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, 2016 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 /* Rotational optimizations only make sense on block devices */ 62 vd->vdev_nonrot = B_TRUE; 63 64 /* 65 * We must have a pathname, and it must be absolute. 66 */ 67 if (vd->vdev_path == NULL || vd->vdev_path[0] != '/') { 68 vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL; 69 return (SET_ERROR(EINVAL)); 70 } 71 72 /* 73 * Reopen the device if it's not currently open. Otherwise, 74 * just update the physical size of the device. 75 */ 76 if (vd->vdev_tsd != NULL) { 77 ASSERT(vd->vdev_reopening); 78 vf = vd->vdev_tsd; 79 goto skip_open; 80 } 81 82 vf = vd->vdev_tsd = kmem_zalloc(sizeof (vdev_file_t), KM_SLEEP); 83 84 /* 85 * We always open the files from the root of the global zone, even if 86 * we're in a local zone. If the user has gotten to this point, the 87 * administrator has already decided that the pool should be available 88 * to local zone users, so the underlying devices should be as well. 89 */ 90 ASSERT(vd->vdev_path != NULL && vd->vdev_path[0] == '/'); 91 error = vn_openat(vd->vdev_path + 1, UIO_SYSSPACE, 92 spa_mode(vd->vdev_spa) | FOFFMAX, 0, &vp, 0, 0, rootdir, -1); 93 94 if (error) { 95 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED; 96 return (error); 97 } 98 99 vf->vf_vnode = vp; 100 101 #ifdef _KERNEL 102 /* 103 * Make sure it's a regular file. 104 */ 105 if (vp->v_type != VREG) { 106 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED; 107 return (SET_ERROR(ENODEV)); 108 } 109 #endif 110 111 skip_open: 112 /* 113 * Determine the physical size of the file. 114 */ 115 vattr.va_mask = AT_SIZE; 116 error = VOP_GETATTR(vf->vf_vnode, &vattr, 0, kcred, NULL); 117 if (error) { 118 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED; 119 return (error); 120 } 121 122 *max_psize = *psize = vattr.va_size; 123 *ashift = SPA_MINBLOCKSHIFT; 124 125 return (0); 126 } 127 128 static void 129 vdev_file_close(vdev_t *vd) 130 { 131 vdev_file_t *vf = vd->vdev_tsd; 132 133 if (vd->vdev_reopening || vf == NULL) 134 return; 135 136 if (vf->vf_vnode != NULL) { 137 (void) VOP_PUTPAGE(vf->vf_vnode, 0, 0, B_INVAL, kcred, NULL); 138 (void) VOP_CLOSE(vf->vf_vnode, spa_mode(vd->vdev_spa), 1, 0, 139 kcred, NULL); 140 VN_RELE(vf->vf_vnode); 141 } 142 143 vd->vdev_delayed_close = B_FALSE; 144 kmem_free(vf, sizeof (vdev_file_t)); 145 vd->vdev_tsd = NULL; 146 } 147 148 /* 149 * Implements the interrupt side for file vdev types. This routine will be 150 * called when the I/O completes allowing us to transfer the I/O to the 151 * interrupt taskqs. For consistency, the code structure mimics disk vdev 152 * types. 153 */ 154 static int 155 vdev_file_io_intr(buf_t *bp) 156 { 157 vdev_buf_t *vb = (vdev_buf_t *)bp; 158 zio_t *zio = vb->vb_io; 159 160 zio->io_error = (geterror(bp) != 0 ? EIO : 0); 161 if (zio->io_error == 0 && bp->b_resid != 0) 162 zio->io_error = SET_ERROR(ENOSPC); 163 164 if (zio->io_type == ZIO_TYPE_READ) { 165 abd_return_buf_copy(zio->io_abd, bp->b_un.b_addr, zio->io_size); 166 } else { 167 abd_return_buf(zio->io_abd, bp->b_un.b_addr, zio->io_size); 168 } 169 170 kmem_free(vb, sizeof (vdev_buf_t)); 171 zio_delay_interrupt(zio); 172 return (0); 173 } 174 175 static void 176 vdev_file_io_strategy(void *arg) 177 { 178 buf_t *bp = arg; 179 vnode_t *vp = bp->b_private; 180 ssize_t resid; 181 int error; 182 183 error = vn_rdwr((bp->b_flags & B_READ) ? UIO_READ : UIO_WRITE, 184 vp, bp->b_un.b_addr, bp->b_bcount, ldbtob(bp->b_lblkno), 185 UIO_SYSSPACE, 0, RLIM64_INFINITY, kcred, &resid); 186 187 if (error == 0) { 188 bp->b_resid = resid; 189 biodone(bp); 190 } else { 191 bioerror(bp, error); 192 biodone(bp); 193 } 194 } 195 196 static void 197 vdev_file_io_start(zio_t *zio) 198 { 199 vdev_t *vd = zio->io_vd; 200 vdev_file_t *vf = vd->vdev_tsd; 201 vdev_buf_t *vb; 202 buf_t *bp; 203 204 if (zio->io_type == ZIO_TYPE_IOCTL) { 205 /* XXPOLICY */ 206 if (!vdev_readable(vd)) { 207 zio->io_error = SET_ERROR(ENXIO); 208 zio_interrupt(zio); 209 return; 210 } 211 212 switch (zio->io_cmd) { 213 case DKIOCFLUSHWRITECACHE: 214 zio->io_error = VOP_FSYNC(vf->vf_vnode, FSYNC | FDSYNC, 215 kcred, NULL); 216 break; 217 default: 218 zio->io_error = SET_ERROR(ENOTSUP); 219 } 220 221 zio_execute(zio); 222 return; 223 } 224 225 ASSERT(zio->io_type == ZIO_TYPE_READ || zio->io_type == ZIO_TYPE_WRITE); 226 zio->io_target_timestamp = zio_handle_io_delay(zio); 227 228 vb = kmem_alloc(sizeof (vdev_buf_t), KM_SLEEP); 229 230 vb->vb_io = zio; 231 bp = &vb->vb_buf; 232 233 bioinit(bp); 234 bp->b_flags = (zio->io_type == ZIO_TYPE_READ ? B_READ : B_WRITE); 235 bp->b_bcount = zio->io_size; 236 237 if (zio->io_type == ZIO_TYPE_READ) { 238 bp->b_un.b_addr = 239 abd_borrow_buf(zio->io_abd, zio->io_size); 240 } else { 241 bp->b_un.b_addr = 242 abd_borrow_buf_copy(zio->io_abd, zio->io_size); 243 } 244 245 bp->b_lblkno = lbtodb(zio->io_offset); 246 bp->b_bufsize = zio->io_size; 247 bp->b_private = vf->vf_vnode; 248 bp->b_iodone = vdev_file_io_intr; 249 250 VERIFY3U(taskq_dispatch(system_taskq, vdev_file_io_strategy, bp, 251 TQ_SLEEP), !=, TASKQID_INVALID); 252 } 253 254 /* ARGSUSED */ 255 static void 256 vdev_file_io_done(zio_t *zio) 257 { 258 } 259 260 vdev_ops_t vdev_file_ops = { 261 .vdev_op_open = vdev_file_open, 262 .vdev_op_close = vdev_file_close, 263 .vdev_op_asize = vdev_default_asize, 264 .vdev_op_io_start = vdev_file_io_start, 265 .vdev_op_io_done = vdev_file_io_done, 266 .vdev_op_state_change = NULL, 267 .vdev_op_need_resilver = NULL, 268 .vdev_op_hold = vdev_file_hold, 269 .vdev_op_rele = vdev_file_rele, 270 .vdev_op_remap = NULL, 271 .vdev_op_xlate = vdev_default_xlate, 272 .vdev_op_type = VDEV_TYPE_FILE, /* name of this vdev type */ 273 .vdev_op_leaf = B_TRUE /* leaf vdev */ 274 }; 275 276 /* 277 * From userland we access disks just like files. 278 */ 279 #ifndef _KERNEL 280 281 vdev_ops_t vdev_disk_ops = { 282 .vdev_op_open = vdev_file_open, 283 .vdev_op_close = vdev_file_close, 284 .vdev_op_asize = vdev_default_asize, 285 .vdev_op_io_start = vdev_file_io_start, 286 .vdev_op_io_done = vdev_file_io_done, 287 .vdev_op_state_change = NULL, 288 .vdev_op_need_resilver = NULL, 289 .vdev_op_hold = vdev_file_hold, 290 .vdev_op_rele = vdev_file_rele, 291 .vdev_op_remap = NULL, 292 .vdev_op_xlate = vdev_default_xlate, 293 .vdev_op_type = VDEV_TYPE_DISK, /* name of this vdev type */ 294 .vdev_op_leaf = B_TRUE /* leaf vdev */ 295 }; 296 297 #endif 298