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 * Rotational optimizations only make sense on block devices
63 */
64 vd->vdev_nonrot = B_TRUE;
65
66 /*
67 * We must have a pathname, and it must be absolute.
68 */
69 if (vd->vdev_path == NULL || vd->vdev_path[0] != '/') {
70 vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
71 return (SET_ERROR(EINVAL));
72 }
73
74 /*
75 * Reopen the device if it's not currently open. Otherwise,
76 * just update the physical size of the device.
77 */
78 if (vd->vdev_tsd != NULL) {
79 ASSERT(vd->vdev_reopening);
80 vf = vd->vdev_tsd;
81 goto skip_open;
82 }
83
84 vf = vd->vdev_tsd = kmem_zalloc(sizeof (vdev_file_t), KM_SLEEP);
85
86 /*
87 * We always open the files from the root of the global zone, even if
88 * we're in a local zone. If the user has gotten to this point, the
89 * administrator has already decided that the pool should be available
90 * to local zone users, so the underlying devices should be as well.
91 */
92 ASSERT(vd->vdev_path != NULL && vd->vdev_path[0] == '/');
93 error = vn_openat(vd->vdev_path + 1, UIO_SYSSPACE,
94 spa_mode(vd->vdev_spa) | FOFFMAX, 0, &vp, 0, 0, rootdir, -1);
95
96 if (error) {
97 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
98 return (error);
99 }
100
101 vf->vf_vnode = vp;
102
103 #ifdef _KERNEL
104 /*
105 * Make sure it's a regular file.
106 */
107 if (vp->v_type != VREG) {
108 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
109 return (SET_ERROR(ENODEV));
110 }
111 #endif
112
113 skip_open:
114 /*
115 * Determine the physical size of the file.
116 */
117 vattr.va_mask = AT_SIZE;
118 error = VOP_GETATTR(vf->vf_vnode, &vattr, 0, kcred, NULL);
119 if (error) {
120 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED;
121 return (error);
122 }
123
124 *max_psize = *psize = vattr.va_size;
125 *ashift = SPA_MINBLOCKSHIFT;
126
127 return (0);
128 }
129
130 static void
131 vdev_file_close(vdev_t *vd)
132 {
133 vdev_file_t *vf = vd->vdev_tsd;
134
135 if (vd->vdev_reopening || vf == NULL)
136 return;
137
138 if (vf->vf_vnode != NULL) {
139 (void) VOP_PUTPAGE(vf->vf_vnode, 0, 0, B_INVAL, kcred, NULL);
140 (void) VOP_CLOSE(vf->vf_vnode, spa_mode(vd->vdev_spa), 1, 0,
141 kcred, NULL);
142 VN_RELE(vf->vf_vnode);
143 }
144
145 vd->vdev_delayed_close = B_FALSE;
146 kmem_free(vf, sizeof (vdev_file_t));
147 vd->vdev_tsd = NULL;
148 }
149
150 /*
151 * Implements the interrupt side for file vdev types. This routine will be
152 * called when the I/O completes allowing us to transfer the I/O to the
153 * interrupt taskqs. For consistency, the code structure mimics disk vdev
154 * types.
155 */
156 static void
157 vdev_file_io_intr(buf_t *bp)
158 {
159 vdev_buf_t *vb = (vdev_buf_t *)bp;
160 zio_t *zio = vb->vb_io;
161
162 zio->io_error = (geterror(bp) != 0 ? EIO : 0);
163 if (zio->io_error == 0 && bp->b_resid != 0)
164 zio->io_error = SET_ERROR(ENOSPC);
165
166 if (zio->io_type == ZIO_TYPE_READ) {
167 abd_return_buf_copy(zio->io_abd, bp->b_un.b_addr, zio->io_size);
168 } else {
169 abd_return_buf(zio->io_abd, bp->b_un.b_addr, zio->io_size);
170 }
171
172 kmem_free(vb, sizeof (vdev_buf_t));
173 zio_delay_interrupt(zio);
174 }
175
176 static void
177 vdev_file_io_strategy(void *arg)
178 {
179 buf_t *bp = arg;
180 vnode_t *vp = bp->b_private;
181 ssize_t resid;
182 int error;
183
184 error = vn_rdwr((bp->b_flags & B_READ) ? UIO_READ : UIO_WRITE,
185 vp, bp->b_un.b_addr, bp->b_bcount, ldbtob(bp->b_lblkno),
186 UIO_SYSSPACE, 0, RLIM64_INFINITY, kcred, &resid);
187
188 if (error == 0) {
189 bp->b_resid = resid;
190 biodone(bp);
191 } else {
192 bioerror(bp, error);
193 biodone(bp);
194 }
195 }
196
197 static void
198 vdev_file_io_start(zio_t *zio)
199 {
200 vdev_t *vd = zio->io_vd;
201 vdev_file_t *vf = vd->vdev_tsd;
202 vdev_buf_t *vb;
203 buf_t *bp;
204
205 if (zio->io_type == ZIO_TYPE_IOCTL) {
206 /* XXPOLICY */
207 if (!vdev_readable(vd)) {
208 zio->io_error = SET_ERROR(ENXIO);
209 zio_interrupt(zio);
210 return;
211 }
212
213 switch (zio->io_cmd) {
214 case DKIOCFLUSHWRITECACHE:
215 zio->io_error = VOP_FSYNC(vf->vf_vnode, FSYNC | FDSYNC,
216 kcred, NULL);
217 break;
218 default:
219 zio->io_error = SET_ERROR(ENOTSUP);
220 }
221
222 zio_execute(zio);
223 return;
224 }
225
226 ASSERT(zio->io_type == ZIO_TYPE_READ || zio->io_type == ZIO_TYPE_WRITE);
227 zio->io_target_timestamp = zio_handle_io_delay(zio);
228
229 vb = kmem_alloc(sizeof (vdev_buf_t), KM_SLEEP);
230
231 vb->vb_io = zio;
232 bp = &vb->vb_buf;
233
234 bioinit(bp);
235 bp->b_flags = (zio->io_type == ZIO_TYPE_READ ? B_READ : B_WRITE);
236 bp->b_bcount = zio->io_size;
237
238 if (zio->io_type == ZIO_TYPE_READ) {
239 bp->b_un.b_addr =
240 abd_borrow_buf(zio->io_abd, zio->io_size);
241 } else {
242 bp->b_un.b_addr =
243 abd_borrow_buf_copy(zio->io_abd, zio->io_size);
244 }
245
246 bp->b_lblkno = lbtodb(zio->io_offset);
247 bp->b_bufsize = zio->io_size;
248 bp->b_private = vf->vf_vnode;
249 bp->b_iodone = (int (*)())vdev_file_io_intr;
250
251 VERIFY3U(taskq_dispatch(system_taskq, vdev_file_io_strategy, bp,
252 TQ_SLEEP), !=, 0);
253 }
254
255 /* ARGSUSED */
256 static void
257 vdev_file_io_done(zio_t *zio)
258 {
259 }
260
261 vdev_ops_t vdev_file_ops = {
262 vdev_file_open,
263 vdev_file_close,
264 vdev_default_asize,
265 vdev_file_io_start,
266 vdev_file_io_done,
267 NULL,
268 vdev_file_hold,
269 vdev_file_rele,
270 VDEV_TYPE_FILE, /* name of this vdev type */
271 B_TRUE /* leaf vdev */
272 };
273
274 /*
275 * From userland we access disks just like files.
276 */
277 #ifndef _KERNEL
278
279 vdev_ops_t vdev_disk_ops = {
280 vdev_file_open,
281 vdev_file_close,
282 vdev_default_asize,
283 vdev_file_io_start,
284 vdev_file_io_done,
285 NULL,
286 vdev_file_hold,
287 vdev_file_rele,
288 VDEV_TYPE_DISK, /* name of this vdev type */
289 B_TRUE /* leaf vdev */
290 };
291
292 #endif