Print this page
5056 ZFS deadlock on db_mtx and dn_holds
Reviewed by: Will Andrews <willa@spectralogic.com>
Reviewed by: Matt Ahrens <mahrens@delphix.com>
Reviewed by: George Wilson <george.wilson@delphix.com>
Approved by: Dan McDonald <danmcd@omniti.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 *
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.
|
↓ open down ↓ |
14 lines elided |
↑ open up ↑ |
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 23 * Copyright (c) 2011, 2014 by Delphix. All rights reserved.
24 24 * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
25 + * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
25 26 */
26 27
27 28 #ifndef _SYS_SPA_IMPL_H
28 29 #define _SYS_SPA_IMPL_H
29 30
30 31 #include <sys/spa.h>
31 32 #include <sys/vdev.h>
32 33 #include <sys/metaslab.h>
33 34 #include <sys/dmu.h>
34 35 #include <sys/dsl_pool.h>
35 36 #include <sys/uberblock_impl.h>
36 37 #include <sys/zfs_context.h>
37 38 #include <sys/avl.h>
38 39 #include <sys/refcount.h>
39 40 #include <sys/bplist.h>
40 41 #include <sys/bpobj.h>
41 42 #include <sys/zfeature.h>
42 43 #include <zfeature_common.h>
43 44
44 45 #ifdef __cplusplus
45 46 extern "C" {
46 47 #endif
47 48
48 49 typedef struct spa_error_entry {
49 50 zbookmark_phys_t se_bookmark;
50 51 char *se_name;
51 52 avl_node_t se_avl;
52 53 } spa_error_entry_t;
53 54
54 55 typedef struct spa_history_phys {
55 56 uint64_t sh_pool_create_len; /* ending offset of zpool create */
56 57 uint64_t sh_phys_max_off; /* physical EOF */
57 58 uint64_t sh_bof; /* logical BOF */
58 59 uint64_t sh_eof; /* logical EOF */
59 60 uint64_t sh_records_lost; /* num of records overwritten */
60 61 } spa_history_phys_t;
61 62
62 63 struct spa_aux_vdev {
63 64 uint64_t sav_object; /* MOS object for device list */
64 65 nvlist_t *sav_config; /* cached device config */
65 66 vdev_t **sav_vdevs; /* devices */
66 67 int sav_count; /* number devices */
67 68 boolean_t sav_sync; /* sync the device list */
68 69 nvlist_t **sav_pending; /* pending device additions */
69 70 uint_t sav_npending; /* # pending devices */
70 71 };
71 72
72 73 typedef struct spa_config_lock {
73 74 kmutex_t scl_lock;
74 75 kthread_t *scl_writer;
75 76 int scl_write_wanted;
76 77 kcondvar_t scl_cv;
77 78 refcount_t scl_count;
78 79 } spa_config_lock_t;
79 80
80 81 typedef struct spa_config_dirent {
81 82 list_node_t scd_link;
82 83 char *scd_path;
83 84 } spa_config_dirent_t;
84 85
85 86 typedef enum zio_taskq_type {
86 87 ZIO_TASKQ_ISSUE = 0,
87 88 ZIO_TASKQ_ISSUE_HIGH,
88 89 ZIO_TASKQ_INTERRUPT,
89 90 ZIO_TASKQ_INTERRUPT_HIGH,
90 91 ZIO_TASKQ_TYPES
91 92 } zio_taskq_type_t;
92 93
93 94 /*
94 95 * State machine for the zpool-poolname process. The states transitions
95 96 * are done as follows:
96 97 *
97 98 * From To Routine
98 99 * PROC_NONE -> PROC_CREATED spa_activate()
99 100 * PROC_CREATED -> PROC_ACTIVE spa_thread()
100 101 * PROC_ACTIVE -> PROC_DEACTIVATE spa_deactivate()
101 102 * PROC_DEACTIVATE -> PROC_GONE spa_thread()
102 103 * PROC_GONE -> PROC_NONE spa_deactivate()
103 104 */
104 105 typedef enum spa_proc_state {
105 106 SPA_PROC_NONE, /* spa_proc = &p0, no process created */
106 107 SPA_PROC_CREATED, /* spa_activate() has proc, is waiting */
107 108 SPA_PROC_ACTIVE, /* taskqs created, spa_proc set */
108 109 SPA_PROC_DEACTIVATE, /* spa_deactivate() requests process exit */
109 110 SPA_PROC_GONE /* spa_thread() is exiting, spa_proc = &p0 */
110 111 } spa_proc_state_t;
111 112
112 113 typedef struct spa_taskqs {
113 114 uint_t stqs_count;
114 115 taskq_t **stqs_taskq;
115 116 } spa_taskqs_t;
116 117
117 118 struct spa {
118 119 /*
119 120 * Fields protected by spa_namespace_lock.
120 121 */
121 122 char spa_name[MAXNAMELEN]; /* pool name */
122 123 char *spa_comment; /* comment */
123 124 avl_node_t spa_avl; /* node in spa_namespace_avl */
124 125 nvlist_t *spa_config; /* last synced config */
125 126 nvlist_t *spa_config_syncing; /* currently syncing config */
126 127 nvlist_t *spa_config_splitting; /* config for splitting */
127 128 nvlist_t *spa_load_info; /* info and errors from load */
128 129 uint64_t spa_config_txg; /* txg of last config change */
129 130 int spa_sync_pass; /* iterate-to-convergence */
130 131 pool_state_t spa_state; /* pool state */
131 132 int spa_inject_ref; /* injection references */
132 133 uint8_t spa_sync_on; /* sync threads are running */
133 134 spa_load_state_t spa_load_state; /* current load operation */
134 135 uint64_t spa_import_flags; /* import specific flags */
135 136 spa_taskqs_t spa_zio_taskq[ZIO_TYPES][ZIO_TASKQ_TYPES];
136 137 dsl_pool_t *spa_dsl_pool;
|
↓ open down ↓ |
102 lines elided |
↑ open up ↑ |
137 138 boolean_t spa_is_initializing; /* true while opening pool */
138 139 metaslab_class_t *spa_normal_class; /* normal data class */
139 140 metaslab_class_t *spa_log_class; /* intent log data class */
140 141 uint64_t spa_first_txg; /* first txg after spa_open() */
141 142 uint64_t spa_final_txg; /* txg of export/destroy */
142 143 uint64_t spa_freeze_txg; /* freeze pool at this txg */
143 144 uint64_t spa_load_max_txg; /* best initial ub_txg */
144 145 uint64_t spa_claim_max_txg; /* highest claimed birth txg */
145 146 timespec_t spa_loaded_ts; /* 1st successful open time */
146 147 objset_t *spa_meta_objset; /* copy of dp->dp_meta_objset */
148 + kmutex_t spa_evicting_os_lock; /* Evicting objset list lock */
149 + list_t spa_evicting_os_list; /* Objsets being evicted. */
150 + kcondvar_t spa_evicting_os_cv; /* Objset Eviction Completion */
147 151 txg_list_t spa_vdev_txg_list; /* per-txg dirty vdev list */
148 152 vdev_t *spa_root_vdev; /* top-level vdev container */
149 153 uint64_t spa_config_guid; /* config pool guid */
150 154 uint64_t spa_load_guid; /* spa_load initialized guid */
151 155 uint64_t spa_last_synced_guid; /* last synced guid */
152 156 list_t spa_config_dirty_list; /* vdevs with dirty config */
153 157 list_t spa_state_dirty_list; /* vdevs with dirty state */
154 158 spa_aux_vdev_t spa_spares; /* hot spares */
155 159 spa_aux_vdev_t spa_l2cache; /* L2ARC cache devices */
156 160 nvlist_t *spa_label_features; /* Features for reading MOS */
157 161 uint64_t spa_config_object; /* MOS object for pool config */
158 162 uint64_t spa_config_generation; /* config generation number */
159 163 uint64_t spa_syncing_txg; /* txg currently syncing */
160 164 bpobj_t spa_deferred_bpobj; /* deferred-free bplist */
161 165 bplist_t spa_free_bplist[TXG_SIZE]; /* bplist of stuff to free */
162 166 uberblock_t spa_ubsync; /* last synced uberblock */
163 167 uberblock_t spa_uberblock; /* current uberblock */
164 168 boolean_t spa_extreme_rewind; /* rewind past deferred frees */
165 169 uint64_t spa_last_io; /* lbolt of last non-scan I/O */
166 170 kmutex_t spa_scrub_lock; /* resilver/scrub lock */
167 171 uint64_t spa_scrub_inflight; /* in-flight scrub I/Os */
168 172 kcondvar_t spa_scrub_io_cv; /* scrub I/O completion */
169 173 uint8_t spa_scrub_active; /* active or suspended? */
170 174 uint8_t spa_scrub_type; /* type of scrub we're doing */
171 175 uint8_t spa_scrub_finished; /* indicator to rotate logs */
172 176 uint8_t spa_scrub_started; /* started since last boot */
173 177 uint8_t spa_scrub_reopen; /* scrub doing vdev_reopen */
174 178 uint64_t spa_scan_pass_start; /* start time per pass/reboot */
175 179 uint64_t spa_scan_pass_exam; /* examined bytes per pass */
176 180 kmutex_t spa_async_lock; /* protect async state */
177 181 kthread_t *spa_async_thread; /* thread doing async task */
178 182 int spa_async_suspended; /* async tasks suspended */
179 183 kcondvar_t spa_async_cv; /* wait for thread_exit() */
180 184 uint16_t spa_async_tasks; /* async task mask */
181 185 char *spa_root; /* alternate root directory */
182 186 uint64_t spa_ena; /* spa-wide ereport ENA */
183 187 int spa_last_open_failed; /* error if last open failed */
184 188 uint64_t spa_last_ubsync_txg; /* "best" uberblock txg */
185 189 uint64_t spa_last_ubsync_txg_ts; /* timestamp from that ub */
186 190 uint64_t spa_load_txg; /* ub txg that loaded */
187 191 uint64_t spa_load_txg_ts; /* timestamp from that ub */
188 192 uint64_t spa_load_meta_errors; /* verify metadata err count */
189 193 uint64_t spa_load_data_errors; /* verify data err count */
190 194 uint64_t spa_verify_min_txg; /* start txg of verify scrub */
191 195 kmutex_t spa_errlog_lock; /* error log lock */
192 196 uint64_t spa_errlog_last; /* last error log object */
193 197 uint64_t spa_errlog_scrub; /* scrub error log object */
194 198 kmutex_t spa_errlist_lock; /* error list/ereport lock */
195 199 avl_tree_t spa_errlist_last; /* last error list */
196 200 avl_tree_t spa_errlist_scrub; /* scrub error list */
197 201 uint64_t spa_deflate; /* should we deflate? */
198 202 uint64_t spa_history; /* history object */
199 203 kmutex_t spa_history_lock; /* history lock */
200 204 vdev_t *spa_pending_vdev; /* pending vdev additions */
201 205 kmutex_t spa_props_lock; /* property lock */
202 206 uint64_t spa_pool_props_object; /* object for properties */
203 207 uint64_t spa_bootfs; /* default boot filesystem */
204 208 uint64_t spa_failmode; /* failure mode for the pool */
205 209 uint64_t spa_delegation; /* delegation on/off */
206 210 list_t spa_config_list; /* previous cache file(s) */
207 211 /* per-CPU array of root of async I/O: */
208 212 zio_t **spa_async_zio_root;
209 213 zio_t *spa_suspend_zio_root; /* root of all suspended I/O */
210 214 kmutex_t spa_suspend_lock; /* protects suspend_zio_root */
211 215 kcondvar_t spa_suspend_cv; /* notification of resume */
212 216 uint8_t spa_suspended; /* pool is suspended */
213 217 uint8_t spa_claiming; /* pool is doing zil_claim() */
214 218 boolean_t spa_debug; /* debug enabled? */
215 219 boolean_t spa_is_root; /* pool is root */
216 220 int spa_minref; /* num refs when first opened */
217 221 int spa_mode; /* FREAD | FWRITE */
218 222 spa_log_state_t spa_log_state; /* log state */
219 223 uint64_t spa_autoexpand; /* lun expansion on/off */
220 224 ddt_t *spa_ddt[ZIO_CHECKSUM_FUNCTIONS]; /* in-core DDTs */
221 225 uint64_t spa_ddt_stat_object; /* DDT statistics */
222 226 uint64_t spa_dedup_ditto; /* dedup ditto threshold */
223 227 uint64_t spa_dedup_checksum; /* default dedup checksum */
224 228 uint64_t spa_dspace; /* dspace in normal class */
225 229 kmutex_t spa_vdev_top_lock; /* dueling offline/remove */
226 230 kmutex_t spa_proc_lock; /* protects spa_proc* */
227 231 kcondvar_t spa_proc_cv; /* spa_proc_state transitions */
228 232 spa_proc_state_t spa_proc_state; /* see definition */
229 233 struct proc *spa_proc; /* "zpool-poolname" process */
230 234 uint64_t spa_did; /* if procp != p0, did of t1 */
231 235 boolean_t spa_autoreplace; /* autoreplace set in open */
232 236 int spa_vdev_locks; /* locks grabbed */
233 237 uint64_t spa_creation_version; /* version at pool creation */
234 238 uint64_t spa_prev_software_version; /* See ub_software_version */
235 239 uint64_t spa_feat_for_write_obj; /* required to write to pool */
236 240 uint64_t spa_feat_for_read_obj; /* required to read from pool */
237 241 uint64_t spa_feat_desc_obj; /* Feature descriptions */
238 242 uint64_t spa_feat_enabled_txg_obj; /* Feature enabled txg */
239 243 /* cache feature refcounts */
240 244 uint64_t spa_feat_refcount_cache[SPA_FEATURES];
241 245 cyclic_id_t spa_deadman_cycid; /* cyclic id */
242 246 uint64_t spa_deadman_calls; /* number of deadman calls */
243 247 hrtime_t spa_sync_starttime; /* starting time fo spa_sync */
244 248 uint64_t spa_deadman_synctime; /* deadman expiration timer */
245 249
246 250 /*
247 251 * spa_iokstat_lock protects spa_iokstat and
248 252 * spa_queue_stats[].
249 253 */
250 254 kmutex_t spa_iokstat_lock;
251 255 struct kstat *spa_iokstat; /* kstat of io to this pool */
252 256 struct {
253 257 int spa_active;
254 258 int spa_queued;
255 259 } spa_queue_stats[ZIO_PRIORITY_NUM_QUEUEABLE];
256 260
257 261 hrtime_t spa_ccw_fail_time; /* Conf cache write fail time */
258 262
259 263 /*
260 264 * spa_refcount & spa_config_lock must be the last elements
261 265 * because refcount_t changes size based on compilation options.
262 266 * In order for the MDB module to function correctly, the other
263 267 * fields must remain in the same location.
264 268 */
265 269 spa_config_lock_t spa_config_lock[SCL_LOCKS]; /* config changes */
266 270 refcount_t spa_refcount; /* number of opens */
267 271 };
268 272
269 273 extern const char *spa_config_path;
270 274
271 275 extern void spa_taskq_dispatch_ent(spa_t *spa, zio_type_t t, zio_taskq_type_t q,
272 276 task_func_t *func, void *arg, uint_t flags, taskq_ent_t *ent);
273 277
274 278 #ifdef __cplusplus
275 279 }
276 280 #endif
277 281
278 282 #endif /* _SYS_SPA_IMPL_H */
|
↓ open down ↓ |
122 lines elided |
↑ open up ↑ |
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX