10a586ceaSMark Shellenbaum /* 20a586ceaSMark Shellenbaum * CDDL HEADER START 30a586ceaSMark Shellenbaum * 40a586ceaSMark Shellenbaum * The contents of this file are subject to the terms of the 50a586ceaSMark Shellenbaum * Common Development and Distribution License (the "License"). 60a586ceaSMark Shellenbaum * You may not use this file except in compliance with the License. 70a586ceaSMark Shellenbaum * 80a586ceaSMark Shellenbaum * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 90a586ceaSMark Shellenbaum * or http://www.opensolaris.org/os/licensing. 100a586ceaSMark Shellenbaum * See the License for the specific language governing permissions 110a586ceaSMark Shellenbaum * and limitations under the License. 120a586ceaSMark Shellenbaum * 130a586ceaSMark Shellenbaum * When distributing Covered Code, include this CDDL HEADER in each 140a586ceaSMark Shellenbaum * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 150a586ceaSMark Shellenbaum * If applicable, add the following below this CDDL HEADER, with the 160a586ceaSMark Shellenbaum * fields enclosed by brackets "[]" replaced with your own identifying 170a586ceaSMark Shellenbaum * information: Portions Copyright [yyyy] [name of copyright owner] 180a586ceaSMark Shellenbaum * 190a586ceaSMark Shellenbaum * CDDL HEADER END 200a586ceaSMark Shellenbaum */ 210a586ceaSMark Shellenbaum /* 22744947dcSTom Erickson * Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved. 2369962b56SMatthew Ahrens * Copyright (c) 2013 by Delphix. All rights reserved. 24bc9014e6SJustin Gibbs * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved. 250a586ceaSMark Shellenbaum */ 260a586ceaSMark Shellenbaum 270a586ceaSMark Shellenbaum #ifndef _SYS_SA_IMPL_H 280a586ceaSMark Shellenbaum #define _SYS_SA_IMPL_H 290a586ceaSMark Shellenbaum 300a586ceaSMark Shellenbaum #include <sys/dmu.h> 310a586ceaSMark Shellenbaum #include <sys/refcount.h> 320a586ceaSMark Shellenbaum #include <sys/list.h> 330a586ceaSMark Shellenbaum 340a586ceaSMark Shellenbaum /* 350a586ceaSMark Shellenbaum * Array of known attributes and their 360a586ceaSMark Shellenbaum * various characteristics. 370a586ceaSMark Shellenbaum */ 380a586ceaSMark Shellenbaum typedef struct sa_attr_table { 390a586ceaSMark Shellenbaum sa_attr_type_t sa_attr; 400a586ceaSMark Shellenbaum uint8_t sa_registered; 410a586ceaSMark Shellenbaum uint16_t sa_length; 420a586ceaSMark Shellenbaum sa_bswap_type_t sa_byteswap; 430a586ceaSMark Shellenbaum char *sa_name; 440a586ceaSMark Shellenbaum } sa_attr_table_t; 450a586ceaSMark Shellenbaum 460a586ceaSMark Shellenbaum /* 470a586ceaSMark Shellenbaum * Zap attribute format for attribute registration 480a586ceaSMark Shellenbaum * 490a586ceaSMark Shellenbaum * 64 56 48 40 32 24 16 8 0 500a586ceaSMark Shellenbaum * +-------+-------+-------+-------+-------+-------+-------+-------+ 510a586ceaSMark Shellenbaum * | unused | len | bswap | attr num | 520a586ceaSMark Shellenbaum * +-------+-------+-------+-------+-------+-------+-------+-------+ 530a586ceaSMark Shellenbaum * 540a586ceaSMark Shellenbaum * Zap attribute format for layout information. 550a586ceaSMark Shellenbaum * 560a586ceaSMark Shellenbaum * layout information is stored as an array of attribute numbers 570a586ceaSMark Shellenbaum * The name of the attribute is the layout number (0, 1, 2, ...) 580a586ceaSMark Shellenbaum * 590a586ceaSMark Shellenbaum * 16 0 600a586ceaSMark Shellenbaum * +---- ---+ 610a586ceaSMark Shellenbaum * | attr # | 620a586ceaSMark Shellenbaum * +--------+ 630a586ceaSMark Shellenbaum * | attr # | 640a586ceaSMark Shellenbaum * +--- ----+ 650a586ceaSMark Shellenbaum * ...... 660a586ceaSMark Shellenbaum * 670a586ceaSMark Shellenbaum */ 680a586ceaSMark Shellenbaum 690a586ceaSMark Shellenbaum #define ATTR_BSWAP(x) BF32_GET(x, 16, 8) 700a586ceaSMark Shellenbaum #define ATTR_LENGTH(x) BF32_GET(x, 24, 16) 710a586ceaSMark Shellenbaum #define ATTR_NUM(x) BF32_GET(x, 0, 16) 720a586ceaSMark Shellenbaum #define ATTR_ENCODE(x, attr, length, bswap) \ 730a586ceaSMark Shellenbaum { \ 740a586ceaSMark Shellenbaum BF64_SET(x, 24, 16, length); \ 750a586ceaSMark Shellenbaum BF64_SET(x, 16, 8, bswap); \ 760a586ceaSMark Shellenbaum BF64_SET(x, 0, 16, attr); \ 770a586ceaSMark Shellenbaum } 780a586ceaSMark Shellenbaum 790a586ceaSMark Shellenbaum #define TOC_OFF(x) BF32_GET(x, 0, 23) 800a586ceaSMark Shellenbaum #define TOC_ATTR_PRESENT(x) BF32_GET(x, 31, 1) 810a586ceaSMark Shellenbaum #define TOC_LEN_IDX(x) BF32_GET(x, 24, 4) 820a586ceaSMark Shellenbaum #define TOC_ATTR_ENCODE(x, len_idx, offset) \ 830a586ceaSMark Shellenbaum { \ 840a586ceaSMark Shellenbaum BF32_SET(x, 31, 1, 1); \ 850a586ceaSMark Shellenbaum BF32_SET(x, 24, 7, len_idx); \ 860a586ceaSMark Shellenbaum BF32_SET(x, 0, 24, offset); \ 870a586ceaSMark Shellenbaum } 880a586ceaSMark Shellenbaum 890a586ceaSMark Shellenbaum #define SA_LAYOUTS "LAYOUTS" 900a586ceaSMark Shellenbaum #define SA_REGISTRY "REGISTRY" 910a586ceaSMark Shellenbaum 920a586ceaSMark Shellenbaum /* 930a586ceaSMark Shellenbaum * Each unique layout will have their own table 940a586ceaSMark Shellenbaum * sa_lot (layout_table) 950a586ceaSMark Shellenbaum */ 960a586ceaSMark Shellenbaum typedef struct sa_lot { 970a586ceaSMark Shellenbaum avl_node_t lot_num_node; 980a586ceaSMark Shellenbaum avl_node_t lot_hash_node; 990a586ceaSMark Shellenbaum uint64_t lot_num; 1000a586ceaSMark Shellenbaum uint64_t lot_hash; 1010a586ceaSMark Shellenbaum sa_attr_type_t *lot_attrs; /* array of attr #'s */ 1020a586ceaSMark Shellenbaum uint32_t lot_var_sizes; /* how many aren't fixed size */ 1030a586ceaSMark Shellenbaum uint32_t lot_attr_count; /* total attr count */ 104*54811da5SToomas Soome list_t lot_idx_tab; /* should be only a couple of entries */ 1050a586ceaSMark Shellenbaum int lot_instance; /* used with lot_hash to identify entry */ 1060a586ceaSMark Shellenbaum } sa_lot_t; 1070a586ceaSMark Shellenbaum 1080a586ceaSMark Shellenbaum /* index table of offsets */ 1090a586ceaSMark Shellenbaum typedef struct sa_idx_tab { 1100a586ceaSMark Shellenbaum list_node_t sa_next; 1110a586ceaSMark Shellenbaum sa_lot_t *sa_layout; 1120a586ceaSMark Shellenbaum uint16_t *sa_variable_lengths; 1130a586ceaSMark Shellenbaum refcount_t sa_refcount; 1140a586ceaSMark Shellenbaum uint32_t *sa_idx_tab; /* array of offsets */ 1150a586ceaSMark Shellenbaum } sa_idx_tab_t; 1160a586ceaSMark Shellenbaum 1170a586ceaSMark Shellenbaum /* 1180a586ceaSMark Shellenbaum * Since the offset/index information into the actual data 1190a586ceaSMark Shellenbaum * will usually be identical we can share that information with 1200a586ceaSMark Shellenbaum * all handles that have the exact same offsets. 1210a586ceaSMark Shellenbaum * 1220a586ceaSMark Shellenbaum * You would typically only have a large number of different table of 1230a586ceaSMark Shellenbaum * contents if you had a several variable sized attributes. 1240a586ceaSMark Shellenbaum * 1250a586ceaSMark Shellenbaum * Two AVL trees are used to track the attribute layout numbers. 1260a586ceaSMark Shellenbaum * one is keyed by number and will be consulted when a DMU_OT_SA 1270a586ceaSMark Shellenbaum * object is first read. The second tree is keyed by the hash signature 1280a586ceaSMark Shellenbaum * of the attributes and will be consulted when an attribute is added 1290a586ceaSMark Shellenbaum * to determine if we already have an instance of that layout. Both 1300a586ceaSMark Shellenbaum * of these tree's are interconnected. The only difference is that 1310a586ceaSMark Shellenbaum * when an entry is found in the "hash" tree the list of attributes will 1320a586ceaSMark Shellenbaum * need to be compared against the list of attributes you have in hand. 1330a586ceaSMark Shellenbaum * The assumption is that typically attributes will just be updated and 1340a586ceaSMark Shellenbaum * adding a completely new attribute is a very rare operation. 1350a586ceaSMark Shellenbaum */ 1360a586ceaSMark Shellenbaum struct sa_os { 137*54811da5SToomas Soome kmutex_t sa_lock; 1380a586ceaSMark Shellenbaum boolean_t sa_need_attr_registration; 1390a586ceaSMark Shellenbaum boolean_t sa_force_spill; 1400a586ceaSMark Shellenbaum uint64_t sa_master_obj; 1410a586ceaSMark Shellenbaum uint64_t sa_reg_attr_obj; 1420a586ceaSMark Shellenbaum uint64_t sa_layout_attr_obj; 1430a586ceaSMark Shellenbaum int sa_num_attrs; 1440a586ceaSMark Shellenbaum sa_attr_table_t *sa_attr_table; /* private attr table */ 1450a586ceaSMark Shellenbaum sa_update_cb_t *sa_update_cb; 1460a586ceaSMark Shellenbaum avl_tree_t sa_layout_num_tree; /* keyed by layout number */ 1470a586ceaSMark Shellenbaum avl_tree_t sa_layout_hash_tree; /* keyed by layout hash value */ 1480a586ceaSMark Shellenbaum int sa_user_table_sz; 1490a586ceaSMark Shellenbaum sa_attr_type_t *sa_user_table; /* user name->attr mapping table */ 1500a586ceaSMark Shellenbaum }; 1510a586ceaSMark Shellenbaum 1520a586ceaSMark Shellenbaum /* 1530a586ceaSMark Shellenbaum * header for all bonus and spill buffers. 154f7170741SWill Andrews * 1550a586ceaSMark Shellenbaum * The header has a fixed portion with a variable number 1560a586ceaSMark Shellenbaum * of "lengths" depending on the number of variable sized 15769962b56SMatthew Ahrens * attributes which are determined by the "layout number" 1580a586ceaSMark Shellenbaum */ 1590a586ceaSMark Shellenbaum 1600a586ceaSMark Shellenbaum #define SA_MAGIC 0x2F505A /* ZFS SA */ 1610a586ceaSMark Shellenbaum typedef struct sa_hdr_phys { 1620a586ceaSMark Shellenbaum uint32_t sa_magic; 16369962b56SMatthew Ahrens /* BEGIN CSTYLED */ 164f7170741SWill Andrews /* 165f7170741SWill Andrews * Encoded with hdrsize and layout number as follows: 166f7170741SWill Andrews * 16 10 0 167f7170741SWill Andrews * +--------+-------+ 168f7170741SWill Andrews * | hdrsz |layout | 169f7170741SWill Andrews * +--------+-------+ 170f7170741SWill Andrews * 171f7170741SWill Andrews * Bits 0-10 are the layout number 172f7170741SWill Andrews * Bits 11-16 are the size of the header. 173f7170741SWill Andrews * The hdrsize is the number * 8 174f7170741SWill Andrews * 175f7170741SWill Andrews * For example. 176f7170741SWill Andrews * hdrsz of 1 ==> 8 byte header 177f7170741SWill Andrews * 2 ==> 16 byte header 178f7170741SWill Andrews * 179f7170741SWill Andrews */ 18069962b56SMatthew Ahrens /* END CSTYLED */ 181f7170741SWill Andrews uint16_t sa_layout_info; 1820a586ceaSMark Shellenbaum uint16_t sa_lengths[1]; /* optional sizes for variable length attrs */ 1830a586ceaSMark Shellenbaum /* ... Data follows the lengths. */ 1840a586ceaSMark Shellenbaum } sa_hdr_phys_t; 1850a586ceaSMark Shellenbaum 1860a586ceaSMark Shellenbaum #define SA_HDR_LAYOUT_NUM(hdr) BF32_GET(hdr->sa_layout_info, 0, 10) 187e828a46dSMatthew Ahrens #define SA_HDR_SIZE(hdr) BF32_GET_SB(hdr->sa_layout_info, 10, 6, 3, 0) 1880a586ceaSMark Shellenbaum #define SA_HDR_LAYOUT_INFO_ENCODE(x, num, size) \ 1890a586ceaSMark Shellenbaum { \ 1900a586ceaSMark Shellenbaum BF32_SET_SB(x, 10, 6, 3, 0, size); \ 1910a586ceaSMark Shellenbaum BF32_SET(x, 0, 10, num); \ 1920a586ceaSMark Shellenbaum } 1930a586ceaSMark Shellenbaum 1940a586ceaSMark Shellenbaum typedef enum sa_buf_type { 1950a586ceaSMark Shellenbaum SA_BONUS = 1, 1960a586ceaSMark Shellenbaum SA_SPILL = 2 1970a586ceaSMark Shellenbaum } sa_buf_type_t; 1980a586ceaSMark Shellenbaum 1990a586ceaSMark Shellenbaum typedef enum sa_data_op { 2000a586ceaSMark Shellenbaum SA_LOOKUP, 2010a586ceaSMark Shellenbaum SA_UPDATE, 2020a586ceaSMark Shellenbaum SA_ADD, 2030a586ceaSMark Shellenbaum SA_REPLACE, 2040a586ceaSMark Shellenbaum SA_REMOVE 2050a586ceaSMark Shellenbaum } sa_data_op_t; 2060a586ceaSMark Shellenbaum 2070a586ceaSMark Shellenbaum /* 2080a586ceaSMark Shellenbaum * Opaque handle used for most sa functions 2090a586ceaSMark Shellenbaum * 2100a586ceaSMark Shellenbaum * This needs to be kept as small as possible. 2110a586ceaSMark Shellenbaum */ 2120a586ceaSMark Shellenbaum 2130a586ceaSMark Shellenbaum struct sa_handle { 214bc9014e6SJustin Gibbs dmu_buf_user_t sa_dbu; 2150a586ceaSMark Shellenbaum kmutex_t sa_lock; 2160a586ceaSMark Shellenbaum dmu_buf_t *sa_bonus; 2170a586ceaSMark Shellenbaum dmu_buf_t *sa_spill; 2180a586ceaSMark Shellenbaum objset_t *sa_os; 219bc9014e6SJustin Gibbs void *sa_userp; 2200a586ceaSMark Shellenbaum sa_idx_tab_t *sa_bonus_tab; /* idx of bonus */ 2210a586ceaSMark Shellenbaum sa_idx_tab_t *sa_spill_tab; /* only present if spill activated */ 2220a586ceaSMark Shellenbaum }; 2230a586ceaSMark Shellenbaum 2240a586ceaSMark Shellenbaum #define SA_GET_DB(hdl, type) \ 2250a586ceaSMark Shellenbaum (dmu_buf_impl_t *)((type == SA_BONUS) ? hdl->sa_bonus : hdl->sa_spill) 2260a586ceaSMark Shellenbaum 2270a586ceaSMark Shellenbaum #define SA_GET_HDR(hdl, type) \ 2280a586ceaSMark Shellenbaum ((sa_hdr_phys_t *)((dmu_buf_impl_t *)(SA_GET_DB(hdl, \ 2290a586ceaSMark Shellenbaum type))->db.db_data)) 2300a586ceaSMark Shellenbaum 2310a586ceaSMark Shellenbaum #define SA_IDX_TAB_GET(hdl, type) \ 2320a586ceaSMark Shellenbaum (type == SA_BONUS ? hdl->sa_bonus_tab : hdl->sa_spill_tab) 2330a586ceaSMark Shellenbaum 2340a586ceaSMark Shellenbaum #define IS_SA_BONUSTYPE(a) \ 2350a586ceaSMark Shellenbaum ((a == DMU_OT_SA) ? B_TRUE : B_FALSE) 2360a586ceaSMark Shellenbaum 2370a586ceaSMark Shellenbaum #define SA_BONUSTYPE_FROM_DB(db) \ 238744947dcSTom Erickson (dmu_get_bonustype((dmu_buf_t *)db)) 2390a586ceaSMark Shellenbaum 240*54811da5SToomas Soome #define SA_BLKPTR_SPACE (DN_OLD_MAX_BONUSLEN - sizeof (blkptr_t)) 2410a586ceaSMark Shellenbaum 2420a586ceaSMark Shellenbaum #define SA_LAYOUT_NUM(x, type) \ 2430a586ceaSMark Shellenbaum ((!IS_SA_BONUSTYPE(type) ? 0 : (((IS_SA_BONUSTYPE(type)) && \ 2440a586ceaSMark Shellenbaum ((SA_HDR_LAYOUT_NUM(x)) == 0)) ? 1 : SA_HDR_LAYOUT_NUM(x)))) 2450a586ceaSMark Shellenbaum 2460a586ceaSMark Shellenbaum 2470a586ceaSMark Shellenbaum #define SA_REGISTERED_LEN(sa, attr) sa->sa_attr_table[attr].sa_length 2480a586ceaSMark Shellenbaum 2490a586ceaSMark Shellenbaum #define SA_ATTR_LEN(sa, idx, attr, hdr) ((SA_REGISTERED_LEN(sa, attr) == 0) ?\ 2500a586ceaSMark Shellenbaum hdr->sa_lengths[TOC_LEN_IDX(idx->sa_idx_tab[attr])] : \ 2510a586ceaSMark Shellenbaum SA_REGISTERED_LEN(sa, attr)) 2520a586ceaSMark Shellenbaum 2530a586ceaSMark Shellenbaum #define SA_SET_HDR(hdr, num, size) \ 2540a586ceaSMark Shellenbaum { \ 2550a586ceaSMark Shellenbaum hdr->sa_magic = SA_MAGIC; \ 2560a586ceaSMark Shellenbaum SA_HDR_LAYOUT_INFO_ENCODE(hdr->sa_layout_info, num, size); \ 2570a586ceaSMark Shellenbaum } 2580a586ceaSMark Shellenbaum 2590a586ceaSMark Shellenbaum #define SA_ATTR_INFO(sa, idx, hdr, attr, bulk, type, hdl) \ 2600a586ceaSMark Shellenbaum { \ 2610a586ceaSMark Shellenbaum bulk.sa_size = SA_ATTR_LEN(sa, idx, attr, hdr); \ 2620a586ceaSMark Shellenbaum bulk.sa_buftype = type; \ 2630a586ceaSMark Shellenbaum bulk.sa_addr = \ 2640a586ceaSMark Shellenbaum (void *)((uintptr_t)TOC_OFF(idx->sa_idx_tab[attr]) + \ 2650a586ceaSMark Shellenbaum (uintptr_t)hdr); \ 2660a586ceaSMark Shellenbaum } 2670a586ceaSMark Shellenbaum 2680a586ceaSMark Shellenbaum #define SA_HDR_SIZE_MATCH_LAYOUT(hdr, tb) \ 2690a586ceaSMark Shellenbaum (SA_HDR_SIZE(hdr) == (sizeof (sa_hdr_phys_t) + \ 2700a586ceaSMark Shellenbaum (tb->lot_var_sizes > 1 ? P2ROUNDUP((tb->lot_var_sizes - 1) * \ 2710a586ceaSMark Shellenbaum sizeof (uint16_t), 8) : 0))) 2720a586ceaSMark Shellenbaum 2730a586ceaSMark Shellenbaum int sa_add_impl(sa_handle_t *, sa_attr_type_t, 2740a586ceaSMark Shellenbaum uint32_t, sa_data_locator_t, void *, dmu_tx_t *); 2750a586ceaSMark Shellenbaum 2760a586ceaSMark Shellenbaum void sa_register_update_callback_locked(objset_t *, sa_update_cb_t *); 2770a586ceaSMark Shellenbaum int sa_size_locked(sa_handle_t *, sa_attr_type_t, int *); 2780a586ceaSMark Shellenbaum 2790a586ceaSMark Shellenbaum void sa_default_locator(void **, uint32_t *, uint32_t, boolean_t, void *); 2800a586ceaSMark Shellenbaum int sa_attr_size(sa_os_t *, sa_idx_tab_t *, sa_attr_type_t, 2810a586ceaSMark Shellenbaum uint16_t *, sa_hdr_phys_t *); 2820a586ceaSMark Shellenbaum 2830a586ceaSMark Shellenbaum #ifdef __cplusplus 2840a586ceaSMark Shellenbaum extern "C" { 2850a586ceaSMark Shellenbaum #endif 2860a586ceaSMark Shellenbaum 2870a586ceaSMark Shellenbaum #ifdef __cplusplus 2880a586ceaSMark Shellenbaum } 2890a586ceaSMark Shellenbaum #endif 2900a586ceaSMark Shellenbaum 2910a586ceaSMark Shellenbaum #endif /* _SYS_SA_IMPL_H */ 292