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) 2011, 2015 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 * Rotational optimizations only make sense on block devices
62 */
63 vd->vdev_nonrot = B_TRUE;
64
65 /*
66 * We must have a pathname, and it must be absolute.
67 */
68 if (vd->vdev_path == NULL || vd->vdev_path[0] != '/') {
69 vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
70 return (SET_ERROR(EINVAL));
71 }
72
73 /*
74 * Reopen the device if it's not currently open. Otherwise,
75 * just update the physical size of the device.
76 */
77 if (vd->vdev_tsd != NULL) {
78 ASSERT(vd->vdev_reopening);
79 vf = vd->vdev_tsd;
80 goto skip_open;
81 }
82
83 vf = vd->vdev_tsd = kmem_zalloc(sizeof (vdev_file_t), KM_SLEEP);
84
85 /*
86 * We always open the files from the root of the global zone, even if
87 * we're in a local zone. If the user has gotten to this point, the
88 * administrator has already decided that the pool should be available
89 * to local zone users, so the underlying devices should be as well.
90 */
91 ASSERT(vd->vdev_path != NULL && vd->vdev_path[0] == '/');
92 error = vn_openat(vd->vdev_path + 1, UIO_SYSSPACE,
93 spa_mode(vd->vdev_spa) | FOFFMAX, 0, &vp, 0, 0, rootdir, -1);
94
95 if (error) {
96 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
97 return (error);
98 }
99
100 vf->vf_vnode = vp;
101
102 #ifdef _KERNEL
103 /*
104 * Make sure it's a regular file.
105 */
106 if (vp->v_type != VREG) {
107 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
108 return (SET_ERROR(ENODEV));
109 }
110 #endif
111
112 skip_open:
113 /*
114 * Determine the physical size of the file.
115 */
116 vattr.va_mask = AT_SIZE;
117 error = VOP_GETATTR(vf->vf_vnode, &vattr, 0, kcred, NULL);
118 if (error) {
119 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
120 return (error);
121 }
122
123 *max_psize = *psize = vattr.va_size;
124 *ashift = SPA_MINBLOCKSHIFT;
125
126 return (0);
127 }
128
129 static void
130 vdev_file_close(vdev_t *vd)
131 {
132 vdev_file_t *vf = vd->vdev_tsd;
133
134 if (vd->vdev_reopening || vf == NULL)
135 return;
136
137 if (vf->vf_vnode != NULL) {
138 (void) VOP_PUTPAGE(vf->vf_vnode, 0, 0, B_INVAL, kcred, NULL);
139 (void) VOP_CLOSE(vf->vf_vnode, spa_mode(vd->vdev_spa), 1, 0,
140 kcred, NULL);
141 VN_RELE(vf->vf_vnode);
142 }
143
144 vd->vdev_delayed_close = B_FALSE;
145 kmem_free(vf, sizeof (vdev_file_t));
146 vd->vdev_tsd = NULL;
147 }
148
149 /*
150 * Implements the interrupt side for file vdev types. This routine will be
151 * called when the I/O completes allowing us to transfer the I/O to the
152 * interrupt taskqs. For consistency, the code structure mimics disk vdev
153 * types.
154 */
155 static void
156 vdev_file_io_intr(buf_t *bp)
157 {
158 vdev_buf_t *vb = (vdev_buf_t *)bp;
159 zio_t *zio = vb->vb_io;
160
161 zio->io_error = (geterror(bp) != 0 ? EIO : 0);
162 if (zio->io_error == 0 && bp->b_resid != 0)
163 zio->io_error = SET_ERROR(ENOSPC);
164
165 kmem_free(vb, sizeof (vdev_buf_t));
166 zio_delay_interrupt(zio);
167 }
168
169 static void
170 vdev_file_io_strategy(void *arg)
171 {
172 buf_t *bp = arg;
173 vnode_t *vp = bp->b_private;
174 ssize_t resid;
175 int error;
176
177 error = vn_rdwr((bp->b_flags & B_READ) ? UIO_READ : UIO_WRITE,
178 vp, bp->b_un.b_addr, bp->b_bcount, ldbtob(bp->b_lblkno),
179 UIO_SYSSPACE, 0, RLIM64_INFINITY, kcred, &resid);
180
181 if (error == 0) {
182 bp->b_resid = resid;
183 biodone(bp);
184 } else {
185 bioerror(bp, error);
186 biodone(bp);
187 }
188 }
189
190 static void
191 vdev_file_io_start(zio_t *zio)
192 {
193 vdev_t *vd = zio->io_vd;
194 vdev_file_t *vf = vd->vdev_tsd;
195 vdev_buf_t *vb;
196 buf_t *bp;
197
198 if (zio->io_type == ZIO_TYPE_IOCTL) {
199 /* XXPOLICY */
200 if (!vdev_readable(vd)) {
201 zio->io_error = SET_ERROR(ENXIO);
202 zio_interrupt(zio);
203 return;
204 }
205
206 switch (zio->io_cmd) {
207 case DKIOCFLUSHWRITECACHE:
208 zio->io_error = VOP_FSYNC(vf->vf_vnode, FSYNC | FDSYNC,
209 kcred, NULL);
210 break;
211 default:
212 zio->io_error = SET_ERROR(ENOTSUP);
213 }
214
215 zio_execute(zio);
216 return;
217 }
218
219 ASSERT(zio->io_type == ZIO_TYPE_READ || zio->io_type == ZIO_TYPE_WRITE);
220 zio->io_target_timestamp = zio_handle_io_delay(zio);
221
222 vb = kmem_alloc(sizeof (vdev_buf_t), KM_SLEEP);
223
224 vb->vb_io = zio;
225 bp = &vb->vb_buf;
226
227 bioinit(bp);
228 bp->b_flags = (zio->io_type == ZIO_TYPE_READ ? B_READ : B_WRITE);
229 bp->b_bcount = zio->io_size;
230 bp->b_un.b_addr = zio->io_data;
231 bp->b_lblkno = lbtodb(zio->io_offset);
232 bp->b_bufsize = zio->io_size;
233 bp->b_private = vf->vf_vnode;
234 bp->b_iodone = (int (*)())vdev_file_io_intr;
235
236 VERIFY3U(taskq_dispatch(system_taskq, vdev_file_io_strategy, bp,
237 TQ_SLEEP), !=, 0);
238 }
239
240 /* ARGSUSED */
241 static void
242 vdev_file_io_done(zio_t *zio)
243 {
244 }
245
246 vdev_ops_t vdev_file_ops = {
247 vdev_file_open,
248 vdev_file_close,
249 vdev_default_asize,
250 vdev_file_io_start,
251 vdev_file_io_done,
252 NULL,
253 vdev_file_hold,
254 vdev_file_rele,
255 VDEV_TYPE_FILE, /* name of this vdev type */
256 B_TRUE /* leaf vdev */
257 };
258
259 /*
260 * From userland we access disks just like files.
261 */
262 #ifndef _KERNEL
263
264 vdev_ops_t vdev_disk_ops = {
265 vdev_file_open,
266 vdev_file_close,
267 vdev_default_asize,
268 vdev_file_io_start,
269 vdev_file_io_done,
270 NULL,
271 vdev_file_hold,
272 vdev_file_rele,
273 VDEV_TYPE_DISK, /* name of this vdev type */
274 B_TRUE /* leaf vdev */
275 };
276
277 #endif