Print this page
4045 zfs write throttle & i/o scheduler performance work
Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Adam Leventhal <ahl@delphix.com>
Reviewed by: Christopher Siden <christopher.siden@delphix.com>
Split |
Close |
Expand all |
Collapse all |
--- old/usr/src/uts/common/fs/zfs/sys/spa_impl.h
+++ new/usr/src/uts/common/fs/zfs/sys/spa_impl.h
1 1 /*
2 2 * CDDL HEADER START
3 3 *
4 4 * The contents of this file are subject to the terms of the
5 5 * Common Development and Distribution License (the "License").
6 6 * You may not use this file except in compliance with the License.
7 7 *
8 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 9 * or http://www.opensolaris.org/os/licensing.
10 10 * See the License for the specific language governing permissions
11 11 * and limitations under the License.
12 12 *
↓ open down ↓ |
12 lines elided |
↑ open up ↑ |
13 13 * When distributing Covered Code, include this CDDL HEADER in each
14 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 15 * If applicable, add the following below this CDDL HEADER, with the
16 16 * fields enclosed by brackets "[]" replaced with your own identifying
17 17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 18 *
19 19 * CDDL HEADER END
20 20 */
21 21 /*
22 22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 - * Copyright (c) 2012 by Delphix. All rights reserved.
23 + * Copyright (c) 2013 by Delphix. All rights reserved.
24 24 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
25 25 */
26 26
27 27 #ifndef _SYS_SPA_IMPL_H
28 28 #define _SYS_SPA_IMPL_H
29 29
30 30 #include <sys/spa.h>
31 31 #include <sys/vdev.h>
32 32 #include <sys/metaslab.h>
33 33 #include <sys/dmu.h>
34 34 #include <sys/dsl_pool.h>
35 35 #include <sys/uberblock_impl.h>
36 36 #include <sys/zfs_context.h>
37 37 #include <sys/avl.h>
38 38 #include <sys/refcount.h>
39 39 #include <sys/bplist.h>
40 40 #include <sys/bpobj.h>
41 41
42 42 #ifdef __cplusplus
43 43 extern "C" {
44 44 #endif
45 45
46 46 typedef struct spa_error_entry {
47 47 zbookmark_t se_bookmark;
48 48 char *se_name;
49 49 avl_node_t se_avl;
50 50 } spa_error_entry_t;
51 51
52 52 typedef struct spa_history_phys {
53 53 uint64_t sh_pool_create_len; /* ending offset of zpool create */
54 54 uint64_t sh_phys_max_off; /* physical EOF */
55 55 uint64_t sh_bof; /* logical BOF */
56 56 uint64_t sh_eof; /* logical EOF */
57 57 uint64_t sh_records_lost; /* num of records overwritten */
58 58 } spa_history_phys_t;
59 59
60 60 struct spa_aux_vdev {
61 61 uint64_t sav_object; /* MOS object for device list */
62 62 nvlist_t *sav_config; /* cached device config */
63 63 vdev_t **sav_vdevs; /* devices */
64 64 int sav_count; /* number devices */
65 65 boolean_t sav_sync; /* sync the device list */
66 66 nvlist_t **sav_pending; /* pending device additions */
67 67 uint_t sav_npending; /* # pending devices */
68 68 };
69 69
70 70 typedef struct spa_config_lock {
71 71 kmutex_t scl_lock;
72 72 kthread_t *scl_writer;
73 73 int scl_write_wanted;
74 74 kcondvar_t scl_cv;
75 75 refcount_t scl_count;
76 76 } spa_config_lock_t;
77 77
78 78 typedef struct spa_config_dirent {
79 79 list_node_t scd_link;
80 80 char *scd_path;
81 81 } spa_config_dirent_t;
82 82
83 83 typedef enum zio_taskq_type {
84 84 ZIO_TASKQ_ISSUE = 0,
85 85 ZIO_TASKQ_ISSUE_HIGH,
86 86 ZIO_TASKQ_INTERRUPT,
87 87 ZIO_TASKQ_INTERRUPT_HIGH,
88 88 ZIO_TASKQ_TYPES
89 89 } zio_taskq_type_t;
90 90
91 91 /*
92 92 * State machine for the zpool-poolname process. The states transitions
93 93 * are done as follows:
94 94 *
95 95 * From To Routine
96 96 * PROC_NONE -> PROC_CREATED spa_activate()
97 97 * PROC_CREATED -> PROC_ACTIVE spa_thread()
98 98 * PROC_ACTIVE -> PROC_DEACTIVATE spa_deactivate()
99 99 * PROC_DEACTIVATE -> PROC_GONE spa_thread()
100 100 * PROC_GONE -> PROC_NONE spa_deactivate()
101 101 */
102 102 typedef enum spa_proc_state {
103 103 SPA_PROC_NONE, /* spa_proc = &p0, no process created */
104 104 SPA_PROC_CREATED, /* spa_activate() has proc, is waiting */
105 105 SPA_PROC_ACTIVE, /* taskqs created, spa_proc set */
106 106 SPA_PROC_DEACTIVATE, /* spa_deactivate() requests process exit */
107 107 SPA_PROC_GONE /* spa_thread() is exiting, spa_proc = &p0 */
108 108 } spa_proc_state_t;
109 109
110 110 typedef struct spa_taskqs {
111 111 uint_t stqs_count;
112 112 taskq_t **stqs_taskq;
113 113 } spa_taskqs_t;
114 114
115 115 struct spa {
116 116 /*
117 117 * Fields protected by spa_namespace_lock.
118 118 */
119 119 char spa_name[MAXNAMELEN]; /* pool name */
120 120 char *spa_comment; /* comment */
121 121 avl_node_t spa_avl; /* node in spa_namespace_avl */
122 122 nvlist_t *spa_config; /* last synced config */
123 123 nvlist_t *spa_config_syncing; /* currently syncing config */
124 124 nvlist_t *spa_config_splitting; /* config for splitting */
125 125 nvlist_t *spa_load_info; /* info and errors from load */
126 126 uint64_t spa_config_txg; /* txg of last config change */
127 127 int spa_sync_pass; /* iterate-to-convergence */
128 128 pool_state_t spa_state; /* pool state */
129 129 int spa_inject_ref; /* injection references */
130 130 uint8_t spa_sync_on; /* sync threads are running */
131 131 spa_load_state_t spa_load_state; /* current load operation */
132 132 uint64_t spa_import_flags; /* import specific flags */
133 133 spa_taskqs_t spa_zio_taskq[ZIO_TYPES][ZIO_TASKQ_TYPES];
134 134 dsl_pool_t *spa_dsl_pool;
135 135 boolean_t spa_is_initializing; /* true while opening pool */
136 136 metaslab_class_t *spa_normal_class; /* normal data class */
137 137 metaslab_class_t *spa_log_class; /* intent log data class */
138 138 uint64_t spa_first_txg; /* first txg after spa_open() */
139 139 uint64_t spa_final_txg; /* txg of export/destroy */
140 140 uint64_t spa_freeze_txg; /* freeze pool at this txg */
141 141 uint64_t spa_load_max_txg; /* best initial ub_txg */
142 142 uint64_t spa_claim_max_txg; /* highest claimed birth txg */
143 143 timespec_t spa_loaded_ts; /* 1st successful open time */
144 144 objset_t *spa_meta_objset; /* copy of dp->dp_meta_objset */
145 145 txg_list_t spa_vdev_txg_list; /* per-txg dirty vdev list */
146 146 vdev_t *spa_root_vdev; /* top-level vdev container */
147 147 uint64_t spa_config_guid; /* config pool guid */
148 148 uint64_t spa_load_guid; /* spa_load initialized guid */
149 149 uint64_t spa_last_synced_guid; /* last synced guid */
150 150 list_t spa_config_dirty_list; /* vdevs with dirty config */
151 151 list_t spa_state_dirty_list; /* vdevs with dirty state */
152 152 spa_aux_vdev_t spa_spares; /* hot spares */
153 153 spa_aux_vdev_t spa_l2cache; /* L2ARC cache devices */
154 154 nvlist_t *spa_label_features; /* Features for reading MOS */
155 155 uint64_t spa_config_object; /* MOS object for pool config */
156 156 uint64_t spa_config_generation; /* config generation number */
157 157 uint64_t spa_syncing_txg; /* txg currently syncing */
158 158 bpobj_t spa_deferred_bpobj; /* deferred-free bplist */
159 159 bplist_t spa_free_bplist[TXG_SIZE]; /* bplist of stuff to free */
160 160 uberblock_t spa_ubsync; /* last synced uberblock */
161 161 uberblock_t spa_uberblock; /* current uberblock */
162 162 boolean_t spa_extreme_rewind; /* rewind past deferred frees */
163 163 uint64_t spa_last_io; /* lbolt of last non-scan I/O */
164 164 kmutex_t spa_scrub_lock; /* resilver/scrub lock */
165 165 uint64_t spa_scrub_inflight; /* in-flight scrub I/Os */
166 166 kcondvar_t spa_scrub_io_cv; /* scrub I/O completion */
167 167 uint8_t spa_scrub_active; /* active or suspended? */
168 168 uint8_t spa_scrub_type; /* type of scrub we're doing */
169 169 uint8_t spa_scrub_finished; /* indicator to rotate logs */
170 170 uint8_t spa_scrub_started; /* started since last boot */
171 171 uint8_t spa_scrub_reopen; /* scrub doing vdev_reopen */
172 172 uint64_t spa_scan_pass_start; /* start time per pass/reboot */
173 173 uint64_t spa_scan_pass_exam; /* examined bytes per pass */
174 174 kmutex_t spa_async_lock; /* protect async state */
175 175 kthread_t *spa_async_thread; /* thread doing async task */
176 176 int spa_async_suspended; /* async tasks suspended */
177 177 kcondvar_t spa_async_cv; /* wait for thread_exit() */
178 178 uint16_t spa_async_tasks; /* async task mask */
179 179 char *spa_root; /* alternate root directory */
180 180 uint64_t spa_ena; /* spa-wide ereport ENA */
181 181 int spa_last_open_failed; /* error if last open failed */
182 182 uint64_t spa_last_ubsync_txg; /* "best" uberblock txg */
183 183 uint64_t spa_last_ubsync_txg_ts; /* timestamp from that ub */
184 184 uint64_t spa_load_txg; /* ub txg that loaded */
185 185 uint64_t spa_load_txg_ts; /* timestamp from that ub */
186 186 uint64_t spa_load_meta_errors; /* verify metadata err count */
187 187 uint64_t spa_load_data_errors; /* verify data err count */
188 188 uint64_t spa_verify_min_txg; /* start txg of verify scrub */
189 189 kmutex_t spa_errlog_lock; /* error log lock */
190 190 uint64_t spa_errlog_last; /* last error log object */
191 191 uint64_t spa_errlog_scrub; /* scrub error log object */
192 192 kmutex_t spa_errlist_lock; /* error list/ereport lock */
193 193 avl_tree_t spa_errlist_last; /* last error list */
194 194 avl_tree_t spa_errlist_scrub; /* scrub error list */
195 195 uint64_t spa_deflate; /* should we deflate? */
196 196 uint64_t spa_history; /* history object */
197 197 kmutex_t spa_history_lock; /* history lock */
198 198 vdev_t *spa_pending_vdev; /* pending vdev additions */
199 199 kmutex_t spa_props_lock; /* property lock */
200 200 uint64_t spa_pool_props_object; /* object for properties */
201 201 uint64_t spa_bootfs; /* default boot filesystem */
202 202 uint64_t spa_failmode; /* failure mode for the pool */
203 203 uint64_t spa_delegation; /* delegation on/off */
204 204 list_t spa_config_list; /* previous cache file(s) */
205 205 zio_t *spa_async_zio_root; /* root of all async I/O */
206 206 zio_t *spa_suspend_zio_root; /* root of all suspended I/O */
207 207 kmutex_t spa_suspend_lock; /* protects suspend_zio_root */
208 208 kcondvar_t spa_suspend_cv; /* notification of resume */
209 209 uint8_t spa_suspended; /* pool is suspended */
210 210 uint8_t spa_claiming; /* pool is doing zil_claim() */
211 211 boolean_t spa_debug; /* debug enabled? */
212 212 boolean_t spa_is_root; /* pool is root */
213 213 int spa_minref; /* num refs when first opened */
214 214 int spa_mode; /* FREAD | FWRITE */
215 215 spa_log_state_t spa_log_state; /* log state */
216 216 uint64_t spa_autoexpand; /* lun expansion on/off */
217 217 ddt_t *spa_ddt[ZIO_CHECKSUM_FUNCTIONS]; /* in-core DDTs */
218 218 uint64_t spa_ddt_stat_object; /* DDT statistics */
219 219 uint64_t spa_dedup_ditto; /* dedup ditto threshold */
220 220 uint64_t spa_dedup_checksum; /* default dedup checksum */
221 221 uint64_t spa_dspace; /* dspace in normal class */
222 222 kmutex_t spa_vdev_top_lock; /* dueling offline/remove */
223 223 kmutex_t spa_proc_lock; /* protects spa_proc* */
224 224 kcondvar_t spa_proc_cv; /* spa_proc_state transitions */
225 225 spa_proc_state_t spa_proc_state; /* see definition */
226 226 struct proc *spa_proc; /* "zpool-poolname" process */
↓ open down ↓ |
193 lines elided |
↑ open up ↑ |
227 227 uint64_t spa_did; /* if procp != p0, did of t1 */
228 228 boolean_t spa_autoreplace; /* autoreplace set in open */
229 229 int spa_vdev_locks; /* locks grabbed */
230 230 uint64_t spa_creation_version; /* version at pool creation */
231 231 uint64_t spa_prev_software_version; /* See ub_software_version */
232 232 uint64_t spa_feat_for_write_obj; /* required to write to pool */
233 233 uint64_t spa_feat_for_read_obj; /* required to read from pool */
234 234 uint64_t spa_feat_desc_obj; /* Feature descriptions */
235 235 cyclic_id_t spa_deadman_cycid; /* cyclic id */
236 236 uint64_t spa_deadman_calls; /* number of deadman calls */
237 - uint64_t spa_sync_starttime; /* starting time fo spa_sync */
237 + hrtime_t spa_sync_starttime; /* starting time fo spa_sync */
238 238 uint64_t spa_deadman_synctime; /* deadman expiration timer */
239 - kmutex_t spa_iokstat_lock; /* protects spa_iokstat_* */
239 +
240 + /*
241 + * spa_iokstat_lock protects spa_iokstat and
242 + * spa_queue_stats[].
243 + */
244 + kmutex_t spa_iokstat_lock;
240 245 struct kstat *spa_iokstat; /* kstat of io to this pool */
246 + struct {
247 + int spa_active;
248 + int spa_queued;
249 + } spa_queue_stats[ZIO_PRIORITY_NUM_QUEUEABLE];
250 +
241 251 hrtime_t spa_ccw_fail_time; /* Conf cache write fail time */
252 +
242 253 /*
243 254 * spa_refcount & spa_config_lock must be the last elements
244 255 * because refcount_t changes size based on compilation options.
245 256 * In order for the MDB module to function correctly, the other
246 257 * fields must remain in the same location.
247 258 */
248 259 spa_config_lock_t spa_config_lock[SCL_LOCKS]; /* config changes */
249 260 refcount_t spa_refcount; /* number of opens */
250 261 };
251 262
252 263 extern const char *spa_config_path;
253 264
254 265 extern void spa_taskq_dispatch_ent(spa_t *spa, zio_type_t t, zio_taskq_type_t q,
255 266 task_func_t *func, void *arg, uint_t flags, taskq_ent_t *ent);
256 267
257 268 #ifdef __cplusplus
258 269 }
259 270 #endif
260 271
261 272 #endif /* _SYS_SPA_IMPL_H */
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX