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) 2013 by Delphix. All rights reserved.
24 */
25
26 #include <sys/zfs_context.h>
27 #include <sys/spa.h>
28 #include <sys/spa_impl.h>
29 #include <sys/vdev_file.h>
30 #include <sys/vdev_impl.h>
31 #include <sys/zio.h>
32 #include <sys/fs/zfs.h>
33 #include <sys/fm/fs/zfs.h>
34
35 /*
36 * Virtual device vector for files.
37 */
38
39 static void
40 vdev_file_hold(vdev_t *vd)
41 {
42 ASSERT(vd->vdev_path != NULL);
43 }
44
45 static void
46 vdev_file_rele(vdev_t *vd)
47 {
48 ASSERT(vd->vdev_path != NULL);
49 }
50
51 static int
52 vdev_file_open(vdev_t *vd, uint64_t *psize, uint64_t *max_psize,
53 uint64_t *ashift)
54 {
55 vdev_file_t *vf;
56 vnode_t *vp;
57 vattr_t vattr;
58 int error;
59
60 /*
61 * We must have a pathname, and it must be absolute.
62 */
63 if (vd->vdev_path == NULL || vd->vdev_path[0] != '/') {
64 vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
65 return (SET_ERROR(EINVAL));
66 }
67
68 /*
69 * Reopen the device if it's not currently open. Otherwise,
70 * just update the physical size of the device.
71 */
72 if (vd->vdev_tsd != NULL) {
73 ASSERT(vd->vdev_reopening);
74 vf = vd->vdev_tsd;
75 goto skip_open;
76 }
77
78 vf = vd->vdev_tsd = kmem_zalloc(sizeof (vdev_file_t), KM_SLEEP);
79
80 /*
81 * We always open the files from the root of the global zone, even if
82 * we're in a local zone. If the user has gotten to this point, the
83 * administrator has already decided that the pool should be available
84 * to local zone users, so the underlying devices should be as well.
85 */
86 ASSERT(vd->vdev_path != NULL && vd->vdev_path[0] == '/');
87 error = vn_openat(vd->vdev_path + 1, UIO_SYSSPACE,
88 spa_mode(vd->vdev_spa) | FOFFMAX, 0, &vp, 0, 0, rootdir, -1);
89
90 if (error) {
91 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
92 return (error);
93 }
94
95 vf->vf_vnode = vp;
96
97 #ifdef _KERNEL
98 /*
99 * Make sure it's a regular file.
100 */
101 if (vp->v_type != VREG) {
102 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
103 return (SET_ERROR(ENODEV));
104 }
105 #endif
106
107 skip_open:
108 /*
109 * Determine the physical size of the file.
110 */
111 vattr.va_mask = AT_SIZE;
112 error = VOP_GETATTR(vf->vf_vnode, &vattr, 0, kcred, NULL);
113 if (error) {
114 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
115 return (error);
116 }
117
118 *max_psize = *psize = vattr.va_size;
119 *ashift = SPA_MINBLOCKSHIFT;
120
121 return (0);
122 }
123
124 static void
125 vdev_file_close(vdev_t *vd)
126 {
127 vdev_file_t *vf = vd->vdev_tsd;
128
129 if (vd->vdev_reopening || vf == NULL)
130 return;
131
132 if (vf->vf_vnode != NULL) {
133 (void) VOP_PUTPAGE(vf->vf_vnode, 0, 0, B_INVAL, kcred, NULL);
134 (void) VOP_CLOSE(vf->vf_vnode, spa_mode(vd->vdev_spa), 1, 0,
135 kcred, NULL);
136 VN_RELE(vf->vf_vnode);
137 }
138
139 vd->vdev_delayed_close = B_FALSE;
140 kmem_free(vf, sizeof (vdev_file_t));
141 vd->vdev_tsd = NULL;
142 }
143
144 /*
145 * Implements the interrupt side for file vdev types. This routine will be
146 * called when the I/O completes allowing us to transfer the I/O to the
147 * interrupt taskqs. For consistency, the code structure mimics disk vdev
148 * types.
149 */
150 static void
151 vdev_file_io_intr(buf_t *bp)
152 {
153 vdev_buf_t *vb = (vdev_buf_t *)bp;
154 zio_t *zio = vb->vb_io;
155
156 zio->io_error = (geterror(bp) != 0 ? EIO : 0);
157 if (zio->io_error == 0 && bp->b_resid != 0)
158 zio->io_error = SET_ERROR(ENOSPC);
159
160 kmem_free(vb, sizeof (vdev_buf_t));
161 zio_interrupt(zio);
162 }
163
164 static void
165 vdev_file_io_strategy(void *arg)
166 {
167 buf_t *bp = arg;
168 vnode_t *vp = bp->b_private;
169 ssize_t resid;
170 int error;
171
172 error = vn_rdwr((bp->b_flags & B_READ) ? UIO_READ : UIO_WRITE,
173 vp, bp->b_un.b_addr, bp->b_bcount, ldbtob(bp->b_lblkno),
174 UIO_SYSSPACE, 0, RLIM64_INFINITY, kcred, &resid);
175
176 if (error == 0) {
177 bp->b_resid = resid;
178 biodone(bp);
179 } else {
180 bioerror(bp, error);
181 biodone(bp);
182 }
183 }
184
185 static int
186 vdev_file_io_start(zio_t *zio)
187 {
188 vdev_t *vd = zio->io_vd;
189 vdev_file_t *vf = vd->vdev_tsd;
190 vdev_buf_t *vb;
191 buf_t *bp;
192
193 if (zio->io_type == ZIO_TYPE_IOCTL) {
194 /* XXPOLICY */
195 if (!vdev_readable(vd)) {
196 zio->io_error = SET_ERROR(ENXIO);
197 return (ZIO_PIPELINE_CONTINUE);
198 }
199
200 switch (zio->io_cmd) {
201 case DKIOCFLUSHWRITECACHE:
202 zio->io_error = VOP_FSYNC(vf->vf_vnode, FSYNC | FDSYNC,
203 kcred, NULL);
204 break;
205 default:
206 zio->io_error = SET_ERROR(ENOTSUP);
207 }
208
209 return (ZIO_PIPELINE_CONTINUE);
210 }
211
212 vb = kmem_alloc(sizeof (vdev_buf_t), KM_SLEEP);
213
214 vb->vb_io = zio;
215 bp = &vb->vb_buf;
216
217 bioinit(bp);
218 bp->b_flags = (zio->io_type == ZIO_TYPE_READ ? B_READ : B_WRITE);
219 bp->b_bcount = zio->io_size;
220 bp->b_un.b_addr = zio->io_data;
221 bp->b_lblkno = lbtodb(zio->io_offset);
222 bp->b_bufsize = zio->io_size;
223 bp->b_private = vf->vf_vnode;
224 bp->b_iodone = (int (*)())vdev_file_io_intr;
225
226 VERIFY3U(taskq_dispatch(system_taskq, vdev_file_io_strategy, bp,
227 TQ_SLEEP), !=, 0);
228
229 return (ZIO_PIPELINE_STOP);
230 }
231
232 /* ARGSUSED */
233 static void
234 vdev_file_io_done(zio_t *zio)
235 {
236 }
237
238 vdev_ops_t vdev_file_ops = {
239 vdev_file_open,
240 vdev_file_close,
241 vdev_default_asize,
242 vdev_file_io_start,
243 vdev_file_io_done,
244 NULL,
245 vdev_file_hold,
246 vdev_file_rele,
247 VDEV_TYPE_FILE, /* name of this vdev type */
248 B_TRUE /* leaf vdev */
249 };
250
251 /*
252 * From userland we access disks just like files.
253 */
254 #ifndef _KERNEL
255
256 vdev_ops_t vdev_disk_ops = {
257 vdev_file_open,
258 vdev_file_close,
259 vdev_default_asize,
260 vdev_file_io_start,
261 vdev_file_io_done,
262 NULL,
263 vdev_file_hold,
264 vdev_file_rele,
265 VDEV_TYPE_DISK, /* name of this vdev type */
266 B_TRUE /* leaf vdev */
267 };
268
269 #endif