1 // SPDX-License-Identifier: CDDL-1.0
2 /*
3 * CDDL HEADER START
4 *
5 * The contents of this file are subject to the terms of the
6 * Common Development and Distribution License (the "License").
7 * You may not use this file except in compliance with the License.
8 *
9 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
10 * or https://opensource.org/licenses/CDDL-1.0.
11 * See the License for the specific language governing permissions
12 * and limitations under the License.
13 *
14 * When distributing Covered Code, include this CDDL HEADER in each
15 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
16 * If applicable, add the following below this CDDL HEADER, with the
17 * fields enclosed by brackets "[]" replaced with your own identifying
18 * information: Portions Copyright [yyyy] [name of copyright owner]
19 *
20 * CDDL HEADER END
21 */
22 /*
23 * Copyright 2010 Sun Microsystems, Inc. All rights reserved.
24 * Use is subject to license terms.
25 */
26
27 /*
28 * Copyright (c) 2012, 2016 by Delphix. All rights reserved.
29 */
30
31 #include <sys/zfs_context.h>
32 #include <sys/spa.h>
33 #include <sys/vdev_impl.h>
34 #include <sys/zio.h>
35 #include <sys/fs/zfs.h>
36
37 /*
38 * Virtual device vector for the pool's root vdev.
39 */
40
41 static uint64_t
vdev_root_core_tvds(vdev_t * vd)42 vdev_root_core_tvds(vdev_t *vd)
43 {
44 uint64_t tvds = 0;
45
46 for (uint64_t c = 0; c < vd->vdev_children; c++) {
47 vdev_t *cvd = vd->vdev_child[c];
48
49 if (!cvd->vdev_ishole && !cvd->vdev_islog &&
50 cvd->vdev_ops != &vdev_indirect_ops) {
51 tvds++;
52 }
53 }
54
55 return (tvds);
56 }
57
58 /*
59 * We should be able to tolerate one failure with absolutely no damage
60 * to our metadata. Two failures will take out space maps, a bunch of
61 * indirect block trees, meta dnodes, dnodes, etc. Probably not a happy
62 * place to live. When we get smarter, we can liberalize this policy.
63 * e.g. If we haven't lost two consecutive top-level vdevs, then we are
64 * probably fine. Adding bean counters during alloc/free can make this
65 * future guesswork more accurate.
66 */
67 static boolean_t
too_many_errors(vdev_t * vd,uint64_t numerrors)68 too_many_errors(vdev_t *vd, uint64_t numerrors)
69 {
70 uint64_t tvds;
71
72 if (numerrors == 0)
73 return (B_FALSE);
74
75 tvds = vdev_root_core_tvds(vd);
76 ASSERT3U(numerrors, <=, tvds);
77
78 if (numerrors == tvds)
79 return (B_TRUE);
80
81 return (numerrors > spa_missing_tvds_allowed(vd->vdev_spa));
82 }
83
84 static int
vdev_root_open(vdev_t * vd,uint64_t * asize,uint64_t * max_asize,uint64_t * ashift,uint64_t * pshift)85 vdev_root_open(vdev_t *vd, uint64_t *asize, uint64_t *max_asize,
86 uint64_t *ashift, uint64_t *pshift)
87 {
88 spa_t *spa = vd->vdev_spa;
89 int lasterror = 0;
90 int numerrors = 0;
91
92 if (vd->vdev_children == 0) {
93 vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
94 return (SET_ERROR(EINVAL));
95 }
96
97 vdev_open_children(vd);
98
99 for (int c = 0; c < vd->vdev_children; c++) {
100 vdev_t *cvd = vd->vdev_child[c];
101
102 if (cvd->vdev_open_error && !cvd->vdev_islog &&
103 cvd->vdev_ops != &vdev_indirect_ops) {
104 lasterror = cvd->vdev_open_error;
105 numerrors++;
106 }
107 }
108
109 if (spa_load_state(spa) != SPA_LOAD_NONE)
110 spa_set_missing_tvds(spa, numerrors);
111
112 if (too_many_errors(vd, numerrors)) {
113 vd->vdev_stat.vs_aux = VDEV_AUX_NO_REPLICAS;
114 return (lasterror);
115 }
116
117 *asize = 0;
118 *max_asize = 0;
119 *ashift = 0;
120 *pshift = 0;
121
122 return (0);
123 }
124
125 static void
vdev_root_close(vdev_t * vd)126 vdev_root_close(vdev_t *vd)
127 {
128 for (int c = 0; c < vd->vdev_children; c++)
129 vdev_close(vd->vdev_child[c]);
130 }
131
132 static void
vdev_root_state_change(vdev_t * vd,int faulted,int degraded)133 vdev_root_state_change(vdev_t *vd, int faulted, int degraded)
134 {
135 if (too_many_errors(vd, faulted)) {
136 vdev_set_state(vd, B_FALSE, VDEV_STATE_CANT_OPEN,
137 VDEV_AUX_NO_REPLICAS);
138 } else if (degraded || faulted) {
139 vdev_set_state(vd, B_FALSE, VDEV_STATE_DEGRADED, VDEV_AUX_NONE);
140 } else {
141 vdev_set_state(vd, B_FALSE, VDEV_STATE_HEALTHY, VDEV_AUX_NONE);
142 }
143 }
144
145 vdev_ops_t vdev_root_ops = {
146 .vdev_op_init = NULL,
147 .vdev_op_fini = NULL,
148 .vdev_op_open = vdev_root_open,
149 .vdev_op_close = vdev_root_close,
150 .vdev_op_asize = vdev_default_asize,
151 .vdev_op_min_asize = vdev_default_min_asize,
152 .vdev_op_min_alloc = NULL,
153 .vdev_op_io_start = NULL, /* not applicable to the root */
154 .vdev_op_io_done = NULL, /* not applicable to the root */
155 .vdev_op_state_change = vdev_root_state_change,
156 .vdev_op_need_resilver = NULL,
157 .vdev_op_hold = NULL,
158 .vdev_op_rele = NULL,
159 .vdev_op_remap = NULL,
160 .vdev_op_xlate = NULL,
161 .vdev_op_rebuild_asize = NULL,
162 .vdev_op_metaslab_init = NULL,
163 .vdev_op_config_generate = NULL,
164 .vdev_op_nparity = NULL,
165 .vdev_op_ndisks = NULL,
166 .vdev_op_type = VDEV_TYPE_ROOT, /* name of this vdev type */
167 .vdev_op_leaf = B_FALSE /* not a leaf vdev */
168 };
169