| @@ -1,659 +1,660 @@ | | | @@ -1,659 +1,660 @@ |
1 | /* | | 1 | /* |
2 | * CDDL HEADER START | | 2 | * CDDL HEADER START |
3 | * | | 3 | * |
4 | * The contents of this file are subject to the terms of the | | 4 | * The contents of this file are subject to the terms of the |
5 | * Common Development and Distribution License (the "License"). | | 5 | * Common Development and Distribution License (the "License"). |
6 | * You may not use this file except in compliance with the License. | | 6 | * You may not use this file except in compliance with the License. |
7 | * | | 7 | * |
8 | * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE | | 8 | * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE |
9 | * or http://www.opensolaris.org/os/licensing. | | 9 | * or http://www.opensolaris.org/os/licensing. |
10 | * See the License for the specific language governing permissions | | 10 | * See the License for the specific language governing permissions |
11 | * and limitations under the License. | | 11 | * and limitations under the License. |
12 | * | | 12 | * |
13 | * When distributing Covered Code, include this CDDL HEADER in each | | 13 | * When distributing Covered Code, include this CDDL HEADER in each |
14 | * file and include the License file at usr/src/OPENSOLARIS.LICENSE. | | 14 | * file and include the License file at usr/src/OPENSOLARIS.LICENSE. |
15 | * If applicable, add the following below this CDDL HEADER, with the | | 15 | * If applicable, add the following below this CDDL HEADER, with the |
16 | * fields enclosed by brackets "[]" replaced with your own identifying | | 16 | * fields enclosed by brackets "[]" replaced with your own identifying |
17 | * information: Portions Copyright [yyyy] [name of copyright owner] | | 17 | * information: Portions Copyright [yyyy] [name of copyright owner] |
18 | * | | 18 | * |
19 | * CDDL HEADER END | | 19 | * CDDL HEADER END |
20 | */ | | 20 | */ |
21 | /* | | 21 | /* |
22 | * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. | | 22 | * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. |
23 | * Copyright (c) 2012, 2015 by Delphix. All rights reserved. | | 23 | * Copyright (c) 2012, 2015 by Delphix. All rights reserved. |
24 | * Copyright 2013 Nexenta Systems, Inc. All rights reserved. | | 24 | * Copyright 2013 Nexenta Systems, Inc. All rights reserved. |
25 | * Copyright (c) 2013 Joyent, Inc. All rights reserved. | | 25 | * Copyright (c) 2013 Joyent, Inc. All rights reserved. |
26 | */ | | 26 | */ |
27 | | | 27 | |
28 | #include <sys/zfs_context.h> | | 28 | #include <sys/zfs_context.h> |
29 | #include <sys/spa.h> | | 29 | #include <sys/spa.h> |
30 | #include <sys/refcount.h> | | 30 | #include <sys/refcount.h> |
31 | #include <sys/vdev_disk.h> | | 31 | #include <sys/vdev_disk.h> |
32 | #include <sys/vdev_impl.h> | | 32 | #include <sys/vdev_impl.h> |
33 | #include <sys/fs/zfs.h> | | 33 | #include <sys/fs/zfs.h> |
34 | #include <sys/zio.h> | | 34 | #include <sys/zio.h> |
35 | #include <sys/sunldi.h> | | 35 | #include <sys/sunldi.h> |
36 | #include <sys/fm/fs/zfs.h> | | 36 | #include <sys/fm/fs/zfs.h> |
37 | #include <sys/disk.h> | | 37 | #include <sys/disk.h> |
38 | #include <sys/disklabel.h> | | 38 | #include <sys/disklabel.h> |
39 | #include <sys/dkio.h> | | 39 | #include <sys/dkio.h> |
40 | #include <sys/workqueue.h> | | 40 | #include <sys/workqueue.h> |
41 | | | 41 | |
42 | #ifdef __NetBSD__ | | 42 | #ifdef __NetBSD__ |
43 | static int | | 43 | static int |
44 | geterror(struct buf *bp) | | 44 | geterror(struct buf *bp) |
45 | { | | 45 | { |
46 | | | 46 | |
47 | return (bp->b_error); | | 47 | return (bp->b_error); |
48 | } | | 48 | } |
49 | #endif | | 49 | #endif |
50 | | | 50 | |
51 | /* | | 51 | /* |
52 | * Virtual device vector for disks. | | 52 | * Virtual device vector for disks. |
53 | */ | | 53 | */ |
54 | | | 54 | |
55 | static void vdev_disk_io_intr(buf_t *); | | 55 | static void vdev_disk_io_intr(buf_t *); |
56 | | | 56 | |
57 | static void | | 57 | static void |
58 | vdev_disk_alloc(vdev_t *vd) | | 58 | vdev_disk_alloc(vdev_t *vd) |
59 | { | | 59 | { |
60 | vdev_disk_t *dvd; | | 60 | vdev_disk_t *dvd; |
61 | | | 61 | |
62 | dvd = vd->vdev_tsd = kmem_zalloc(sizeof (vdev_disk_t), KM_SLEEP); | | 62 | dvd = vd->vdev_tsd = kmem_zalloc(sizeof (vdev_disk_t), KM_SLEEP); |
63 | | | 63 | |
64 | #ifdef illumos | | 64 | #ifdef illumos |
65 | /* | | 65 | /* |
66 | * Create the LDI event callback list. | | 66 | * Create the LDI event callback list. |
67 | */ | | 67 | */ |
68 | list_create(&dvd->vd_ldi_cbs, sizeof (vdev_disk_ldi_cb_t), | | 68 | list_create(&dvd->vd_ldi_cbs, sizeof (vdev_disk_ldi_cb_t), |
69 | offsetof(vdev_disk_ldi_cb_t, lcb_next)); | | 69 | offsetof(vdev_disk_ldi_cb_t, lcb_next)); |
70 | #endif | | 70 | #endif |
71 | } | | 71 | } |
72 | | | 72 | |
73 | | | 73 | |
74 | static void | | 74 | static void |
75 | vdev_disk_free(vdev_t *vd) | | 75 | vdev_disk_free(vdev_t *vd) |
76 | { | | 76 | { |
77 | vdev_disk_t *dvd = vd->vdev_tsd; | | 77 | vdev_disk_t *dvd = vd->vdev_tsd; |
78 | #ifdef illumos | | 78 | #ifdef illumos |
79 | vdev_disk_ldi_cb_t *lcb; | | 79 | vdev_disk_ldi_cb_t *lcb; |
80 | #endif | | 80 | #endif |
81 | | | 81 | |
82 | if (dvd == NULL) | | 82 | if (dvd == NULL) |
83 | return; | | 83 | return; |
84 | | | 84 | |
85 | #ifdef illumos | | 85 | #ifdef illumos |
86 | /* | | 86 | /* |
87 | * We have already closed the LDI handle. Clean up the LDI event | | 87 | * We have already closed the LDI handle. Clean up the LDI event |
88 | * callbacks and free vd->vdev_tsd. | | 88 | * callbacks and free vd->vdev_tsd. |
89 | */ | | 89 | */ |
90 | while ((lcb = list_head(&dvd->vd_ldi_cbs)) != NULL) { | | 90 | while ((lcb = list_head(&dvd->vd_ldi_cbs)) != NULL) { |
91 | list_remove(&dvd->vd_ldi_cbs, lcb); | | 91 | list_remove(&dvd->vd_ldi_cbs, lcb); |
92 | (void) ldi_ev_remove_callbacks(lcb->lcb_id); | | 92 | (void) ldi_ev_remove_callbacks(lcb->lcb_id); |
93 | kmem_free(lcb, sizeof (vdev_disk_ldi_cb_t)); | | 93 | kmem_free(lcb, sizeof (vdev_disk_ldi_cb_t)); |
94 | } | | 94 | } |
95 | list_destroy(&dvd->vd_ldi_cbs); | | 95 | list_destroy(&dvd->vd_ldi_cbs); |
96 | #endif | | 96 | #endif |
97 | kmem_free(dvd, sizeof (vdev_disk_t)); | | 97 | kmem_free(dvd, sizeof (vdev_disk_t)); |
98 | vd->vdev_tsd = NULL; | | 98 | vd->vdev_tsd = NULL; |
99 | } | | 99 | } |
100 | | | 100 | |
101 | | | 101 | |
102 | /* | | 102 | /* |
103 | * It's not clear what these hold/rele functions are supposed to do. | | 103 | * It's not clear what these hold/rele functions are supposed to do. |
104 | */ | | 104 | */ |
105 | static void | | 105 | static void |
106 | vdev_disk_hold(vdev_t *vd) | | 106 | vdev_disk_hold(vdev_t *vd) |
107 | { | | 107 | { |
108 | | | 108 | |
109 | ASSERT(spa_config_held(vd->vdev_spa, SCL_STATE, RW_WRITER)); | | 109 | ASSERT(spa_config_held(vd->vdev_spa, SCL_STATE, RW_WRITER)); |
110 | | | 110 | |
111 | } | | 111 | } |
112 | | | 112 | |
113 | static void | | 113 | static void |
114 | vdev_disk_rele(vdev_t *vd) | | 114 | vdev_disk_rele(vdev_t *vd) |
115 | { | | 115 | { |
116 | | | 116 | |
117 | ASSERT(spa_config_held(vd->vdev_spa, SCL_STATE, RW_WRITER)); | | 117 | ASSERT(spa_config_held(vd->vdev_spa, SCL_STATE, RW_WRITER)); |
118 | | | 118 | |
119 | } | | 119 | } |
120 | | | 120 | |
121 | static void | | 121 | static void |
122 | vdev_disk_flush(struct work *work, void *cookie) | | 122 | vdev_disk_flush(struct work *work, void *cookie) |
123 | { | | 123 | { |
124 | vdev_disk_t *dvd; | | 124 | vdev_disk_t *dvd; |
125 | int error, cmd; | | 125 | int error, cmd; |
126 | buf_t *bp; | | 126 | buf_t *bp; |
127 | vnode_t *vp; | | 127 | vnode_t *vp; |
128 | | | 128 | |
129 | bp = (struct buf *)work; | | 129 | bp = (struct buf *)work; |
130 | vp = bp->b_vp; | | 130 | vp = bp->b_vp; |
131 | dvd = cookie; | | 131 | dvd = cookie; |
132 | | | 132 | |
133 | KASSERT(vp == dvd->vd_vp); | | 133 | KASSERT(vp == dvd->vd_vp); |
134 | | | 134 | |
135 | cmd = 1; | | 135 | cmd = 1; |
136 | vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); | | 136 | vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); |
137 | error = VOP_IOCTL(vp, DIOCCACHESYNC, &cmd, FREAD|FWRITE, kcred); | | 137 | error = VOP_IOCTL(vp, DIOCCACHESYNC, &cmd, FREAD|FWRITE, kcred); |
138 | VOP_UNLOCK(vp, 0); | | 138 | VOP_UNLOCK(vp, 0); |
139 | bp->b_error = error; | | 139 | bp->b_error = error; |
140 | vdev_disk_io_intr(bp); | | 140 | vdev_disk_io_intr(bp); |
141 | } | | 141 | } |
142 | | | 142 | |
143 | static int | | 143 | static int |
144 | vdev_disk_open(vdev_t *vd, uint64_t *psize, uint64_t *max_psize, | | 144 | vdev_disk_open(vdev_t *vd, uint64_t *psize, uint64_t *max_psize, |
145 | uint64_t *ashift, uint64_t *pashift) | | 145 | uint64_t *ashift, uint64_t *pashift) |
146 | { | | 146 | { |
147 | spa_t *spa = vd->vdev_spa; | | 147 | spa_t *spa = vd->vdev_spa; |
148 | vdev_disk_t *dvd; | | 148 | vdev_disk_t *dvd; |
149 | vnode_t *vp; | | 149 | vnode_t *vp; |
150 | struct dkwedge_info dkw; | | 150 | struct dkwedge_info dkw; |
151 | struct disk *pdk; | | 151 | struct disk *pdk; |
152 | int error, cmd; | | 152 | int error, cmd; |
153 | struct partinfo pinfo; | | 153 | struct partinfo pinfo; |
154 | | | 154 | |
155 | /* | | 155 | /* |
156 | * We must have a pathname, and it must be absolute. | | 156 | * We must have a pathname, and it must be absolute. |
157 | */ | | 157 | */ |
158 | if (vd->vdev_path == NULL || vd->vdev_path[0] != '/') { | | 158 | if (vd->vdev_path == NULL || vd->vdev_path[0] != '/') { |
159 | vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL; | | 159 | vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL; |
160 | return (SET_ERROR(EINVAL)); | | 160 | return (SET_ERROR(EINVAL)); |
161 | } | | 161 | } |
162 | | | 162 | |
163 | /* | | 163 | /* |
164 | * Reopen the device if it's not currently open. Otherwise, | | 164 | * Reopen the device if it's not currently open. Otherwise, |
165 | * just update the physical size of the device. | | 165 | * just update the physical size of the device. |
166 | */ | | 166 | */ |
167 | if (vd->vdev_tsd != NULL) { | | 167 | if (vd->vdev_tsd != NULL) { |
168 | ASSERT(vd->vdev_reopening); | | 168 | ASSERT(vd->vdev_reopening); |
169 | dvd = vd->vdev_tsd; | | 169 | dvd = vd->vdev_tsd; |
170 | vp = dvd->vd_vp; | | 170 | vp = dvd->vd_vp; |
171 | KASSERT(vp != NULL); | | 171 | KASSERT(vp != NULL); |
172 | goto skip_open; | | 172 | goto skip_open; |
173 | } | | 173 | } |
174 | | | 174 | |
175 | /* | | 175 | /* |
176 | * Create vd->vdev_tsd. | | 176 | * Create vd->vdev_tsd. |
177 | */ | | 177 | */ |
178 | vdev_disk_alloc(vd); | | 178 | vdev_disk_alloc(vd); |
179 | dvd = vd->vdev_tsd; | | 179 | dvd = vd->vdev_tsd; |
180 | | | 180 | |
181 | /* | | 181 | /* |
182 | * When opening a disk device, we want to preserve the user's original | | 182 | * When opening a disk device, we want to preserve the user's original |
183 | * intent. We always want to open the device by the path the user gave | | 183 | * intent. We always want to open the device by the path the user gave |
184 | * us, even if it is one of multiple paths to the save device. But we | | 184 | * us, even if it is one of multiple paths to the save device. But we |
185 | * also want to be able to survive disks being removed/recabled. | | 185 | * also want to be able to survive disks being removed/recabled. |
186 | * Therefore the sequence of opening devices is: | | 186 | * Therefore the sequence of opening devices is: |
187 | * | | 187 | * |
188 | * 1. Try opening the device by path. For legacy pools without the | | 188 | * 1. Try opening the device by path. For legacy pools without the |
189 | * 'whole_disk' property, attempt to fix the path by appending 's0'. | | 189 | * 'whole_disk' property, attempt to fix the path by appending 's0'. |
190 | * | | 190 | * |
191 | * 2. If the devid of the device matches the stored value, return | | 191 | * 2. If the devid of the device matches the stored value, return |
192 | * success. | | 192 | * success. |
193 | * | | 193 | * |
194 | * 3. Otherwise, the device may have moved. Try opening the device | | 194 | * 3. Otherwise, the device may have moved. Try opening the device |
195 | * by the devid instead. | | 195 | * by the devid instead. |
196 | */ | | 196 | */ |
197 | if (vd->vdev_devid != NULL) { | | 197 | if (vd->vdev_devid != NULL) { |
198 | /* XXXNETBSD wedges */ | | 198 | /* XXXNETBSD wedges */ |
199 | #ifdef illumos | | 199 | #ifdef illumos |
200 | if (ddi_devid_str_decode(vd->vdev_devid, &dvd->vd_devid, | | 200 | if (ddi_devid_str_decode(vd->vdev_devid, &dvd->vd_devid, |
201 | &dvd->vd_minor) != 0) { | | 201 | &dvd->vd_minor) != 0) { |
202 | vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL; | | 202 | vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL; |
203 | return (SET_ERROR(EINVAL)); | | 203 | return (SET_ERROR(EINVAL)); |
204 | } | | 204 | } |
205 | #endif | | 205 | #endif |
206 | } | | 206 | } |
207 | | | 207 | |
208 | error = EINVAL; /* presume failure */ | | 208 | error = EINVAL; /* presume failure */ |
209 | | | 209 | |
210 | error = vn_open(vd->vdev_path, UIO_SYSSPACE, FREAD|FWRITE, 0, | | 210 | error = vn_open(vd->vdev_path, UIO_SYSSPACE, FREAD|FWRITE, 0, |
211 | &vp, CRCREAT, 0); | | 211 | &vp, CRCREAT, 0); |
212 | if (error != 0) { | | 212 | if (error != 0) { |
213 | vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED; | | 213 | vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED; |
214 | return (SET_ERROR(error)); | | 214 | return (SET_ERROR(error)); |
215 | } | | 215 | } |
216 | if (vp->v_type != VBLK) { | | 216 | if (vp->v_type != VBLK) { |
217 | vrele(vp); | | 217 | vrele(vp); |
218 | vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED; | | 218 | vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED; |
219 | return (SET_ERROR(EINVAL)); | | 219 | return (SET_ERROR(EINVAL)); |
220 | } | | 220 | } |
221 | | | 221 | |
222 | /* XXXNETBSD Once tls-maxphys gets merged this block becomes: | | 222 | /* XXXNETBSD Once tls-maxphys gets merged this block becomes: |
223 | if (VOP_IOCTL(vp, DIOCGWEDGEINFO, &dkw, FREAD, NOCRED) == 0) | | 223 | if (VOP_IOCTL(vp, DIOCGWEDGEINFO, &dkw, FREAD, NOCRED) == 0) |
224 | pdk = disk_find(dkw.dkw_parent); | | 224 | pdk = disk_find(dkw.dkw_parent); |
225 | else | | 225 | else |
226 | pdk = disk_find_blk(vp->v_rdev); | | 226 | pdk = disk_find_blk(vp->v_rdev); |
227 | dvd->vd_maxphys = (pdk ? disk_maxphys(pdk) : MACHINE_MAXPHYS); | | 227 | dvd->vd_maxphys = (pdk ? disk_maxphys(pdk) : MACHINE_MAXPHYS); |
228 | */ | | 228 | */ |
229 | { | | 229 | { |
230 | struct buf buf = { .b_bcount = MAXPHYS }; | | 230 | struct buf buf = { .b_bcount = MAXPHYS }; |
231 | | | 231 | |
232 | if (VOP_IOCTL(vp, DIOCGWEDGEINFO, &dkw, FREAD, NOCRED) == 0) { | | 232 | if (VOP_IOCTL(vp, DIOCGWEDGEINFO, &dkw, FREAD, NOCRED) == 0) { |
233 | pdk = disk_find(dkw.dkw_parent); | | 233 | pdk = disk_find(dkw.dkw_parent); |
234 | } else { | | 234 | } else { |
235 | const char *dev_name; | | 235 | const char *dev_name; |
236 | | | 236 | |
237 | dev_name = devsw_blk2name(major(vp->v_rdev)); | | 237 | dev_name = devsw_blk2name(major(vp->v_rdev)); |
238 | if (dev_name) { | | 238 | if (dev_name) { |
239 | char disk_name[16]; | | 239 | char disk_name[16]; |
240 | | | 240 | |
241 | snprintf(disk_name, sizeof(disk_name), "%s%d", | | 241 | snprintf(disk_name, sizeof(disk_name), "%s%d", |
242 | dev_name, DISKUNIT(vp->v_rdev)); | | 242 | dev_name, DISKUNIT(vp->v_rdev)); |
243 | pdk = disk_find(disk_name); | | 243 | pdk = disk_find(disk_name); |
244 | } else { | | 244 | } else { |
245 | pdk = NULL; | | 245 | pdk = NULL; |
246 | } | | 246 | } |
| | | 247 | buf.b_dev = vp->v_rdev; |
247 | } | | 248 | } |
248 | if (pdk && pdk->dk_driver && pdk->dk_driver->d_minphys) | | 249 | if (pdk && pdk->dk_driver && pdk->dk_driver->d_minphys) |
249 | (*pdk->dk_driver->d_minphys)(&buf); | | 250 | (*pdk->dk_driver->d_minphys)(&buf); |
250 | dvd->vd_maxphys = buf.b_bcount; | | 251 | dvd->vd_maxphys = buf.b_bcount; |
251 | } | | 252 | } |
252 | | | 253 | |
253 | /* | | 254 | /* |
254 | * XXXNETBSD Compare the devid to the stored value. | | 255 | * XXXNETBSD Compare the devid to the stored value. |
255 | */ | | 256 | */ |
256 | | | 257 | |
257 | /* | | 258 | /* |
258 | * Create a workqueue to process cache-flushes concurrently. | | 259 | * Create a workqueue to process cache-flushes concurrently. |
259 | */ | | 260 | */ |
260 | error = workqueue_create(&dvd->vd_wq, "vdevsync", | | 261 | error = workqueue_create(&dvd->vd_wq, "vdevsync", |
261 | vdev_disk_flush, dvd, PRI_NONE, IPL_NONE, WQ_MPSAFE); | | 262 | vdev_disk_flush, dvd, PRI_NONE, IPL_NONE, WQ_MPSAFE); |
262 | if (error != 0) { | | 263 | if (error != 0) { |
263 | vrele(vp); | | 264 | vrele(vp); |
264 | return (SET_ERROR(error)); | | 265 | return (SET_ERROR(error)); |
265 | } | | 266 | } |
266 | | | 267 | |
267 | dvd->vd_vp = vp; | | 268 | dvd->vd_vp = vp; |
268 | | | 269 | |
269 | skip_open: | | 270 | skip_open: |
270 | /* | | 271 | /* |
271 | * Determine the actual size of the device. | | 272 | * Determine the actual size of the device. |
272 | * Try wedge info first as it supports larger disks. | | 273 | * Try wedge info first as it supports larger disks. |
273 | */ | | 274 | */ |
274 | error = VOP_IOCTL(vp, DIOCGWEDGEINFO, &dkw, FREAD, NOCRED); | | 275 | error = VOP_IOCTL(vp, DIOCGWEDGEINFO, &dkw, FREAD, NOCRED); |
275 | if (error == 0) { | | 276 | if (error == 0) { |
276 | pdk = disk_find(dkw.dkw_parent); | | 277 | pdk = disk_find(dkw.dkw_parent); |
277 | if (pdk) { | | 278 | if (pdk) { |
278 | pinfo.pi_secsize = (1 << pdk->dk_byteshift); | | 279 | pinfo.pi_secsize = (1 << pdk->dk_byteshift); |
279 | pinfo.pi_size = dkw.dkw_size; | | 280 | pinfo.pi_size = dkw.dkw_size; |
280 | pinfo.pi_offset = dkw.dkw_offset; | | 281 | pinfo.pi_offset = dkw.dkw_offset; |
281 | } else | | 282 | } else |
282 | error = ENODEV; | | 283 | error = ENODEV; |
283 | } | | 284 | } |
284 | if (error) | | 285 | if (error) |
285 | error = VOP_IOCTL(vp, DIOCGPARTINFO, &pinfo, FREAD, kcred); | | 286 | error = VOP_IOCTL(vp, DIOCGPARTINFO, &pinfo, FREAD, kcred); |
286 | if (error != 0) { | | 287 | if (error != 0) { |
287 | vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED; | | 288 | vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED; |
288 | return (SET_ERROR(error)); | | 289 | return (SET_ERROR(error)); |
289 | } | | 290 | } |
290 | *psize = pinfo.pi_size * pinfo.pi_secsize; | | 291 | *psize = pinfo.pi_size * pinfo.pi_secsize; |
291 | *max_psize = *psize; | | 292 | *max_psize = *psize; |
292 | | | 293 | |
293 | *ashift = highbit(MAX(pinfo.pi_secsize, SPA_MINBLOCKSIZE)) - 1; | | 294 | *ashift = highbit(MAX(pinfo.pi_secsize, SPA_MINBLOCKSIZE)) - 1; |
294 | *pashift = *ashift; | | 295 | *pashift = *ashift; |
295 | vd->vdev_wholedisk = (pinfo.pi_offset == 0); /* XXXNETBSD */ | | 296 | vd->vdev_wholedisk = (pinfo.pi_offset == 0); /* XXXNETBSD */ |
296 | | | 297 | |
297 | /* | | 298 | /* |
298 | * Clear the nowritecache bit, so that on a vdev_reopen() we will | | 299 | * Clear the nowritecache bit, so that on a vdev_reopen() we will |
299 | * try again. | | 300 | * try again. |
300 | */ | | 301 | */ |
301 | vd->vdev_nowritecache = B_FALSE; | | 302 | vd->vdev_nowritecache = B_FALSE; |
302 | | | 303 | |
303 | return (0); | | 304 | return (0); |
304 | } | | 305 | } |
305 | | | 306 | |
306 | static void | | 307 | static void |
307 | vdev_disk_close(vdev_t *vd) | | 308 | vdev_disk_close(vdev_t *vd) |
308 | { | | 309 | { |
309 | vdev_disk_t *dvd = vd->vdev_tsd; | | 310 | vdev_disk_t *dvd = vd->vdev_tsd; |
310 | | | 311 | |
311 | if (vd->vdev_reopening || dvd == NULL) | | 312 | if (vd->vdev_reopening || dvd == NULL) |
312 | return; | | 313 | return; |
313 | | | 314 | |
314 | #ifdef illumos | | 315 | #ifdef illumos |
315 | if (dvd->vd_minor != NULL) { | | 316 | if (dvd->vd_minor != NULL) { |
316 | ddi_devid_str_free(dvd->vd_minor); | | 317 | ddi_devid_str_free(dvd->vd_minor); |
317 | dvd->vd_minor = NULL; | | 318 | dvd->vd_minor = NULL; |
318 | } | | 319 | } |
319 | | | 320 | |
320 | if (dvd->vd_devid != NULL) { | | 321 | if (dvd->vd_devid != NULL) { |
321 | ddi_devid_free(dvd->vd_devid); | | 322 | ddi_devid_free(dvd->vd_devid); |
322 | dvd->vd_devid = NULL; | | 323 | dvd->vd_devid = NULL; |
323 | } | | 324 | } |
324 | | | 325 | |
325 | if (dvd->vd_lh != NULL) { | | 326 | if (dvd->vd_lh != NULL) { |
326 | (void) ldi_close(dvd->vd_lh, spa_mode(vd->vdev_spa), kcred); | | 327 | (void) ldi_close(dvd->vd_lh, spa_mode(vd->vdev_spa), kcred); |
327 | dvd->vd_lh = NULL; | | 328 | dvd->vd_lh = NULL; |
328 | } | | 329 | } |
329 | #endif | | 330 | #endif |
330 | | | 331 | |
331 | #ifdef __NetBSD__ | | 332 | #ifdef __NetBSD__ |
332 | if (dvd->vd_vp != NULL) { | | 333 | if (dvd->vd_vp != NULL) { |
333 | vn_close(dvd->vd_vp, FREAD|FWRITE, kcred); | | 334 | vn_close(dvd->vd_vp, FREAD|FWRITE, kcred); |
334 | dvd->vd_vp = NULL; | | 335 | dvd->vd_vp = NULL; |
335 | } | | 336 | } |
336 | if (dvd->vd_wq != NULL) { | | 337 | if (dvd->vd_wq != NULL) { |
337 | workqueue_destroy(dvd->vd_wq); | | 338 | workqueue_destroy(dvd->vd_wq); |
338 | dvd->vd_wq = NULL; | | 339 | dvd->vd_wq = NULL; |
339 | } | | 340 | } |
340 | #endif | | 341 | #endif |
341 | | | 342 | |
342 | vd->vdev_delayed_close = B_FALSE; | | 343 | vd->vdev_delayed_close = B_FALSE; |
343 | #ifdef illumos | | 344 | #ifdef illumos |
344 | /* | | 345 | /* |
345 | * If we closed the LDI handle due to an offline notify from LDI, | | 346 | * If we closed the LDI handle due to an offline notify from LDI, |
346 | * don't free vd->vdev_tsd or unregister the callbacks here; | | 347 | * don't free vd->vdev_tsd or unregister the callbacks here; |
347 | * the offline finalize callback or a reopen will take care of it. | | 348 | * the offline finalize callback or a reopen will take care of it. |
348 | */ | | 349 | */ |
349 | if (dvd->vd_ldi_offline) | | 350 | if (dvd->vd_ldi_offline) |
350 | return; | | 351 | return; |
351 | #endif | | 352 | #endif |
352 | | | 353 | |
353 | vdev_disk_free(vd); | | 354 | vdev_disk_free(vd); |
354 | } | | 355 | } |
355 | | | 356 | |
356 | int | | 357 | int |
357 | vdev_disk_physio(vdev_t *vd, caddr_t data, | | 358 | vdev_disk_physio(vdev_t *vd, caddr_t data, |
358 | size_t size, uint64_t offset, int flags, boolean_t isdump) | | 359 | size_t size, uint64_t offset, int flags, boolean_t isdump) |
359 | { | | 360 | { |
360 | #ifdef illumos | | 361 | #ifdef illumos |
361 | vdev_disk_t *dvd = vd->vdev_tsd; | | 362 | vdev_disk_t *dvd = vd->vdev_tsd; |
362 | | | 363 | |
363 | /* | | 364 | /* |
364 | * If the vdev is closed, it's likely in the REMOVED or FAULTED state. | | 365 | * If the vdev is closed, it's likely in the REMOVED or FAULTED state. |
365 | * Nothing to be done here but return failure. | | 366 | * Nothing to be done here but return failure. |
366 | */ | | 367 | */ |
367 | if (dvd == NULL || (dvd->vd_ldi_offline && dvd->vd_lh == NULL)) | | 368 | if (dvd == NULL || (dvd->vd_ldi_offline && dvd->vd_lh == NULL)) |
368 | return (EIO); | | 369 | return (EIO); |
369 | | | 370 | |
370 | ASSERT(vd->vdev_ops == &vdev_disk_ops); | | 371 | ASSERT(vd->vdev_ops == &vdev_disk_ops); |
371 | | | 372 | |
372 | /* | | 373 | /* |
373 | * If in the context of an active crash dump, use the ldi_dump(9F) | | 374 | * If in the context of an active crash dump, use the ldi_dump(9F) |
374 | * call instead of ldi_strategy(9F) as usual. | | 375 | * call instead of ldi_strategy(9F) as usual. |
375 | */ | | 376 | */ |
376 | if (isdump) { | | 377 | if (isdump) { |
377 | ASSERT3P(dvd, !=, NULL); | | 378 | ASSERT3P(dvd, !=, NULL); |
378 | return (ldi_dump(dvd->vd_lh, data, lbtodb(offset), | | 379 | return (ldi_dump(dvd->vd_lh, data, lbtodb(offset), |
379 | lbtodb(size))); | | 380 | lbtodb(size))); |
380 | } | | 381 | } |
381 | | | 382 | |
382 | return (vdev_disk_ldi_physio(dvd->vd_lh, data, size, offset, flags)); | | 383 | return (vdev_disk_ldi_physio(dvd->vd_lh, data, size, offset, flags)); |
383 | #endif | | 384 | #endif |
384 | #ifdef __NetBSD__ | | 385 | #ifdef __NetBSD__ |
385 | return (EIO); | | 386 | return (EIO); |
386 | #endif | | 387 | #endif |
387 | } | | 388 | } |
388 | | | 389 | |
389 | static void | | 390 | static void |
390 | vdev_disk_io_intr(buf_t *bp) | | 391 | vdev_disk_io_intr(buf_t *bp) |
391 | { | | 392 | { |
392 | zio_t *zio = bp->b_private; | | 393 | zio_t *zio = bp->b_private; |
393 | | | 394 | |
394 | /* | | 395 | /* |
395 | * The rest of the zio stack only deals with EIO, ECKSUM, and ENXIO. | | 396 | * The rest of the zio stack only deals with EIO, ECKSUM, and ENXIO. |
396 | * Rather than teach the rest of the stack about other error | | 397 | * Rather than teach the rest of the stack about other error |
397 | * possibilities (EFAULT, etc), we normalize the error value here. | | 398 | * possibilities (EFAULT, etc), we normalize the error value here. |
398 | */ | | 399 | */ |
399 | zio->io_error = (geterror(bp) != 0 ? SET_ERROR(EIO) : 0); | | 400 | zio->io_error = (geterror(bp) != 0 ? SET_ERROR(EIO) : 0); |
400 | | | 401 | |
401 | if (zio->io_error == 0 && bp->b_resid != 0) | | 402 | if (zio->io_error == 0 && bp->b_resid != 0) |
402 | zio->io_error = SET_ERROR(EIO); | | 403 | zio->io_error = SET_ERROR(EIO); |
403 | | | 404 | |
404 | putiobuf(bp); | | 405 | putiobuf(bp); |
405 | zio_delay_interrupt(zio); | | 406 | zio_delay_interrupt(zio); |
406 | } | | 407 | } |
407 | | | 408 | |
408 | static void | | 409 | static void |
409 | vdev_disk_ioctl_free(zio_t *zio) | | 410 | vdev_disk_ioctl_free(zio_t *zio) |
410 | { | | 411 | { |
411 | kmem_free(zio->io_vsd, sizeof (struct dk_callback)); | | 412 | kmem_free(zio->io_vsd, sizeof (struct dk_callback)); |
412 | } | | 413 | } |
413 | | | 414 | |
414 | static const zio_vsd_ops_t vdev_disk_vsd_ops = { | | 415 | static const zio_vsd_ops_t vdev_disk_vsd_ops = { |
415 | vdev_disk_ioctl_free, | | 416 | vdev_disk_ioctl_free, |
416 | zio_vsd_default_cksum_report | | 417 | zio_vsd_default_cksum_report |
417 | }; | | 418 | }; |
418 | | | 419 | |
419 | static void | | 420 | static void |
420 | vdev_disk_ioctl_done(void *zio_arg, int error) | | 421 | vdev_disk_ioctl_done(void *zio_arg, int error) |
421 | { | | 422 | { |
422 | zio_t *zio = zio_arg; | | 423 | zio_t *zio = zio_arg; |
423 | | | 424 | |
424 | zio->io_error = error; | | 425 | zio->io_error = error; |
425 | | | 426 | |
426 | zio_interrupt(zio); | | 427 | zio_interrupt(zio); |
427 | } | | 428 | } |
428 | | | 429 | |
429 | static void | | 430 | static void |
430 | vdev_disk_io_start(zio_t *zio) | | 431 | vdev_disk_io_start(zio_t *zio) |
431 | { | | 432 | { |
432 | vdev_t *vd = zio->io_vd; | | 433 | vdev_t *vd = zio->io_vd; |
433 | vdev_disk_t *dvd = vd->vdev_tsd; | | 434 | vdev_disk_t *dvd = vd->vdev_tsd; |
434 | vnode_t *vp; | | 435 | vnode_t *vp; |
435 | buf_t *bp, *nbp; | | 436 | buf_t *bp, *nbp; |
436 | int error, size, off, resid; | | 437 | int error, size, off, resid; |
437 | | | 438 | |
438 | /* | | 439 | /* |
439 | * If the vdev is closed, it's likely in the REMOVED or FAULTED state. | | 440 | * If the vdev is closed, it's likely in the REMOVED or FAULTED state. |
440 | * Nothing to be done here but return failure. | | 441 | * Nothing to be done here but return failure. |
441 | */ | | 442 | */ |
442 | #ifdef illumos | | 443 | #ifdef illumos |
443 | if (dvd == NULL || (dvd->vd_ldi_offline && dvd->vd_lh == NULL)) { | | 444 | if (dvd == NULL || (dvd->vd_ldi_offline && dvd->vd_lh == NULL)) { |
444 | zio->io_error = SET_ERROR(ENXIO); | | 445 | zio->io_error = SET_ERROR(ENXIO); |
445 | zio_interrupt(zio); | | 446 | zio_interrupt(zio); |
446 | return; | | 447 | return; |
447 | } | | 448 | } |
448 | #endif | | 449 | #endif |
449 | #ifdef __NetBSD__ | | 450 | #ifdef __NetBSD__ |
450 | if (dvd == NULL) { | | 451 | if (dvd == NULL) { |
451 | zio->io_error = SET_ERROR(ENXIO); | | 452 | zio->io_error = SET_ERROR(ENXIO); |
452 | zio_interrupt(zio); | | 453 | zio_interrupt(zio); |
453 | return; | | 454 | return; |
454 | } | | 455 | } |
455 | ASSERT3U(dvd->vd_maxphys, >, 0); | | 456 | ASSERT3U(dvd->vd_maxphys, >, 0); |
456 | vp = dvd->vd_vp; | | 457 | vp = dvd->vd_vp; |
457 | #endif | | 458 | #endif |
458 | | | 459 | |
459 | if (zio->io_type == ZIO_TYPE_IOCTL) { | | 460 | if (zio->io_type == ZIO_TYPE_IOCTL) { |
460 | /* XXPOLICY */ | | 461 | /* XXPOLICY */ |
461 | if (!vdev_readable(vd)) { | | 462 | if (!vdev_readable(vd)) { |
462 | zio->io_error = SET_ERROR(ENXIO); | | 463 | zio->io_error = SET_ERROR(ENXIO); |
463 | zio_interrupt(zio); | | 464 | zio_interrupt(zio); |
464 | return; | | 465 | return; |
465 | } | | 466 | } |
466 | | | 467 | |
467 | switch (zio->io_cmd) { | | 468 | switch (zio->io_cmd) { |
468 | case DKIOCFLUSHWRITECACHE: | | 469 | case DKIOCFLUSHWRITECACHE: |
469 | | | 470 | |
470 | if (zfs_nocacheflush) | | 471 | if (zfs_nocacheflush) |
471 | break; | | 472 | break; |
472 | | | 473 | |
473 | if (vd->vdev_nowritecache) { | | 474 | if (vd->vdev_nowritecache) { |
474 | zio->io_error = ENOTSUP; | | 475 | zio->io_error = ENOTSUP; |
475 | break; | | 476 | break; |
476 | } | | 477 | } |
477 | | | 478 | |
478 | bp = getiobuf(vp, true); | | 479 | bp = getiobuf(vp, true); |
479 | bp->b_private = zio; | | 480 | bp->b_private = zio; |
480 | workqueue_enqueue(dvd->vd_wq, &bp->b_work, NULL); | | 481 | workqueue_enqueue(dvd->vd_wq, &bp->b_work, NULL); |
481 | return; | | 482 | return; |
482 | | | 483 | |
483 | default: | | 484 | default: |
484 | zio->io_error = SET_ERROR(ENOTSUP); | | 485 | zio->io_error = SET_ERROR(ENOTSUP); |
485 | break; | | 486 | break; |
486 | } | | 487 | } |
487 | | | 488 | |
488 | zio_execute(zio); | | 489 | zio_execute(zio); |
489 | return; | | 490 | return; |
490 | } | | 491 | } |
491 | | | 492 | |
492 | bp = getiobuf(vp, true); | | 493 | bp = getiobuf(vp, true); |
493 | bp->b_flags = (zio->io_type == ZIO_TYPE_READ ? B_READ : B_WRITE); | | 494 | bp->b_flags = (zio->io_type == ZIO_TYPE_READ ? B_READ : B_WRITE); |
494 | bp->b_cflags = BC_BUSY | BC_NOCACHE; | | 495 | bp->b_cflags = BC_BUSY | BC_NOCACHE; |
495 | bp->b_data = zio->io_data; | | 496 | bp->b_data = zio->io_data; |
496 | bp->b_blkno = btodb(zio->io_offset); | | 497 | bp->b_blkno = btodb(zio->io_offset); |
497 | bp->b_bcount = zio->io_size; | | 498 | bp->b_bcount = zio->io_size; |
498 | bp->b_resid = zio->io_size; | | 499 | bp->b_resid = zio->io_size; |
499 | bp->b_iodone = vdev_disk_io_intr; | | 500 | bp->b_iodone = vdev_disk_io_intr; |
500 | bp->b_private = zio; | | 501 | bp->b_private = zio; |
501 | | | 502 | |
502 | if (!(bp->b_flags & B_READ)) { | | 503 | if (!(bp->b_flags & B_READ)) { |
503 | mutex_enter(vp->v_interlock); | | 504 | mutex_enter(vp->v_interlock); |
504 | vp->v_numoutput++; | | 505 | vp->v_numoutput++; |
505 | mutex_exit(vp->v_interlock); | | 506 | mutex_exit(vp->v_interlock); |
506 | } | | 507 | } |
507 | | | 508 | |
508 | if (bp->b_bcount <= dvd->vd_maxphys) { | | 509 | if (bp->b_bcount <= dvd->vd_maxphys) { |
509 | /* We can do this I/O in one pass. */ | | 510 | /* We can do this I/O in one pass. */ |
510 | (void)VOP_STRATEGY(vp, bp); | | 511 | (void)VOP_STRATEGY(vp, bp); |
511 | } else { | | 512 | } else { |
512 | /* | | 513 | /* |
513 | * The I/O is larger than we can process in one pass. | | 514 | * The I/O is larger than we can process in one pass. |
514 | * Split it into smaller pieces. | | 515 | * Split it into smaller pieces. |
515 | */ | | 516 | */ |
516 | resid = zio->io_size; | | 517 | resid = zio->io_size; |
517 | off = 0; | | 518 | off = 0; |
518 | while (resid != 0) { | | 519 | while (resid != 0) { |
519 | size = uimin(resid, dvd->vd_maxphys); | | 520 | size = uimin(resid, dvd->vd_maxphys); |
520 | nbp = getiobuf(vp, true); | | 521 | nbp = getiobuf(vp, true); |
521 | nbp->b_blkno = btodb(zio->io_offset + off); | | 522 | nbp->b_blkno = btodb(zio->io_offset + off); |
522 | /* Below call increments v_numoutput. */ | | 523 | /* Below call increments v_numoutput. */ |
523 | nestiobuf_setup(bp, nbp, off, size); | | 524 | nestiobuf_setup(bp, nbp, off, size); |
524 | (void)VOP_STRATEGY(vp, nbp); | | 525 | (void)VOP_STRATEGY(vp, nbp); |
525 | resid -= size; | | 526 | resid -= size; |
526 | off += size; | | 527 | off += size; |
527 | } | | 528 | } |
528 | } | | 529 | } |
529 | } | | 530 | } |
530 | | | 531 | |
531 | static void | | 532 | static void |
532 | vdev_disk_io_done(zio_t *zio) | | 533 | vdev_disk_io_done(zio_t *zio) |
533 | { | | 534 | { |
534 | #ifdef illumos | | 535 | #ifdef illumos |
535 | vdev_t *vd = zio->io_vd; | | 536 | vdev_t *vd = zio->io_vd; |
536 | | | 537 | |
537 | /* | | 538 | /* |
538 | * If the device returned EIO, then attempt a DKIOCSTATE ioctl to see if | | 539 | * If the device returned EIO, then attempt a DKIOCSTATE ioctl to see if |
539 | * the device has been removed. If this is the case, then we trigger an | | 540 | * the device has been removed. If this is the case, then we trigger an |
540 | * asynchronous removal of the device. Otherwise, probe the device and | | 541 | * asynchronous removal of the device. Otherwise, probe the device and |
541 | * make sure it's still accessible. | | 542 | * make sure it's still accessible. |
542 | */ | | 543 | */ |
543 | if (zio->io_error == EIO && !vd->vdev_remove_wanted) { | | 544 | if (zio->io_error == EIO && !vd->vdev_remove_wanted) { |
544 | vdev_disk_t *dvd = vd->vdev_tsd; | | 545 | vdev_disk_t *dvd = vd->vdev_tsd; |
545 | int state = DKIO_NONE; | | 546 | int state = DKIO_NONE; |
546 | | | 547 | |
547 | if (ldi_ioctl(dvd->vd_lh, DKIOCSTATE, (intptr_t)&state, | | 548 | if (ldi_ioctl(dvd->vd_lh, DKIOCSTATE, (intptr_t)&state, |
548 | FKIOCTL, kcred, NULL) == 0 && state != DKIO_INSERTED) { | | 549 | FKIOCTL, kcred, NULL) == 0 && state != DKIO_INSERTED) { |
549 | /* | | 550 | /* |
550 | * We post the resource as soon as possible, instead of | | 551 | * We post the resource as soon as possible, instead of |
551 | * when the async removal actually happens, because the | | 552 | * when the async removal actually happens, because the |
552 | * DE is using this information to discard previous I/O | | 553 | * DE is using this information to discard previous I/O |
553 | * errors. | | 554 | * errors. |
554 | */ | | 555 | */ |
555 | zfs_post_remove(zio->io_spa, vd); | | 556 | zfs_post_remove(zio->io_spa, vd); |
556 | vd->vdev_remove_wanted = B_TRUE; | | 557 | vd->vdev_remove_wanted = B_TRUE; |
557 | spa_async_request(zio->io_spa, SPA_ASYNC_REMOVE); | | 558 | spa_async_request(zio->io_spa, SPA_ASYNC_REMOVE); |
558 | } else if (!vd->vdev_delayed_close) { | | 559 | } else if (!vd->vdev_delayed_close) { |
559 | vd->vdev_delayed_close = B_TRUE; | | 560 | vd->vdev_delayed_close = B_TRUE; |
560 | } | | 561 | } |
561 | } | | 562 | } |
562 | #endif | | 563 | #endif |
563 | } | | 564 | } |
564 | | | 565 | |
565 | vdev_ops_t vdev_disk_ops = { | | 566 | vdev_ops_t vdev_disk_ops = { |
566 | vdev_disk_open, | | 567 | vdev_disk_open, |
567 | vdev_disk_close, | | 568 | vdev_disk_close, |
568 | vdev_default_asize, | | 569 | vdev_default_asize, |
569 | vdev_disk_io_start, | | 570 | vdev_disk_io_start, |
570 | vdev_disk_io_done, | | 571 | vdev_disk_io_done, |
571 | NULL, | | 572 | NULL, |
572 | vdev_disk_hold, | | 573 | vdev_disk_hold, |
573 | vdev_disk_rele, | | 574 | vdev_disk_rele, |
574 | VDEV_TYPE_DISK, /* name of this vdev type */ | | 575 | VDEV_TYPE_DISK, /* name of this vdev type */ |
575 | B_TRUE /* leaf vdev */ | | 576 | B_TRUE /* leaf vdev */ |
576 | }; | | 577 | }; |
577 | | | 578 | |
578 | /* | | 579 | /* |
579 | * Given the root disk device devid or pathname, read the label from | | 580 | * Given the root disk device devid or pathname, read the label from |
580 | * the device, and construct a configuration nvlist. | | 581 | * the device, and construct a configuration nvlist. |
581 | */ | | 582 | */ |
582 | int | | 583 | int |
583 | vdev_disk_read_rootlabel(char *devpath, char *devid, nvlist_t **config) | | 584 | vdev_disk_read_rootlabel(char *devpath, char *devid, nvlist_t **config) |
584 | { | | 585 | { |
585 | #ifdef __NetBSD__ | | 586 | #ifdef __NetBSD__ |
586 | return (ENOTSUP); | | 587 | return (ENOTSUP); |
587 | #else | | 588 | #else |
588 | ldi_handle_t vd_lh; | | 589 | ldi_handle_t vd_lh; |
589 | vdev_label_t *label; | | 590 | vdev_label_t *label; |
590 | uint64_t s, size; | | 591 | uint64_t s, size; |
591 | int l; | | 592 | int l; |
592 | ddi_devid_t tmpdevid; | | 593 | ddi_devid_t tmpdevid; |
593 | int error = -1; | | 594 | int error = -1; |
594 | char *minor_name; | | 595 | char *minor_name; |
595 | | | 596 | |
596 | /* | | 597 | /* |
597 | * Read the device label and build the nvlist. | | 598 | * Read the device label and build the nvlist. |
598 | */ | | 599 | */ |
599 | if (devid != NULL && ddi_devid_str_decode(devid, &tmpdevid, | | 600 | if (devid != NULL && ddi_devid_str_decode(devid, &tmpdevid, |
600 | &minor_name) == 0) { | | 601 | &minor_name) == 0) { |
601 | error = ldi_open_by_devid(tmpdevid, minor_name, | | 602 | error = ldi_open_by_devid(tmpdevid, minor_name, |
602 | FREAD, kcred, &vd_lh, zfs_li); | | 603 | FREAD, kcred, &vd_lh, zfs_li); |
603 | ddi_devid_free(tmpdevid); | | 604 | ddi_devid_free(tmpdevid); |
604 | ddi_devid_str_free(minor_name); | | 605 | ddi_devid_str_free(minor_name); |
605 | } | | 606 | } |
606 | | | 607 | |
607 | if (error && (error = ldi_open_by_name(devpath, FREAD, kcred, &vd_lh, | | 608 | if (error && (error = ldi_open_by_name(devpath, FREAD, kcred, &vd_lh, |
608 | zfs_li))) | | 609 | zfs_li))) |
609 | return (error); | | 610 | return (error); |
610 | | | 611 | |
611 | if (ldi_get_size(vd_lh, &s)) { | | 612 | if (ldi_get_size(vd_lh, &s)) { |
612 | (void) ldi_close(vd_lh, FREAD, kcred); | | 613 | (void) ldi_close(vd_lh, FREAD, kcred); |
613 | return (SET_ERROR(EIO)); | | 614 | return (SET_ERROR(EIO)); |
614 | } | | 615 | } |
615 | | | 616 | |
616 | size = P2ALIGN_TYPED(s, sizeof (vdev_label_t), uint64_t); | | 617 | size = P2ALIGN_TYPED(s, sizeof (vdev_label_t), uint64_t); |
617 | label = kmem_alloc(sizeof (vdev_label_t), KM_SLEEP); | | 618 | label = kmem_alloc(sizeof (vdev_label_t), KM_SLEEP); |
618 | | | 619 | |
619 | *config = NULL; | | 620 | *config = NULL; |
620 | for (l = 0; l < VDEV_LABELS; l++) { | | 621 | for (l = 0; l < VDEV_LABELS; l++) { |
621 | uint64_t offset, state, txg = 0; | | 622 | uint64_t offset, state, txg = 0; |
622 | | | 623 | |
623 | /* read vdev label */ | | 624 | /* read vdev label */ |
624 | offset = vdev_label_offset(size, l, 0); | | 625 | offset = vdev_label_offset(size, l, 0); |
625 | if (vdev_disk_ldi_physio(vd_lh, (caddr_t)label, | | 626 | if (vdev_disk_ldi_physio(vd_lh, (caddr_t)label, |
626 | VDEV_SKIP_SIZE + VDEV_PHYS_SIZE, offset, B_READ) != 0) | | 627 | VDEV_SKIP_SIZE + VDEV_PHYS_SIZE, offset, B_READ) != 0) |
627 | continue; | | 628 | continue; |
628 | | | 629 | |
629 | if (nvlist_unpack(label->vl_vdev_phys.vp_nvlist, | | 630 | if (nvlist_unpack(label->vl_vdev_phys.vp_nvlist, |
630 | sizeof (label->vl_vdev_phys.vp_nvlist), config, 0) != 0) { | | 631 | sizeof (label->vl_vdev_phys.vp_nvlist), config, 0) != 0) { |
631 | *config = NULL; | | 632 | *config = NULL; |
632 | continue; | | 633 | continue; |
633 | } | | 634 | } |
634 | | | 635 | |
635 | if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_STATE, | | 636 | if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_STATE, |
636 | &state) != 0 || state >= POOL_STATE_DESTROYED) { | | 637 | &state) != 0 || state >= POOL_STATE_DESTROYED) { |
637 | nvlist_free(*config); | | 638 | nvlist_free(*config); |
638 | *config = NULL; | | 639 | *config = NULL; |
639 | continue; | | 640 | continue; |
640 | } | | 641 | } |
641 | | | 642 | |
642 | if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_TXG, | | 643 | if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_TXG, |
643 | &txg) != 0 || txg == 0) { | | 644 | &txg) != 0 || txg == 0) { |
644 | nvlist_free(*config); | | 645 | nvlist_free(*config); |
645 | *config = NULL; | | 646 | *config = NULL; |
646 | continue; | | 647 | continue; |
647 | } | | 648 | } |
648 | | | 649 | |
649 | break; | | 650 | break; |
650 | } | | 651 | } |
651 | | | 652 | |
652 | kmem_free(label, sizeof (vdev_label_t)); | | 653 | kmem_free(label, sizeof (vdev_label_t)); |
653 | (void) ldi_close(vd_lh, FREAD, kcred); | | 654 | (void) ldi_close(vd_lh, FREAD, kcred); |
654 | if (*config == NULL) | | 655 | if (*config == NULL) |
655 | error = SET_ERROR(EIDRM); | | 656 | error = SET_ERROR(EIDRM); |
656 | | | 657 | |
657 | return (error); | | 658 | return (error); |
658 | #endif | | 659 | #endif |
659 | } | | 660 | } |