be only a couple of entries */ 104 int lot_instance; /* used with lot_hash to identify entry */ 105} sa_lot_t; 106 107/* index table of offsets */ 108typedef struct sa_idx_tab { 109 list_node_t sa_next; 110 sa_lot_t *sa_layout; 111 uint16_t *sa_variable_lengths; 112 refcount_t sa_refcount; 113 uint32_t *sa_idx_tab; /* array of offsets */ 114} sa_idx_tab_t; 115 116/* 117 * Since the offset/index information into the actual data 118 * will usually be identical we can share that information with 119 * all handles that have the exact same offsets. 120 * 121 * You would typically only have a large number of different table of 122 * contents if you had a several variable sized attributes. 123 * 124 * Two AVL trees are used to track the attribute layout numbers. 125 * one is keyed by number and will be consulted when a DMU_OT_SA 126 * object is first read. The second tree is keyed by the hash signature 127 * of the attributes and will be consulted when an attribute is added 128 * to determine if we already have an instance of that layout. Both 129 * of these tree's are interconnected. The only difference is that 130 * when an entry is found in the "hash" tree the list of attributes will 131 * need to be compared against the list of attributes you have in hand. 132 * The assumption is that typically attributes will just be updated and 133 * adding a completely new attribute is a very rare operation. 134 */ 135struct sa_os { 136 kmutex_t sa_lock; 137 boolean_t sa_need_attr_registration; 138 boolean_t sa_force_spill; 139 uint64_t sa_master_obj; 140 uint64_t sa_reg_attr_obj; 141 uint64_t sa_layout_attr_obj; 142 int sa_num_attrs; 143 sa_attr_table_t *sa_attr_table; /* private attr table */ 144 sa_update_cb_t *sa_update_cb; 145 avl_tree_t sa_layout_num_tree; /* keyed by layout number */ 146 avl_tree_t sa_layout_hash_tree; /* keyed by layout hash value */ 147 int sa_user_table_sz; 148 sa_attr_type_t *sa_user_table; /* user name->attr mapping table */ 149}; 150 151/* 152 * header for all bonus and spill buffers. 153 * The header has a fixed portion with a variable number 154 * of "lengths" depending on the number of variable sized 155 * attribues which are determined by the "layout number" 156 */ 157 158#define SA_MAGIC 0x2F505A /* ZFS SA */ 159typedef struct sa_hdr_phys { 160 uint32_t sa_magic; 161 uint16_t sa_layout_info; /* Encoded with hdrsize and layout number */ 162 uint16_t sa_lengths[1]; /* optional sizes for variable length attrs */ 163 /* ... Data follows the lengths. */ 164} sa_hdr_phys_t; 165 166/* 167 * sa_hdr_phys -> sa_layout_info 168 * 169 * 16 10 0 170 * +--------+-------+ 171 * | hdrsz |layout | 172 * +--------+-------+ 173 * 174 * Bits 0-10 are the layout number 175 * Bits 11-16 are the size of the header. 176 * The hdrsize is the number * 8 177 * 178 * For example. 179 * hdrsz of 1 ==> 8 byte header 180 * 2 ==> 16 byte header 181 * 182 */ 183 184#define SA_HDR_LAYOUT_NUM(hdr) BF32_GET(hdr->sa_layout_info, 0, 10)
|
184#define SA_HDR_SIZE(hdr) BF32_GET_SB(hdr->sa_layout_info, 10, 16, 3, 0)
| 185#define SA_HDR_SIZE(hdr) BF32_GET_SB(hdr->sa_layout_info, 10, 6, 3, 0)
|
185#define SA_HDR_LAYOUT_INFO_ENCODE(x, num, size) \ 186{ \ 187 BF32_SET_SB(x, 10, 6, 3, 0, size); \ 188 BF32_SET(x, 0, 10, num); \ 189} 190 191typedef enum sa_buf_type { 192 SA_BONUS = 1, 193 SA_SPILL = 2 194} sa_buf_type_t; 195 196typedef enum sa_data_op { 197 SA_LOOKUP, 198 SA_UPDATE, 199 SA_ADD, 200 SA_REPLACE, 201 SA_REMOVE 202} sa_data_op_t; 203 204/* 205 * Opaque handle used for most sa functions 206 * 207 * This needs to be kept as small as possible. 208 */ 209 210struct sa_handle { 211 kmutex_t sa_lock; 212 dmu_buf_t *sa_bonus; 213 dmu_buf_t *sa_spill; 214 objset_t *sa_os; 215 void *sa_userp; 216 sa_idx_tab_t *sa_bonus_tab; /* idx of bonus */ 217 sa_idx_tab_t *sa_spill_tab; /* only present if spill activated */ 218}; 219 220#define SA_GET_DB(hdl, type) \ 221 (dmu_buf_impl_t *)((type == SA_BONUS) ? hdl->sa_bonus : hdl->sa_spill) 222 223#define SA_GET_HDR(hdl, type) \ 224 ((sa_hdr_phys_t *)((dmu_buf_impl_t *)(SA_GET_DB(hdl, \ 225 type))->db.db_data)) 226 227#define SA_IDX_TAB_GET(hdl, type) \ 228 (type == SA_BONUS ? hdl->sa_bonus_tab : hdl->sa_spill_tab) 229 230#define IS_SA_BONUSTYPE(a) \ 231 ((a == DMU_OT_SA) ? B_TRUE : B_FALSE) 232 233#define SA_BONUSTYPE_FROM_DB(db) \ 234 (dmu_get_bonustype((dmu_buf_t *)db)) 235 236#define SA_BLKPTR_SPACE (DN_MAX_BONUSLEN - sizeof (blkptr_t)) 237 238#define SA_LAYOUT_NUM(x, type) \ 239 ((!IS_SA_BONUSTYPE(type) ? 0 : (((IS_SA_BONUSTYPE(type)) && \ 240 ((SA_HDR_LAYOUT_NUM(x)) == 0)) ? 1 : SA_HDR_LAYOUT_NUM(x)))) 241 242 243#define SA_REGISTERED_LEN(sa, attr) sa->sa_attr_table[attr].sa_length 244 245#define SA_ATTR_LEN(sa, idx, attr, hdr) ((SA_REGISTERED_LEN(sa, attr) == 0) ?\ 246 hdr->sa_lengths[TOC_LEN_IDX(idx->sa_idx_tab[attr])] : \ 247 SA_REGISTERED_LEN(sa, attr)) 248 249#define SA_SET_HDR(hdr, num, size) \ 250 { \ 251 hdr->sa_magic = SA_MAGIC; \ 252 SA_HDR_LAYOUT_INFO_ENCODE(hdr->sa_layout_info, num, size); \ 253 } 254 255#define SA_ATTR_INFO(sa, idx, hdr, attr, bulk, type, hdl) \ 256 { \ 257 bulk.sa_size = SA_ATTR_LEN(sa, idx, attr, hdr); \ 258 bulk.sa_buftype = type; \ 259 bulk.sa_addr = \ 260 (void *)((uintptr_t)TOC_OFF(idx->sa_idx_tab[attr]) + \ 261 (uintptr_t)hdr); \ 262} 263 264#define SA_HDR_SIZE_MATCH_LAYOUT(hdr, tb) \ 265 (SA_HDR_SIZE(hdr) == (sizeof (sa_hdr_phys_t) + \ 266 (tb->lot_var_sizes > 1 ? P2ROUNDUP((tb->lot_var_sizes - 1) * \ 267 sizeof (uint16_t), 8) : 0))) 268 269int sa_add_impl(sa_handle_t *, sa_attr_type_t, 270 uint32_t, sa_data_locator_t, void *, dmu_tx_t *); 271 272void sa_register_update_callback_locked(objset_t *, sa_update_cb_t *); 273int sa_size_locked(sa_handle_t *, sa_attr_type_t, int *); 274 275void sa_default_locator(void **, uint32_t *, uint32_t, boolean_t, void *); 276int sa_attr_size(sa_os_t *, sa_idx_tab_t *, sa_attr_type_t, 277 uint16_t *, sa_hdr_phys_t *); 278 279#ifdef __cplusplus 280extern "C" { 281#endif 282 283#ifdef __cplusplus 284} 285#endif 286 287#endif /* _SYS_SA_IMPL_H */
| 186#define SA_HDR_LAYOUT_INFO_ENCODE(x, num, size) \ 187{ \ 188 BF32_SET_SB(x, 10, 6, 3, 0, size); \ 189 BF32_SET(x, 0, 10, num); \ 190} 191 192typedef enum sa_buf_type { 193 SA_BONUS = 1, 194 SA_SPILL = 2 195} sa_buf_type_t; 196 197typedef enum sa_data_op { 198 SA_LOOKUP, 199 SA_UPDATE, 200 SA_ADD, 201 SA_REPLACE, 202 SA_REMOVE 203} sa_data_op_t; 204 205/* 206 * Opaque handle used for most sa functions 207 * 208 * This needs to be kept as small as possible. 209 */ 210 211struct sa_handle { 212 kmutex_t sa_lock; 213 dmu_buf_t *sa_bonus; 214 dmu_buf_t *sa_spill; 215 objset_t *sa_os; 216 void *sa_userp; 217 sa_idx_tab_t *sa_bonus_tab; /* idx of bonus */ 218 sa_idx_tab_t *sa_spill_tab; /* only present if spill activated */ 219}; 220 221#define SA_GET_DB(hdl, type) \ 222 (dmu_buf_impl_t *)((type == SA_BONUS) ? hdl->sa_bonus : hdl->sa_spill) 223 224#define SA_GET_HDR(hdl, type) \ 225 ((sa_hdr_phys_t *)((dmu_buf_impl_t *)(SA_GET_DB(hdl, \ 226 type))->db.db_data)) 227 228#define SA_IDX_TAB_GET(hdl, type) \ 229 (type == SA_BONUS ? hdl->sa_bonus_tab : hdl->sa_spill_tab) 230 231#define IS_SA_BONUSTYPE(a) \ 232 ((a == DMU_OT_SA) ? B_TRUE : B_FALSE) 233 234#define SA_BONUSTYPE_FROM_DB(db) \ 235 (dmu_get_bonustype((dmu_buf_t *)db)) 236 237#define SA_BLKPTR_SPACE (DN_MAX_BONUSLEN - sizeof (blkptr_t)) 238 239#define SA_LAYOUT_NUM(x, type) \ 240 ((!IS_SA_BONUSTYPE(type) ? 0 : (((IS_SA_BONUSTYPE(type)) && \ 241 ((SA_HDR_LAYOUT_NUM(x)) == 0)) ? 1 : SA_HDR_LAYOUT_NUM(x)))) 242 243 244#define SA_REGISTERED_LEN(sa, attr) sa->sa_attr_table[attr].sa_length 245 246#define SA_ATTR_LEN(sa, idx, attr, hdr) ((SA_REGISTERED_LEN(sa, attr) == 0) ?\ 247 hdr->sa_lengths[TOC_LEN_IDX(idx->sa_idx_tab[attr])] : \ 248 SA_REGISTERED_LEN(sa, attr)) 249 250#define SA_SET_HDR(hdr, num, size) \ 251 { \ 252 hdr->sa_magic = SA_MAGIC; \ 253 SA_HDR_LAYOUT_INFO_ENCODE(hdr->sa_layout_info, num, size); \ 254 } 255 256#define SA_ATTR_INFO(sa, idx, hdr, attr, bulk, type, hdl) \ 257 { \ 258 bulk.sa_size = SA_ATTR_LEN(sa, idx, attr, hdr); \ 259 bulk.sa_buftype = type; \ 260 bulk.sa_addr = \ 261 (void *)((uintptr_t)TOC_OFF(idx->sa_idx_tab[attr]) + \ 262 (uintptr_t)hdr); \ 263} 264 265#define SA_HDR_SIZE_MATCH_LAYOUT(hdr, tb) \ 266 (SA_HDR_SIZE(hdr) == (sizeof (sa_hdr_phys_t) + \ 267 (tb->lot_var_sizes > 1 ? P2ROUNDUP((tb->lot_var_sizes - 1) * \ 268 sizeof (uint16_t), 8) : 0))) 269 270int sa_add_impl(sa_handle_t *, sa_attr_type_t, 271 uint32_t, sa_data_locator_t, void *, dmu_tx_t *); 272 273void sa_register_update_callback_locked(objset_t *, sa_update_cb_t *); 274int sa_size_locked(sa_handle_t *, sa_attr_type_t, int *); 275 276void sa_default_locator(void **, uint32_t *, uint32_t, boolean_t, void *); 277int sa_attr_size(sa_os_t *, sa_idx_tab_t *, sa_attr_type_t, 278 uint16_t *, sa_hdr_phys_t *); 279 280#ifdef __cplusplus 281extern "C" { 282#endif 283 284#ifdef __cplusplus 285} 286#endif 287 288#endif /* _SYS_SA_IMPL_H */
|