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.
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.
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]
22 * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
23 * Copyright (c) 2013 by Delphix. All rights reserved.
26 #include <sys/zfs_context.h>
28 #include <sys/spa_impl.h>
29 #include <sys/vdev_file.h>
30 #include <sys/vdev_impl.h>
32 #include <sys/fs/zfs.h>
33 #include <sys/fm/fs/zfs.h>
36 * Virtual device vector for files.
40 vdev_file_hold(vdev_t
*vd
)
42 ASSERT(vd
->vdev_path
!= NULL
);
46 vdev_file_rele(vdev_t
*vd
)
48 ASSERT(vd
->vdev_path
!= NULL
);
52 vdev_file_open(vdev_t
*vd
, uint64_t *psize
, uint64_t *max_psize
,
61 * We must have a pathname, and it must be absolute.
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
));
69 * Reopen the device if it's not currently open. Otherwise,
70 * just update the physical size of the device.
72 if (vd
->vdev_tsd
!= NULL
) {
73 ASSERT(vd
->vdev_reopening
);
78 vf
= vd
->vdev_tsd
= kmem_zalloc(sizeof (vdev_file_t
), KM_SLEEP
);
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.
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);
91 vd
->vdev_stat
.vs_aux
= VDEV_AUX_OPEN_FAILED
;
99 * Make sure it's a regular file.
101 if (vp
->v_type
!= VREG
) {
102 vd
->vdev_stat
.vs_aux
= VDEV_AUX_OPEN_FAILED
;
103 return (SET_ERROR(ENODEV
));
109 * Determine the physical size of the file.
111 vattr
.va_mask
= AT_SIZE
;
112 error
= VOP_GETATTR(vf
->vf_vnode
, &vattr
, 0, kcred
, NULL
);
114 vd
->vdev_stat
.vs_aux
= VDEV_AUX_OPEN_FAILED
;
118 *max_psize
= *psize
= vattr
.va_size
;
119 *ashift
= SPA_MINBLOCKSHIFT
;
125 vdev_file_close(vdev_t
*vd
)
127 vdev_file_t
*vf
= vd
->vdev_tsd
;
129 if (vd
->vdev_reopening
|| vf
== NULL
)
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,
136 VN_RELE(vf
->vf_vnode
);
139 vd
->vdev_delayed_close
= B_FALSE
;
140 kmem_free(vf
, sizeof (vdev_file_t
));
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
151 vdev_file_io_intr(buf_t
*bp
)
153 vdev_buf_t
*vb
= (vdev_buf_t
*)bp
;
154 zio_t
*zio
= vb
->vb_io
;
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
);
160 kmem_free(vb
, sizeof (vdev_buf_t
));
165 vdev_file_io_strategy(void *arg
)
168 vnode_t
*vp
= bp
->b_private
;
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
);
186 vdev_file_io_start(zio_t
*zio
)
188 vdev_t
*vd
= zio
->io_vd
;
189 vdev_file_t
*vf
= vd
->vdev_tsd
;
193 if (zio
->io_type
== ZIO_TYPE_IOCTL
) {
195 if (!vdev_readable(vd
)) {
196 zio
->io_error
= SET_ERROR(ENXIO
);
197 return (ZIO_PIPELINE_CONTINUE
);
200 switch (zio
->io_cmd
) {
201 case DKIOCFLUSHWRITECACHE
:
202 zio
->io_error
= VOP_FSYNC(vf
->vf_vnode
, FSYNC
| FDSYNC
,
206 zio
->io_error
= SET_ERROR(ENOTSUP
);
209 return (ZIO_PIPELINE_CONTINUE
);
212 vb
= kmem_alloc(sizeof (vdev_buf_t
), KM_SLEEP
);
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
;
226 VERIFY3U(taskq_dispatch(system_taskq
, vdev_file_io_strategy
, bp
,
229 return (ZIO_PIPELINE_STOP
);
234 vdev_file_io_done(zio_t
*zio
)
238 vdev_ops_t vdev_file_ops
= {
247 VDEV_TYPE_FILE
, /* name of this vdev type */
248 B_TRUE
/* leaf vdev */
252 * From userland we access disks just like files.
256 vdev_ops_t vdev_disk_ops
= {
265 VDEV_TYPE_DISK
, /* name of this vdev type */
266 B_TRUE
/* leaf vdev */