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 2006 Sun Microsystems, Inc. All rights reserved. 23 * Use is subject to license terms. 24 */ 25 26 #pragma ident "%Z%%M% %I% %E% SMI" 27 28 #include <sys/zfs_context.h> 29 #include <sys/spa.h> 30 #include <sys/vdev_disk.h> 31 #include <sys/vdev_impl.h> 32 #include <sys/fs/zfs.h> 33 #include <sys/zio.h> 34 #include <sys/sunldi.h> 35 36 /* 37 * Virtual device vector for disks. 38 */ 39 40 extern ldi_ident_t zfs_li; 41 42 typedef struct vdev_disk_buf { 43 buf_t vdb_buf; 44 zio_t *vdb_io; 45 } vdev_disk_buf_t; 46 47 static int 48 vdev_disk_open(vdev_t *vd, uint64_t *psize, uint64_t *ashift) 49 { 50 vdev_disk_t *dvd; 51 struct dk_minfo dkm; 52 int error; 53 54 /* 55 * We must have a pathname, and it must be absolute. 56 */ 57 if (vd->vdev_path == NULL || vd->vdev_path[0] != '/') { 58 vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL; 59 return (EINVAL); 60 } 61 62 dvd = vd->vdev_tsd = kmem_zalloc(sizeof (vdev_disk_t), KM_SLEEP); 63 64 /* 65 * When opening a disk device, we want to preserve the user's original 66 * intent. We always want to open the device by the path the user gave 67 * us, even if it is one of multiple paths to the save device. But we 68 * also want to be able to survive disks being removed/recabled. 69 * Therefore the sequence of opening devices is: 70 * 71 * 1. Try opening the device by path. For legacy pools without the 72 * 'whole_disk' property, attempt to fix the path by appending 's0'. 73 * 74 * 2. If the devid of the device matches the stored value, return 75 * success. 76 * 77 * 3. Otherwise, the device may have moved. Try opening the device 78 * by the devid instead. 79 * 80 */ 81 if (vd->vdev_devid != NULL) { 82 if (ddi_devid_str_decode(vd->vdev_devid, &dvd->vd_devid, 83 &dvd->vd_minor) != 0) { 84 vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL; 85 return (EINVAL); 86 } 87 } 88 89 error = EINVAL; /* presume failure */ 90 91 if (vd->vdev_path != NULL) { 92 ddi_devid_t devid; 93 94 if (vd->vdev_wholedisk == -1ULL) { 95 size_t len = strlen(vd->vdev_path) + 3; 96 char *buf = kmem_alloc(len, KM_SLEEP); 97 ldi_handle_t lh; 98 99 (void) snprintf(buf, len, "%ss0", vd->vdev_path); 100 101 if (ldi_open_by_name(buf, spa_mode, kcred, 102 &lh, zfs_li) == 0) { 103 spa_strfree(vd->vdev_path); 104 vd->vdev_path = buf; 105 vd->vdev_wholedisk = 1ULL; 106 (void) ldi_close(lh, spa_mode, kcred); 107 } else { 108 kmem_free(buf, len); 109 } 110 } 111 112 error = ldi_open_by_name(vd->vdev_path, spa_mode, kcred, 113 &dvd->vd_lh, zfs_li); 114 115 /* 116 * Compare the devid to the stored value. 117 */ 118 if (error == 0 && vd->vdev_devid != NULL && 119 ldi_get_devid(dvd->vd_lh, &devid) == 0) { 120 if (ddi_devid_compare(devid, dvd->vd_devid) != 0) { 121 error = EINVAL; 122 (void) ldi_close(dvd->vd_lh, spa_mode, kcred); 123 dvd->vd_lh = NULL; 124 } 125 ddi_devid_free(devid); 126 } 127 128 /* 129 * If we succeeded in opening the device, but 'vdev_wholedisk' 130 * is not yet set, then this must be a slice. 131 */ 132 if (error == 0 && vd->vdev_wholedisk == -1ULL) 133 vd->vdev_wholedisk = 0; 134 } 135 136 /* 137 * If we were unable to open by path, or the devid check fails, open by 138 * devid instead. 139 */ 140 if (error != 0 && vd->vdev_devid != NULL) 141 error = ldi_open_by_devid(dvd->vd_devid, dvd->vd_minor, 142 spa_mode, kcred, &dvd->vd_lh, zfs_li); 143 144 if (error) { 145 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED; 146 return (error); 147 } 148 149 /* 150 * Determine the actual size of the device. 151 */ 152 if (ldi_get_size(dvd->vd_lh, psize) != 0) { 153 vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED; 154 return (EINVAL); 155 } 156 157 /* 158 * If we own the whole disk, try to enable disk write caching. 159 * We ignore errors because it's OK if we can't do it. 160 */ 161 if (vd->vdev_wholedisk == 1) { 162 int wce = 1; 163 (void) ldi_ioctl(dvd->vd_lh, DKIOCSETWCE, (intptr_t)&wce, 164 FKIOCTL, kcred, NULL); 165 } 166 167 /* 168 * Determine the device's minimum transfer size. 169 * If the ioctl isn't supported, assume DEV_BSIZE. 170 */ 171 if (ldi_ioctl(dvd->vd_lh, DKIOCGMEDIAINFO, (intptr_t)&dkm, 172 FKIOCTL, kcred, NULL) != 0) 173 dkm.dki_lbsize = DEV_BSIZE; 174 175 *ashift = highbit(MAX(dkm.dki_lbsize, SPA_MINBLOCKSIZE)) - 1; 176 177 /* 178 * Clear the nowritecache bit, so that on a vdev_reopen() we will 179 * try again. 180 */ 181 vd->vdev_nowritecache = B_FALSE; 182 183 return (0); 184 } 185 186 static void 187 vdev_disk_close(vdev_t *vd) 188 { 189 vdev_disk_t *dvd = vd->vdev_tsd; 190 191 if (dvd == NULL) 192 return; 193 194 dprintf("removing disk %s, devid %s\n", 195 vd->vdev_path ? vd->vdev_path : "<none>", 196 vd->vdev_devid ? vd->vdev_devid : "<none>"); 197 198 if (dvd->vd_minor != NULL) 199 ddi_devid_str_free(dvd->vd_minor); 200 201 if (dvd->vd_devid != NULL) 202 ddi_devid_free(dvd->vd_devid); 203 204 if (dvd->vd_lh != NULL) 205 (void) ldi_close(dvd->vd_lh, spa_mode, kcred); 206 207 kmem_free(dvd, sizeof (vdev_disk_t)); 208 vd->vdev_tsd = NULL; 209 } 210 211 static void 212 vdev_disk_io_intr(buf_t *bp) 213 { 214 vdev_disk_buf_t *vdb = (vdev_disk_buf_t *)bp; 215 zio_t *zio = vdb->vdb_io; 216 217 if ((zio->io_error = geterror(bp)) == 0 && bp->b_resid != 0) 218 zio->io_error = EIO; 219 220 kmem_free(vdb, sizeof (vdev_disk_buf_t)); 221 222 zio_next_stage_async(zio); 223 } 224 225 static void 226 vdev_disk_ioctl_done(void *zio_arg, int error) 227 { 228 zio_t *zio = zio_arg; 229 230 zio->io_error = error; 231 232 zio_next_stage_async(zio); 233 } 234 235 static void 236 vdev_disk_io_start(zio_t *zio) 237 { 238 vdev_t *vd = zio->io_vd; 239 vdev_disk_t *dvd = vd->vdev_tsd; 240 vdev_disk_buf_t *vdb; 241 buf_t *bp; 242 int flags, error; 243 244 if (zio->io_type == ZIO_TYPE_IOCTL) { 245 zio_vdev_io_bypass(zio); 246 247 /* XXPOLICY */ 248 if (vdev_is_dead(vd)) { 249 zio->io_error = ENXIO; 250 zio_next_stage_async(zio); 251 return; 252 } 253 254 switch (zio->io_cmd) { 255 256 case DKIOCFLUSHWRITECACHE: 257 258 if (vd->vdev_nowritecache) { 259 zio->io_error = ENOTSUP; 260 break; 261 } 262 263 zio->io_dk_callback.dkc_callback = vdev_disk_ioctl_done; 264 zio->io_dk_callback.dkc_cookie = zio; 265 266 error = ldi_ioctl(dvd->vd_lh, zio->io_cmd, 267 (uintptr_t)&zio->io_dk_callback, 268 FKIOCTL, kcred, NULL); 269 270 if (error == 0) { 271 /* 272 * The ioctl will be done asychronously, 273 * and will call vdev_disk_ioctl_done() 274 * upon completion. 275 */ 276 return; 277 } else if (error == ENOTSUP) { 278 /* 279 * If we get ENOTSUP, we know that no future 280 * attempts will ever succeed. In this case we 281 * set a persistent bit so that we don't bother 282 * with the ioctl in the future. 283 */ 284 vd->vdev_nowritecache = B_TRUE; 285 } 286 zio->io_error = error; 287 288 break; 289 290 default: 291 zio->io_error = ENOTSUP; 292 } 293 294 zio_next_stage_async(zio); 295 return; 296 } 297 298 if (zio->io_type == ZIO_TYPE_READ && vdev_cache_read(zio) == 0) 299 return; 300 301 if ((zio = vdev_queue_io(zio)) == NULL) 302 return; 303 304 flags = (zio->io_type == ZIO_TYPE_READ ? B_READ : B_WRITE); 305 flags |= B_BUSY | B_NOCACHE; 306 if (zio->io_flags & ZIO_FLAG_FAILFAST) 307 flags |= B_FAILFAST; 308 309 vdb = kmem_alloc(sizeof (vdev_disk_buf_t), KM_SLEEP); 310 311 vdb->vdb_io = zio; 312 bp = &vdb->vdb_buf; 313 314 bioinit(bp); 315 bp->b_flags = flags; 316 bp->b_bcount = zio->io_size; 317 bp->b_un.b_addr = zio->io_data; 318 bp->b_lblkno = lbtodb(zio->io_offset); 319 bp->b_bufsize = zio->io_size; 320 bp->b_iodone = (int (*)())vdev_disk_io_intr; 321 322 /* XXPOLICY */ 323 error = vdev_is_dead(vd) ? ENXIO : vdev_error_inject(vd, zio); 324 if (error) { 325 zio->io_error = error; 326 bioerror(bp, error); 327 bp->b_resid = bp->b_bcount; 328 bp->b_iodone(bp); 329 return; 330 } 331 332 error = ldi_strategy(dvd->vd_lh, bp); 333 /* ldi_strategy() will return non-zero only on programming errors */ 334 ASSERT(error == 0); 335 } 336 337 static void 338 vdev_disk_io_done(zio_t *zio) 339 { 340 vdev_queue_io_done(zio); 341 342 if (zio->io_type == ZIO_TYPE_WRITE) 343 vdev_cache_write(zio); 344 345 if (zio_injection_enabled && zio->io_error == 0) 346 zio->io_error = zio_handle_device_injection(zio->io_vd, EIO); 347 348 zio_next_stage(zio); 349 } 350 351 vdev_ops_t vdev_disk_ops = { 352 vdev_disk_open, 353 vdev_disk_close, 354 vdev_default_asize, 355 vdev_disk_io_start, 356 vdev_disk_io_done, 357 NULL, 358 VDEV_TYPE_DISK, /* name of this vdev type */ 359 B_TRUE /* leaf vdev */ 360 }; 361