zil_impl.h revision 276081 
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 */
25
26/* Portions Copyright 2010 Robert Milkowski */
27
28#ifndef	_SYS_ZIL_IMPL_H
29#define	_SYS_ZIL_IMPL_H
30
31#include <sys/zil.h>
32#include <sys/dmu_objset.h>
33
34#ifdef	__cplusplus
35extern "C" {
36#endif
37
38/*
39 * Log write buffer.
40 */
41typedef struct lwb {
42	zilog_t		*lwb_zilog;	/* back pointer to log struct */
43	blkptr_t	lwb_blk;	/* on disk address of this log blk */
44	int		lwb_nused;	/* # used bytes in buffer */
45	int		lwb_sz;		/* size of block and buffer */
46	char		*lwb_buf;	/* log write buffer */
47	zio_t		*lwb_zio;	/* zio for this buffer */
48	dmu_tx_t	*lwb_tx;	/* tx for log block allocation */
49	uint64_t	lwb_max_txg;	/* highest txg in this lwb */
50	list_node_t	lwb_node;	/* zilog->zl_lwb_list linkage */
51} lwb_t;
52
53/*
54 * Intent log transaction lists
55 */
56typedef struct itxs {
57	list_t		i_sync_list;	/* list of synchronous itxs */
58	avl_tree_t	i_async_tree;	/* tree of foids for async itxs */
59} itxs_t;
60
61typedef struct itxg {
62	kmutex_t	itxg_lock;	/* lock for this structure */
63	uint64_t	itxg_txg;	/* txg for this chain */
64	uint64_t	itxg_sod;	/* total size on disk for this txg */
65	itxs_t		*itxg_itxs;	/* sync and async itxs */
66} itxg_t;
67
68/* for async nodes we build up an AVL tree of lists of async itxs per file */
69typedef struct itx_async_node {
70	uint64_t	ia_foid;	/* file object id */
71	list_t		ia_list;	/* list of async itxs for this foid */
72	avl_node_t	ia_node;	/* AVL tree linkage */
73} itx_async_node_t;
74
75/*
76 * Vdev flushing: during a zil_commit(), we build up an AVL tree of the vdevs
77 * we've touched so we know which ones need a write cache flush at the end.
78 */
79typedef struct zil_vdev_node {
80	uint64_t	zv_vdev;	/* vdev to be flushed */
81	avl_node_t	zv_node;	/* AVL tree linkage */
82} zil_vdev_node_t;
83
84#define	ZIL_PREV_BLKS 16
85
86/*
87 * Stable storage intent log management structure.  One per dataset.
88 */
89struct zilog {
90	kmutex_t	zl_lock;	/* protects most zilog_t fields */
91	struct dsl_pool	*zl_dmu_pool;	/* DSL pool */
92	spa_t		*zl_spa;	/* handle for read/write log */
93	const zil_header_t *zl_header;	/* log header buffer */
94	objset_t	*zl_os;		/* object set we're logging */
95	zil_get_data_t	*zl_get_data;	/* callback to get object content */
96	zio_t		*zl_root_zio;	/* log writer root zio */
97	uint64_t	zl_lr_seq;	/* on-disk log record sequence number */
98	uint64_t	zl_commit_lr_seq; /* last committed on-disk lr seq */
99	uint64_t	zl_destroy_txg;	/* txg of last zil_destroy() */
100	uint64_t	zl_replayed_seq[TXG_SIZE]; /* last replayed rec seq */
101	uint64_t	zl_replaying_seq; /* current replay seq number */
102	uint32_t	zl_suspend;	/* log suspend count */
103	kcondvar_t	zl_cv_writer;	/* log writer thread completion */
104	kcondvar_t	zl_cv_suspend;	/* log suspend completion */
105	uint8_t		zl_suspending;	/* log is currently suspending */
106	uint8_t		zl_keep_first;	/* keep first log block in destroy */
107	uint8_t		zl_replay;	/* replaying records while set */
108	uint8_t		zl_stop_sync;	/* for debugging */
109	uint8_t		zl_writer;	/* boolean: write setup in progress */
110	uint8_t		zl_logbias;	/* latency or throughput */
111	uint8_t		zl_sync;	/* synchronous or asynchronous */
112	int		zl_parse_error;	/* last zil_parse() error */
113	uint64_t	zl_parse_blk_seq; /* highest blk seq on last parse */
114	uint64_t	zl_parse_lr_seq; /* highest lr seq on last parse */
115	uint64_t	zl_parse_blk_count; /* number of blocks parsed */
116	uint64_t	zl_parse_lr_count; /* number of log records parsed */
117	uint64_t	zl_next_batch;	/* next batch number */
118	uint64_t	zl_com_batch;	/* committed batch number */
119	kcondvar_t	zl_cv_batch[2];	/* batch condition variables */
120	itxg_t		zl_itxg[TXG_SIZE]; /* intent log txg chains */
121	list_t		zl_itx_commit_list; /* itx list to be committed */
122	uint64_t	zl_itx_list_sz;	/* total size of records on list */
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
137typedef struct zil_bp_node {
138	dva_t		zn_dva;
139	avl_node_t	zn_node;
140} zil_bp_node_t;
141
142#define	ZIL_MAX_LOG_DATA (SPA_OLD_MAXBLOCKSIZE - sizeof (zil_chain_t) - \
143    sizeof (lr_write_t))
144
145#ifdef	__cplusplus
146}
147#endif
148
149#endif	/* _SYS_ZIL_IMPL_H */
150