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 by Delphix. All rights reserved. 24 * Copyright (c) 2014 Integros [integros.com] 25 */ 26 27 /* Portions Copyright 2010 Robert Milkowski */ 28 29 #ifndef _SYS_ZIL_IMPL_H 30 #define _SYS_ZIL_IMPL_H 31 32 #include <sys/zil.h> 33 #include <sys/dmu_objset.h> 34 35 #ifdef __cplusplus 36 extern "C" { 37 #endif 38 39 /* 40 * Log write buffer. 41 */ 42 typedef struct lwb { 43 zilog_t *lwb_zilog; /* back pointer to log struct */ 44 blkptr_t lwb_blk; /* on disk address of this log blk */ 45 boolean_t lwb_slog; /* lwb_blk is on SLOG device */ 46 int lwb_nused; /* # used bytes in buffer */ 47 int lwb_sz; /* size of block and buffer */ 48 char *lwb_buf; /* log write buffer */ 49 zio_t *lwb_zio; /* zio for this buffer */ 50 dmu_tx_t *lwb_tx; /* tx for log block allocation */ 51 uint64_t lwb_max_txg; /* highest txg in this lwb */ 52 list_node_t lwb_node; /* zilog->zl_lwb_list linkage */ 53 } lwb_t; 54 55 /* 56 * Intent log transaction lists 57 */ 58 typedef struct itxs { 59 list_t i_sync_list; /* list of synchronous itxs */ 60 avl_tree_t i_async_tree; /* tree of foids for async itxs */ 61 } itxs_t; 62 63 typedef struct itxg { 64 kmutex_t itxg_lock; /* lock for this structure */ 65 uint64_t itxg_txg; /* txg for this chain */ 66 itxs_t *itxg_itxs; /* sync and async itxs */ 67 } itxg_t; 68 69 /* for async nodes we build up an AVL tree of lists of async itxs per file */ 70 typedef struct itx_async_node { 71 uint64_t ia_foid; /* file object id */ 72 list_t ia_list; /* list of async itxs for this foid */ 73 avl_node_t ia_node; /* AVL tree linkage */ 74 } itx_async_node_t; 75 76 /* 77 * Vdev flushing: during a zil_commit(), we build up an AVL tree of the vdevs 78 * we've touched so we know which ones need a write cache flush at the end. 79 */ 80 typedef struct zil_vdev_node { 81 uint64_t zv_vdev; /* vdev to be flushed */ 82 avl_node_t zv_node; /* AVL tree linkage */ 83 } zil_vdev_node_t; 84 85 #define ZIL_PREV_BLKS 16 86 87 /* 88 * Stable storage intent log management structure. One per dataset. 89 */ 90 struct zilog { 91 kmutex_t zl_lock; /* protects most zilog_t fields */ 92 struct dsl_pool *zl_dmu_pool; /* DSL pool */ 93 spa_t *zl_spa; /* handle for read/write log */ 94 const zil_header_t *zl_header; /* log header buffer */ 95 objset_t *zl_os; /* object set we're logging */ 96 zil_get_data_t *zl_get_data; /* callback to get object content */ 97 zio_t *zl_root_zio; /* log writer root zio */ 98 uint64_t zl_lr_seq; /* on-disk log record sequence number */ 99 uint64_t zl_commit_lr_seq; /* last committed on-disk lr seq */ 100 uint64_t zl_destroy_txg; /* txg of last zil_destroy() */ 101 uint64_t zl_replayed_seq[TXG_SIZE]; /* last replayed rec seq */ 102 uint64_t zl_replaying_seq; /* current replay seq number */ 103 uint32_t zl_suspend; /* log suspend count */ 104 kcondvar_t zl_cv_writer; /* log writer thread completion */ 105 kcondvar_t zl_cv_suspend; /* log suspend completion */ 106 uint8_t zl_suspending; /* log is currently suspending */ 107 uint8_t zl_keep_first; /* keep first log block in destroy */ 108 uint8_t zl_replay; /* replaying records while set */ 109 uint8_t zl_stop_sync; /* for debugging */ 110 uint8_t zl_writer; /* boolean: write setup in progress */ 111 uint8_t zl_logbias; /* latency or throughput */ 112 uint8_t zl_sync; /* synchronous or asynchronous */ 113 int zl_parse_error; /* last zil_parse() error */ 114 uint64_t zl_parse_blk_seq; /* highest blk seq on last parse */ 115 uint64_t zl_parse_lr_seq; /* highest lr seq on last parse */ 116 uint64_t zl_parse_blk_count; /* number of blocks parsed */ 117 uint64_t zl_parse_lr_count; /* number of log records parsed */ 118 uint64_t zl_next_batch; /* next batch number */ 119 uint64_t zl_com_batch; /* committed batch number */ 120 kcondvar_t zl_cv_batch[2]; /* batch condition variables */ 121 itxg_t zl_itxg[TXG_SIZE]; /* intent log txg chains */ 122 list_t zl_itx_commit_list; /* itx list to be committed */ 123 uint64_t zl_cur_used; /* current commit log size used */ 124 list_t zl_lwb_list; /* in-flight log write list */ 125 kmutex_t zl_vdev_lock; /* protects zl_vdev_tree */ 126 avl_tree_t zl_vdev_tree; /* vdevs to flush in zil_commit() */ 127 taskq_t *zl_clean_taskq; /* runs lwb and itx clean tasks */ 128 avl_tree_t zl_bp_tree; /* track bps during log parse */ 129 clock_t zl_replay_time; /* lbolt of when replay started */ 130 uint64_t zl_replay_blks; /* number of log blocks replayed */ 131 zil_header_t zl_old_header; /* debugging aid */ 132 uint_t zl_prev_blks[ZIL_PREV_BLKS]; /* size - sector rounded */ 133 uint_t zl_prev_rotor; /* rotor for zl_prev[] */ 134 txg_node_t zl_dirty_link; /* protected by dp_dirty_zilogs list */ 135 }; 136 137 typedef struct zil_bp_node { 138 dva_t zn_dva; 139 avl_node_t zn_node; 140 } zil_bp_node_t; 141 142 /* 143 * Maximum amount of write data that can be put into single log block. 144 */ 145 #define ZIL_MAX_LOG_DATA (SPA_OLD_MAXBLOCKSIZE - sizeof (zil_chain_t) - \ 146 sizeof (lr_write_t)) 147 148 /* 149 * Maximum amount of log space we agree to waste to reduce number of 150 * WR_NEED_COPY chunks to reduce zl_get_data() overhead (~12%). 151 */ 152 #define ZIL_MAX_WASTE_SPACE (ZIL_MAX_LOG_DATA / 8) 153 154 /* 155 * Maximum amount of write data for WR_COPIED. Fall back to WR_NEED_COPY 156 * as more space efficient if we can't fit at least two log records into 157 * maximum sized log block. 158 */ 159 #define ZIL_MAX_COPIED_DATA ((SPA_OLD_MAXBLOCKSIZE - \ 160 sizeof (zil_chain_t)) / 2 - sizeof (lr_write_t)) 161 162 #ifdef __cplusplus 163 } 164 #endif 165 166 #endif /* _SYS_ZIL_IMPL_H */ 167