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 #include <sys/abd.h>
35
36 /*
37 * Virtual device vector for files.
38 */
39
40 static void
41 vdev_file_hold(vdev_t *vd)
42 {
43 ASSERT(vd->vdev_path != NULL);
44 }
45
46 static void
47 vdev_file_rele(vdev_t *vd)
48 {
49 ASSERT(vd->vdev_path != NULL);
50 }
51
52 static int
53 vdev_file_open(vdev_t *vd, uint64_t *psize, uint64_t *max_psize,
54 uint64_t *ashift)
55 {
56 vdev_file_t *vf;
57 vnode_t *vp;
58 vattr_t vattr;
59 int error;
60
61 /*
62 * We must have a pathname, and it must be absolute.
63 */
64 if (vd->vdev_path == NULL || vd->vdev_path[0] != '/') {
65 vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
66 return (SET_ERROR(EINVAL));
67 }
68
69 /*
70 * Reopen the device if it's not currently open. Otherwise,
71 * just update the physical size of the device.
72 */
73 if (vd->vdev_tsd != NULL) {
74 ASSERT(vd->vdev_reopening);
75 vf = vd->vdev_tsd;
76 goto skip_open;
77 }
78
79 vf = vd->vdev_tsd = kmem_zalloc(sizeof (vdev_file_t), KM_SLEEP);
80
81 /*
82 * We always open the files from the root of the global zone, even if
83 * we're in a local zone. If the user has gotten to this point, the
84 * administrator has already decided that the pool should be available
85 * to local zone users, so the underlying devices should be as well.
86 */
87 ASSERT(vd->vdev_path != NULL && vd->vdev_path[0] == '/');
88 error = vn_openat(vd->vdev_path + 1, UIO_SYSSPACE,
89 spa_mode(vd->vdev_spa) | FOFFMAX, 0, &vp, 0, 0, rootdir, -1);
90
91 if (error) {
92 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
93 return (error);
94 }
95
96 vf->vf_vnode = vp;
97
98 #ifdef _KERNEL
99 /*
100 * Make sure it's a regular file.
101 */
102 if (vp->v_type != VREG) {
103 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
104 return (SET_ERROR(ENODEV));
105 }
106 #endif
107
108 skip_open:
109 /*
110 * Determine the physical size of the file.
111 */
112 vattr.va_mask = AT_SIZE;
113 error = VOP_GETATTR(vf->vf_vnode, &vattr, 0, kcred, NULL);
114 if (error) {
115 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
116 return (error);
117 }
118
119 *max_psize = *psize = vattr.va_size;
120 *ashift = SPA_MINBLOCKSHIFT;
121
122 return (0);
123 }
124
125 static void
126 vdev_file_close(vdev_t *vd)
127 {
128 vdev_file_t *vf = vd->vdev_tsd;
129
130 if (vd->vdev_reopening || vf == NULL)
131 return;
132
133 if (vf->vf_vnode != NULL) {
134 (void) VOP_PUTPAGE(vf->vf_vnode, 0, 0, B_INVAL, kcred, NULL);
135 (void) VOP_CLOSE(vf->vf_vnode, spa_mode(vd->vdev_spa), 1, 0,
136 kcred, NULL);
137 VN_RELE(vf->vf_vnode);
138 }
139
140 vd->vdev_delayed_close = B_FALSE;
141 kmem_free(vf, sizeof (vdev_file_t));
142 vd->vdev_tsd = NULL;
143 }
144
145 /*
146 * Implements the interrupt side for file vdev types. This routine will be
147 * called when the I/O completes allowing us to transfer the I/O to the
148 * interrupt taskqs. For consistency, the code structure mimics disk vdev
149 * types.
150 */
151 static void
152 vdev_file_io_intr(buf_t *bp)
153 {
154 vdev_buf_t *vb = (vdev_buf_t *)bp;
155 zio_t *zio = vb->vb_io;
156
157 zio->io_error = (geterror(bp) != 0 ? EIO : 0);
158 if (zio->io_error == 0 && bp->b_resid != 0)
159 zio->io_error = SET_ERROR(ENOSPC);
160
161 if (zio->io_type == ZIO_TYPE_READ) {
162 abd_return_buf_copy(zio->io_abd, bp->b_un.b_addr, zio->io_size);
163 } else {
164 abd_return_buf(zio->io_abd, bp->b_un.b_addr, zio->io_size);
165 }
166
167 kmem_free(vb, sizeof (vdev_buf_t));
168 zio_delay_interrupt(zio);
169 }
170
171 static void
172 vdev_file_io_strategy(void *arg)
173 {
174 buf_t *bp = arg;
175 vnode_t *vp = bp->b_private;
176 ssize_t resid;
177 int error;
178
179 error = vn_rdwr((bp->b_flags & B_READ) ? UIO_READ : UIO_WRITE,
180 vp, bp->b_un.b_addr, bp->b_bcount, ldbtob(bp->b_lblkno),
181 UIO_SYSSPACE, 0, RLIM64_INFINITY, kcred, &resid);
182
183 if (error == 0) {
184 bp->b_resid = resid;
185 biodone(bp);
186 } else {
187 bioerror(bp, error);
188 biodone(bp);
189 }
190 }
191
192 static void
193 vdev_file_io_start(zio_t *zio)
194 {
195 vdev_t *vd = zio->io_vd;
196 vdev_file_t *vf = vd->vdev_tsd;
197 vdev_buf_t *vb;
198 buf_t *bp;
199
200 if (zio->io_type == ZIO_TYPE_IOCTL) {
201 /* XXPOLICY */
202 if (!vdev_readable(vd)) {
203 zio->io_error = SET_ERROR(ENXIO);
204 zio_interrupt(zio);
205 return;
206 }
207
208 switch (zio->io_cmd) {
209 case DKIOCFLUSHWRITECACHE:
210 zio->io_error = VOP_FSYNC(vf->vf_vnode, FSYNC | FDSYNC,
211 kcred, NULL);
212 break;
213 default:
214 zio->io_error = SET_ERROR(ENOTSUP);
215 }
216
217 zio_execute(zio);
218 return;
219 }
220
221 ASSERT(zio->io_type == ZIO_TYPE_READ || zio->io_type == ZIO_TYPE_WRITE);
222 zio->io_target_timestamp = zio_handle_io_delay(zio);
223
224 vb = kmem_alloc(sizeof (vdev_buf_t), KM_SLEEP);
225
226 vb->vb_io = zio;
227 bp = &vb->vb_buf;
228
229 bioinit(bp);
230 bp->b_flags = (zio->io_type == ZIO_TYPE_READ ? B_READ : B_WRITE);
231 bp->b_bcount = zio->io_size;
232
233 if (zio->io_type == ZIO_TYPE_READ) {
234 bp->b_un.b_addr =
235 abd_borrow_buf(zio->io_abd, zio->io_size);
236 } else {
237 bp->b_un.b_addr =
238 abd_borrow_buf_copy(zio->io_abd, zio->io_size);
239 }
240
241 bp->b_lblkno = lbtodb(zio->io_offset);
242 bp->b_bufsize = zio->io_size;
243 bp->b_private = vf->vf_vnode;
244 bp->b_iodone = (int (*)())vdev_file_io_intr;
245
246 VERIFY3U(taskq_dispatch(system_taskq, vdev_file_io_strategy, bp,
247 TQ_SLEEP), !=, 0);
248 }
249
250 /* ARGSUSED */
251 static void
252 vdev_file_io_done(zio_t *zio)
253 {
254 }
255
256 vdev_ops_t vdev_file_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_FILE, /* name of this vdev type */
266 B_TRUE /* leaf vdev */
267 };
268
269 /*
270 * From userland we access disks just like files.
271 */
272 #ifndef _KERNEL
273
274 vdev_ops_t vdev_disk_ops = {
275 vdev_file_open,
276 vdev_file_close,
277 vdev_default_asize,
278 vdev_file_io_start,
279 vdev_file_io_done,
280 NULL,
281 vdev_file_hold,
282 vdev_file_rele,
283 VDEV_TYPE_DISK, /* name of this vdev type */
284 B_TRUE /* leaf vdev */
285 };
286
287 #endif