1fa9e4066Sahrens /* 2fa9e4066Sahrens * CDDL HEADER START 3fa9e4066Sahrens * 4fa9e4066Sahrens * The contents of this file are subject to the terms of the 5bef6b7d2Swebaker * Common Development and Distribution License (the "License"). 6bef6b7d2Swebaker * You may not use this file except in compliance with the License. 7fa9e4066Sahrens * 8fa9e4066Sahrens * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9fa9e4066Sahrens * or http://www.opensolaris.org/os/licensing. 10fa9e4066Sahrens * See the License for the specific language governing permissions 11fa9e4066Sahrens * and limitations under the License. 12fa9e4066Sahrens * 13fa9e4066Sahrens * When distributing Covered Code, include this CDDL HEADER in each 14fa9e4066Sahrens * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15fa9e4066Sahrens * If applicable, add the following below this CDDL HEADER, with the 16fa9e4066Sahrens * fields enclosed by brackets "[]" replaced with your own identifying 17fa9e4066Sahrens * information: Portions Copyright [yyyy] [name of copyright owner] 18fa9e4066Sahrens * 19fa9e4066Sahrens * CDDL HEADER END 20fa9e4066Sahrens */ 21fa9e4066Sahrens /* 22e7cbe64fSgw * Copyright 2008 Sun Microsystems, Inc. All rights reserved. 23fa9e4066Sahrens * Use is subject to license terms. 24fa9e4066Sahrens */ 25fa9e4066Sahrens 26fa9e4066Sahrens #include <sys/zfs_context.h> 27fa9e4066Sahrens #include <sys/spa.h> 28e7cbe64fSgw #include <sys/refcount.h> 29fa9e4066Sahrens #include <sys/vdev_disk.h> 30fa9e4066Sahrens #include <sys/vdev_impl.h> 31fa9e4066Sahrens #include <sys/fs/zfs.h> 32fa9e4066Sahrens #include <sys/zio.h> 33afefbcddSeschrock #include <sys/sunldi.h> 3451ece835Seschrock #include <sys/fm/fs/zfs.h> 35fa9e4066Sahrens 36fa9e4066Sahrens /* 37fa9e4066Sahrens * Virtual device vector for disks. 38fa9e4066Sahrens */ 39fa9e4066Sahrens 40fa9e4066Sahrens extern ldi_ident_t zfs_li; 41fa9e4066Sahrens 42fa9e4066Sahrens typedef struct vdev_disk_buf { 43fa9e4066Sahrens buf_t vdb_buf; 44fa9e4066Sahrens zio_t *vdb_io; 45fa9e4066Sahrens } vdev_disk_buf_t; 46fa9e4066Sahrens 47fa9e4066Sahrens static int 48e14bb325SJeff Bonwick vdev_disk_open(vdev_t *vd, uint64_t *psize, uint64_t *ashift) 49fa9e4066Sahrens { 50*8ad4d6ddSJeff Bonwick spa_t *spa = vd->vdev_spa; 51fa9e4066Sahrens vdev_disk_t *dvd; 52e14bb325SJeff Bonwick struct dk_minfo dkm; 530a4e9518Sgw int error; 54e14bb325SJeff Bonwick dev_t dev; 55e14bb325SJeff Bonwick int otyp; 56fa9e4066Sahrens 57fa9e4066Sahrens /* 58fa9e4066Sahrens * We must have a pathname, and it must be absolute. 59fa9e4066Sahrens */ 60fa9e4066Sahrens if (vd->vdev_path == NULL || vd->vdev_path[0] != '/') { 61fa9e4066Sahrens vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL; 62fa9e4066Sahrens return (EINVAL); 63fa9e4066Sahrens } 64fa9e4066Sahrens 65fa9e4066Sahrens dvd = vd->vdev_tsd = kmem_zalloc(sizeof (vdev_disk_t), KM_SLEEP); 66fa9e4066Sahrens 67fa9e4066Sahrens /* 68fa9e4066Sahrens * When opening a disk device, we want to preserve the user's original 69fa9e4066Sahrens * intent. We always want to open the device by the path the user gave 70fa9e4066Sahrens * us, even if it is one of multiple paths to the save device. But we 71fa9e4066Sahrens * also want to be able to survive disks being removed/recabled. 72fa9e4066Sahrens * Therefore the sequence of opening devices is: 73fa9e4066Sahrens * 74afefbcddSeschrock * 1. Try opening the device by path. For legacy pools without the 75afefbcddSeschrock * 'whole_disk' property, attempt to fix the path by appending 's0'. 76fa9e4066Sahrens * 77fa9e4066Sahrens * 2. If the devid of the device matches the stored value, return 78fa9e4066Sahrens * success. 79fa9e4066Sahrens * 80fa9e4066Sahrens * 3. Otherwise, the device may have moved. Try opening the device 81fa9e4066Sahrens * by the devid instead. 82fa9e4066Sahrens * 83bf82a41bSeschrock * If the vdev is part of the root pool, we avoid opening it by path. 84bf82a41bSeschrock * We do this because there is no /dev path available early in boot, 85bf82a41bSeschrock * and if we try to open the device by path at a later point, we can 86bf82a41bSeschrock * deadlock when devfsadm attempts to open the underlying backing store 87bf82a41bSeschrock * file. 88fa9e4066Sahrens */ 89fa9e4066Sahrens if (vd->vdev_devid != NULL) { 90fa9e4066Sahrens if (ddi_devid_str_decode(vd->vdev_devid, &dvd->vd_devid, 91fa9e4066Sahrens &dvd->vd_minor) != 0) { 92fa9e4066Sahrens vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL; 93fa9e4066Sahrens return (EINVAL); 94fa9e4066Sahrens } 95fa9e4066Sahrens } 96fa9e4066Sahrens 97fa9e4066Sahrens error = EINVAL; /* presume failure */ 98fa9e4066Sahrens 99*8ad4d6ddSJeff Bonwick if (vd->vdev_path != NULL && !spa_is_root(spa)) { 100fa9e4066Sahrens ddi_devid_t devid; 101fa9e4066Sahrens 102afefbcddSeschrock if (vd->vdev_wholedisk == -1ULL) { 103afefbcddSeschrock size_t len = strlen(vd->vdev_path) + 3; 104afefbcddSeschrock char *buf = kmem_alloc(len, KM_SLEEP); 105afefbcddSeschrock ldi_handle_t lh; 106afefbcddSeschrock 107afefbcddSeschrock (void) snprintf(buf, len, "%ss0", vd->vdev_path); 108afefbcddSeschrock 109*8ad4d6ddSJeff Bonwick if (ldi_open_by_name(buf, spa_mode(spa), kcred, 110afefbcddSeschrock &lh, zfs_li) == 0) { 111afefbcddSeschrock spa_strfree(vd->vdev_path); 112afefbcddSeschrock vd->vdev_path = buf; 113afefbcddSeschrock vd->vdev_wholedisk = 1ULL; 114*8ad4d6ddSJeff Bonwick (void) ldi_close(lh, spa_mode(spa), kcred); 115afefbcddSeschrock } else { 116afefbcddSeschrock kmem_free(buf, len); 117afefbcddSeschrock } 118afefbcddSeschrock } 119fa9e4066Sahrens 120*8ad4d6ddSJeff Bonwick error = ldi_open_by_name(vd->vdev_path, spa_mode(spa), kcred, 121afefbcddSeschrock &dvd->vd_lh, zfs_li); 122fa9e4066Sahrens 123fa9e4066Sahrens /* 124fa9e4066Sahrens * Compare the devid to the stored value. 125fa9e4066Sahrens */ 126fa9e4066Sahrens if (error == 0 && vd->vdev_devid != NULL && 127fa9e4066Sahrens ldi_get_devid(dvd->vd_lh, &devid) == 0) { 128fa9e4066Sahrens if (ddi_devid_compare(devid, dvd->vd_devid) != 0) { 129fa9e4066Sahrens error = EINVAL; 130*8ad4d6ddSJeff Bonwick (void) ldi_close(dvd->vd_lh, spa_mode(spa), 131*8ad4d6ddSJeff Bonwick kcred); 132fa9e4066Sahrens dvd->vd_lh = NULL; 133fa9e4066Sahrens } 134fa9e4066Sahrens ddi_devid_free(devid); 135fa9e4066Sahrens } 136afefbcddSeschrock 137afefbcddSeschrock /* 138afefbcddSeschrock * If we succeeded in opening the device, but 'vdev_wholedisk' 139afefbcddSeschrock * is not yet set, then this must be a slice. 140afefbcddSeschrock */ 141afefbcddSeschrock if (error == 0 && vd->vdev_wholedisk == -1ULL) 142afefbcddSeschrock vd->vdev_wholedisk = 0; 143fa9e4066Sahrens } 144fa9e4066Sahrens 145fa9e4066Sahrens /* 146fa9e4066Sahrens * If we were unable to open by path, or the devid check fails, open by 147fa9e4066Sahrens * devid instead. 148fa9e4066Sahrens */ 149fa9e4066Sahrens if (error != 0 && vd->vdev_devid != NULL) 150fa9e4066Sahrens error = ldi_open_by_devid(dvd->vd_devid, dvd->vd_minor, 151*8ad4d6ddSJeff Bonwick spa_mode(spa), kcred, &dvd->vd_lh, zfs_li); 152fa9e4066Sahrens 1533d7072f8Seschrock /* 1543d7072f8Seschrock * If all else fails, then try opening by physical path (if available) 1553d7072f8Seschrock * or the logical path (if we failed due to the devid check). While not 1563d7072f8Seschrock * as reliable as the devid, this will give us something, and the higher 1573d7072f8Seschrock * level vdev validation will prevent us from opening the wrong device. 1583d7072f8Seschrock */ 1593d7072f8Seschrock if (error) { 1603d7072f8Seschrock if (vd->vdev_physpath != NULL && 1613d7072f8Seschrock (dev = ddi_pathname_to_dev_t(vd->vdev_physpath)) != ENODEV) 162*8ad4d6ddSJeff Bonwick error = ldi_open_by_dev(&dev, OTYP_BLK, spa_mode(spa), 1633d7072f8Seschrock kcred, &dvd->vd_lh, zfs_li); 1643d7072f8Seschrock 1653d7072f8Seschrock /* 1663d7072f8Seschrock * Note that we don't support the legacy auto-wholedisk support 1673d7072f8Seschrock * as above. This hasn't been used in a very long time and we 1683d7072f8Seschrock * don't need to propagate its oddities to this edge condition. 1693d7072f8Seschrock */ 170*8ad4d6ddSJeff Bonwick if (error && vd->vdev_path != NULL && !spa_is_root(spa)) 171*8ad4d6ddSJeff Bonwick error = ldi_open_by_name(vd->vdev_path, spa_mode(spa), 172*8ad4d6ddSJeff Bonwick kcred, &dvd->vd_lh, zfs_li); 1733d7072f8Seschrock } 1743d7072f8Seschrock 175e14bb325SJeff Bonwick if (error) { 176fa9e4066Sahrens vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED; 177fa9e4066Sahrens return (error); 178e14bb325SJeff Bonwick } 179fa9e4066Sahrens 1803d7072f8Seschrock /* 1813d7072f8Seschrock * Once a device is opened, verify that the physical device path (if 1823d7072f8Seschrock * available) is up to date. 1833d7072f8Seschrock */ 1843d7072f8Seschrock if (ldi_get_dev(dvd->vd_lh, &dev) == 0 && 1853d7072f8Seschrock ldi_get_otyp(dvd->vd_lh, &otyp) == 0) { 1860a4e9518Sgw char *physpath, *minorname; 1870a4e9518Sgw 1883d7072f8Seschrock physpath = kmem_alloc(MAXPATHLEN, KM_SLEEP); 1893d7072f8Seschrock minorname = NULL; 1903d7072f8Seschrock if (ddi_dev_pathname(dev, otyp, physpath) == 0 && 1913d7072f8Seschrock ldi_get_minor_name(dvd->vd_lh, &minorname) == 0 && 1923d7072f8Seschrock (vd->vdev_physpath == NULL || 1933d7072f8Seschrock strcmp(vd->vdev_physpath, physpath) != 0)) { 1943d7072f8Seschrock if (vd->vdev_physpath) 1953d7072f8Seschrock spa_strfree(vd->vdev_physpath); 1963d7072f8Seschrock (void) strlcat(physpath, ":", MAXPATHLEN); 1973d7072f8Seschrock (void) strlcat(physpath, minorname, MAXPATHLEN); 1983d7072f8Seschrock vd->vdev_physpath = spa_strdup(physpath); 1993d7072f8Seschrock } 2003d7072f8Seschrock if (minorname) 2013d7072f8Seschrock kmem_free(minorname, strlen(minorname) + 1); 2023d7072f8Seschrock kmem_free(physpath, MAXPATHLEN); 2033d7072f8Seschrock } 2043d7072f8Seschrock 205fa9e4066Sahrens /* 206fa9e4066Sahrens * Determine the actual size of the device. 207fa9e4066Sahrens */ 208fa9e4066Sahrens if (ldi_get_size(dvd->vd_lh, psize) != 0) { 209fa9e4066Sahrens vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED; 210fa9e4066Sahrens return (EINVAL); 211fa9e4066Sahrens } 212fa9e4066Sahrens 213ecc2d604Sbonwick /* 214ecc2d604Sbonwick * If we own the whole disk, try to enable disk write caching. 215ecc2d604Sbonwick * We ignore errors because it's OK if we can't do it. 216ecc2d604Sbonwick */ 217bef6b7d2Swebaker if (vd->vdev_wholedisk == 1) { 218ecc2d604Sbonwick int wce = 1; 219ecc2d604Sbonwick (void) ldi_ioctl(dvd->vd_lh, DKIOCSETWCE, (intptr_t)&wce, 220ecc2d604Sbonwick FKIOCTL, kcred, NULL); 221ecc2d604Sbonwick } 222bef6b7d2Swebaker 223ecc2d604Sbonwick /* 224ecc2d604Sbonwick * Determine the device's minimum transfer size. 225ecc2d604Sbonwick * If the ioctl isn't supported, assume DEV_BSIZE. 226ecc2d604Sbonwick */ 227ecc2d604Sbonwick if (ldi_ioctl(dvd->vd_lh, DKIOCGMEDIAINFO, (intptr_t)&dkm, 228ecc2d604Sbonwick FKIOCTL, kcred, NULL) != 0) 229ecc2d604Sbonwick dkm.dki_lbsize = DEV_BSIZE; 230bef6b7d2Swebaker 231ecc2d604Sbonwick *ashift = highbit(MAX(dkm.dki_lbsize, SPA_MINBLOCKSIZE)) - 1; 232bef6b7d2Swebaker 233b468a217Seschrock /* 234b468a217Seschrock * Clear the nowritecache bit, so that on a vdev_reopen() we will 235b468a217Seschrock * try again. 236b468a217Seschrock */ 237b468a217Seschrock vd->vdev_nowritecache = B_FALSE; 238b468a217Seschrock 239fa9e4066Sahrens return (0); 240fa9e4066Sahrens } 241fa9e4066Sahrens 242fa9e4066Sahrens static void 243fa9e4066Sahrens vdev_disk_close(vdev_t *vd) 244fa9e4066Sahrens { 245fa9e4066Sahrens vdev_disk_t *dvd = vd->vdev_tsd; 246fa9e4066Sahrens 247fa9e4066Sahrens if (dvd == NULL) 248fa9e4066Sahrens return; 249fa9e4066Sahrens 250fa9e4066Sahrens if (dvd->vd_minor != NULL) 251fa9e4066Sahrens ddi_devid_str_free(dvd->vd_minor); 252fa9e4066Sahrens 253fa9e4066Sahrens if (dvd->vd_devid != NULL) 254fa9e4066Sahrens ddi_devid_free(dvd->vd_devid); 255fa9e4066Sahrens 256fa9e4066Sahrens if (dvd->vd_lh != NULL) 257*8ad4d6ddSJeff Bonwick (void) ldi_close(dvd->vd_lh, spa_mode(vd->vdev_spa), kcred); 258fa9e4066Sahrens 259fa9e4066Sahrens kmem_free(dvd, sizeof (vdev_disk_t)); 260fa9e4066Sahrens vd->vdev_tsd = NULL; 261fa9e4066Sahrens } 262fa9e4066Sahrens 263e7cbe64fSgw int 264e7cbe64fSgw vdev_disk_physio(ldi_handle_t vd_lh, caddr_t data, size_t size, 265e7cbe64fSgw uint64_t offset, int flags) 266e7cbe64fSgw { 267e7cbe64fSgw buf_t *bp; 268e7cbe64fSgw int error = 0; 269e7cbe64fSgw 270e7cbe64fSgw if (vd_lh == NULL) 271e7cbe64fSgw return (EINVAL); 272e7cbe64fSgw 273e7cbe64fSgw ASSERT(flags & B_READ || flags & B_WRITE); 274e7cbe64fSgw 275e7cbe64fSgw bp = getrbuf(KM_SLEEP); 276e7cbe64fSgw bp->b_flags = flags | B_BUSY | B_NOCACHE | B_FAILFAST; 277e7cbe64fSgw bp->b_bcount = size; 278e7cbe64fSgw bp->b_un.b_addr = (void *)data; 279e7cbe64fSgw bp->b_lblkno = lbtodb(offset); 280e7cbe64fSgw bp->b_bufsize = size; 281e7cbe64fSgw 282e7cbe64fSgw error = ldi_strategy(vd_lh, bp); 283e7cbe64fSgw ASSERT(error == 0); 284e7cbe64fSgw if ((error = biowait(bp)) == 0 && bp->b_resid != 0) 285e7cbe64fSgw error = EIO; 286e7cbe64fSgw freerbuf(bp); 287e7cbe64fSgw 288e7cbe64fSgw return (error); 289e7cbe64fSgw } 290e7cbe64fSgw 291fa9e4066Sahrens static void 292fa9e4066Sahrens vdev_disk_io_intr(buf_t *bp) 293fa9e4066Sahrens { 294fa9e4066Sahrens vdev_disk_buf_t *vdb = (vdev_disk_buf_t *)bp; 295fa9e4066Sahrens zio_t *zio = vdb->vdb_io; 296fa9e4066Sahrens 29751ece835Seschrock /* 29851ece835Seschrock * The rest of the zio stack only deals with EIO, ECKSUM, and ENXIO. 29951ece835Seschrock * Rather than teach the rest of the stack about other error 30051ece835Seschrock * possibilities (EFAULT, etc), we normalize the error value here. 30151ece835Seschrock */ 30251ece835Seschrock zio->io_error = (geterror(bp) != 0 ? EIO : 0); 30351ece835Seschrock 30451ece835Seschrock if (zio->io_error == 0 && bp->b_resid != 0) 305fa9e4066Sahrens zio->io_error = EIO; 306fa9e4066Sahrens 307fa9e4066Sahrens kmem_free(vdb, sizeof (vdev_disk_buf_t)); 308fa9e4066Sahrens 309e05725b1Sbonwick zio_interrupt(zio); 310fa9e4066Sahrens } 311fa9e4066Sahrens 312f4a72450SJeff Bonwick static void 313f4a72450SJeff Bonwick vdev_disk_ioctl_free(zio_t *zio) 314f4a72450SJeff Bonwick { 315f4a72450SJeff Bonwick kmem_free(zio->io_vsd, sizeof (struct dk_callback)); 316f4a72450SJeff Bonwick } 317f4a72450SJeff Bonwick 318fa9e4066Sahrens static void 319fa9e4066Sahrens vdev_disk_ioctl_done(void *zio_arg, int error) 320fa9e4066Sahrens { 321fa9e4066Sahrens zio_t *zio = zio_arg; 322fa9e4066Sahrens 323fa9e4066Sahrens zio->io_error = error; 324fa9e4066Sahrens 325e05725b1Sbonwick zio_interrupt(zio); 326fa9e4066Sahrens } 327fa9e4066Sahrens 328e05725b1Sbonwick static int 329fa9e4066Sahrens vdev_disk_io_start(zio_t *zio) 330fa9e4066Sahrens { 331fa9e4066Sahrens vdev_t *vd = zio->io_vd; 332fa9e4066Sahrens vdev_disk_t *dvd = vd->vdev_tsd; 333fa9e4066Sahrens vdev_disk_buf_t *vdb; 334e14bb325SJeff Bonwick struct dk_callback *dkc; 335fa9e4066Sahrens buf_t *bp; 336e14bb325SJeff Bonwick int error; 337fa9e4066Sahrens 338fa9e4066Sahrens if (zio->io_type == ZIO_TYPE_IOCTL) { 339fa9e4066Sahrens /* XXPOLICY */ 3400a4e9518Sgw if (!vdev_readable(vd)) { 341fa9e4066Sahrens zio->io_error = ENXIO; 342e05725b1Sbonwick return (ZIO_PIPELINE_CONTINUE); 343fa9e4066Sahrens } 344fa9e4066Sahrens 345fa9e4066Sahrens switch (zio->io_cmd) { 346fa9e4066Sahrens 347fa9e4066Sahrens case DKIOCFLUSHWRITECACHE: 348fa9e4066Sahrens 349a2eea2e1Sahrens if (zfs_nocacheflush) 350a2eea2e1Sahrens break; 351a2eea2e1Sahrens 352b468a217Seschrock if (vd->vdev_nowritecache) { 353b468a217Seschrock zio->io_error = ENOTSUP; 354b468a217Seschrock break; 355b468a217Seschrock } 356b468a217Seschrock 357e14bb325SJeff Bonwick zio->io_vsd = dkc = kmem_alloc(sizeof (*dkc), KM_SLEEP); 358f4a72450SJeff Bonwick zio->io_vsd_free = vdev_disk_ioctl_free; 359e14bb325SJeff Bonwick 360e14bb325SJeff Bonwick dkc->dkc_callback = vdev_disk_ioctl_done; 361e14bb325SJeff Bonwick dkc->dkc_flag = FLUSH_VOLATILE; 362e14bb325SJeff Bonwick dkc->dkc_cookie = zio; 363fa9e4066Sahrens 364fa9e4066Sahrens error = ldi_ioctl(dvd->vd_lh, zio->io_cmd, 365e14bb325SJeff Bonwick (uintptr_t)dkc, FKIOCTL, kcred, NULL); 366fa9e4066Sahrens 367fa9e4066Sahrens if (error == 0) { 368fa9e4066Sahrens /* 369fa9e4066Sahrens * The ioctl will be done asychronously, 370fa9e4066Sahrens * and will call vdev_disk_ioctl_done() 371fa9e4066Sahrens * upon completion. 372fa9e4066Sahrens */ 373e05725b1Sbonwick return (ZIO_PIPELINE_STOP); 374e05725b1Sbonwick } 375e05725b1Sbonwick 376e05725b1Sbonwick if (error == ENOTSUP || error == ENOTTY) { 377b468a217Seschrock /* 378d5782879Smishra * If we get ENOTSUP or ENOTTY, we know that 379d5782879Smishra * no future attempts will ever succeed. 380d5782879Smishra * In this case we set a persistent bit so 381d5782879Smishra * that we don't bother with the ioctl in the 382d5782879Smishra * future. 383b468a217Seschrock */ 384b468a217Seschrock vd->vdev_nowritecache = B_TRUE; 385fa9e4066Sahrens } 386fa9e4066Sahrens zio->io_error = error; 387b468a217Seschrock 388fa9e4066Sahrens break; 389fa9e4066Sahrens 390fa9e4066Sahrens default: 391fa9e4066Sahrens zio->io_error = ENOTSUP; 392fa9e4066Sahrens } 393fa9e4066Sahrens 394e05725b1Sbonwick return (ZIO_PIPELINE_CONTINUE); 395fa9e4066Sahrens } 396fa9e4066Sahrens 397fa9e4066Sahrens vdb = kmem_alloc(sizeof (vdev_disk_buf_t), KM_SLEEP); 398fa9e4066Sahrens 399fa9e4066Sahrens vdb->vdb_io = zio; 400fa9e4066Sahrens bp = &vdb->vdb_buf; 401fa9e4066Sahrens 402fa9e4066Sahrens bioinit(bp); 403e14bb325SJeff Bonwick bp->b_flags = B_BUSY | B_NOCACHE | 404e14bb325SJeff Bonwick (zio->io_type == ZIO_TYPE_READ ? B_READ : B_WRITE) | 405e14bb325SJeff Bonwick ((zio->io_flags & ZIO_FLAG_IO_RETRY) ? 0 : B_FAILFAST); 406fa9e4066Sahrens bp->b_bcount = zio->io_size; 407fa9e4066Sahrens bp->b_un.b_addr = zio->io_data; 408fa9e4066Sahrens bp->b_lblkno = lbtodb(zio->io_offset); 409fa9e4066Sahrens bp->b_bufsize = zio->io_size; 410fa9e4066Sahrens bp->b_iodone = (int (*)())vdev_disk_io_intr; 411fa9e4066Sahrens 412fa9e4066Sahrens /* ldi_strategy() will return non-zero only on programming errors */ 413e14bb325SJeff Bonwick VERIFY(ldi_strategy(dvd->vd_lh, bp) == 0); 414e05725b1Sbonwick 415e05725b1Sbonwick return (ZIO_PIPELINE_STOP); 416fa9e4066Sahrens } 417fa9e4066Sahrens 418e14bb325SJeff Bonwick static void 419fa9e4066Sahrens vdev_disk_io_done(zio_t *zio) 420fa9e4066Sahrens { 421e14bb325SJeff Bonwick vdev_t *vd = zio->io_vd; 422ea8dc4b6Seschrock 4233d7072f8Seschrock /* 4243d7072f8Seschrock * If the device returned EIO, then attempt a DKIOCSTATE ioctl to see if 4253d7072f8Seschrock * the device has been removed. If this is the case, then we trigger an 4260a4e9518Sgw * asynchronous removal of the device. Otherwise, probe the device and 4271f7ad2e1Sgw * make sure it's still accessible. 4283d7072f8Seschrock */ 4293d7072f8Seschrock if (zio->io_error == EIO) { 4300a4e9518Sgw vdev_disk_t *dvd = vd->vdev_tsd; 431e14bb325SJeff Bonwick int state = DKIO_NONE; 4320a4e9518Sgw 433e14bb325SJeff Bonwick if (ldi_ioctl(dvd->vd_lh, DKIOCSTATE, (intptr_t)&state, 434e14bb325SJeff Bonwick FKIOCTL, kcred, NULL) == 0 && state != DKIO_INSERTED) { 4353d7072f8Seschrock vd->vdev_remove_wanted = B_TRUE; 4363d7072f8Seschrock spa_async_request(zio->io_spa, SPA_ASYNC_REMOVE); 4373d7072f8Seschrock } 4383d7072f8Seschrock } 439fa9e4066Sahrens } 440fa9e4066Sahrens 441fa9e4066Sahrens vdev_ops_t vdev_disk_ops = { 442fa9e4066Sahrens vdev_disk_open, 443fa9e4066Sahrens vdev_disk_close, 444fa9e4066Sahrens vdev_default_asize, 445fa9e4066Sahrens vdev_disk_io_start, 446fa9e4066Sahrens vdev_disk_io_done, 447fa9e4066Sahrens NULL, 448fa9e4066Sahrens VDEV_TYPE_DISK, /* name of this vdev type */ 449fa9e4066Sahrens B_TRUE /* leaf vdev */ 450fa9e4066Sahrens }; 451e7cbe64fSgw 452e7cbe64fSgw /* 453051aabe6Staylor * Given the root disk device devid or pathname, read the label from 454051aabe6Staylor * the device, and construct a configuration nvlist. 455e7cbe64fSgw */ 456f940fbb1SLin Ling int 457f940fbb1SLin Ling vdev_disk_read_rootlabel(char *devpath, char *devid, nvlist_t **config) 458e7cbe64fSgw { 459e7cbe64fSgw ldi_handle_t vd_lh; 460e7cbe64fSgw vdev_label_t *label; 461e7cbe64fSgw uint64_t s, size; 462e7cbe64fSgw int l; 463051aabe6Staylor ddi_devid_t tmpdevid; 464f4565e39SLin Ling int error = -1; 465051aabe6Staylor char *minor_name; 466e7cbe64fSgw 467e7cbe64fSgw /* 468e7cbe64fSgw * Read the device label and build the nvlist. 469e7cbe64fSgw */ 470f4565e39SLin Ling if (devid != NULL && ddi_devid_str_decode(devid, &tmpdevid, 471051aabe6Staylor &minor_name) == 0) { 472051aabe6Staylor error = ldi_open_by_devid(tmpdevid, minor_name, 473*8ad4d6ddSJeff Bonwick FREAD, kcred, &vd_lh, zfs_li); 474051aabe6Staylor ddi_devid_free(tmpdevid); 475051aabe6Staylor ddi_devid_str_free(minor_name); 476051aabe6Staylor } 477051aabe6Staylor 478f4565e39SLin Ling if (error && (error = ldi_open_by_name(devpath, FREAD, kcred, &vd_lh, 479f4565e39SLin Ling zfs_li))) 480f940fbb1SLin Ling return (error); 481e7cbe64fSgw 482bf82a41bSeschrock if (ldi_get_size(vd_lh, &s)) { 483bf82a41bSeschrock (void) ldi_close(vd_lh, FREAD, kcred); 484f940fbb1SLin Ling return (EIO); 485bf82a41bSeschrock } 486e7cbe64fSgw 487e7cbe64fSgw size = P2ALIGN_TYPED(s, sizeof (vdev_label_t), uint64_t); 488e7cbe64fSgw label = kmem_alloc(sizeof (vdev_label_t), KM_SLEEP); 489e7cbe64fSgw 490e7cbe64fSgw for (l = 0; l < VDEV_LABELS; l++) { 491e7cbe64fSgw uint64_t offset, state, txg = 0; 492e7cbe64fSgw 493e7cbe64fSgw /* read vdev label */ 494e7cbe64fSgw offset = vdev_label_offset(size, l, 0); 495e7cbe64fSgw if (vdev_disk_physio(vd_lh, (caddr_t)label, 496e7cbe64fSgw VDEV_SKIP_SIZE + VDEV_BOOT_HEADER_SIZE + 497e7cbe64fSgw VDEV_PHYS_SIZE, offset, B_READ) != 0) 498e7cbe64fSgw continue; 499e7cbe64fSgw 500e7cbe64fSgw if (nvlist_unpack(label->vl_vdev_phys.vp_nvlist, 501f940fbb1SLin Ling sizeof (label->vl_vdev_phys.vp_nvlist), config, 0) != 0) { 502f940fbb1SLin Ling *config = NULL; 503e7cbe64fSgw continue; 504e7cbe64fSgw } 505e7cbe64fSgw 506f940fbb1SLin Ling if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_STATE, 507e7cbe64fSgw &state) != 0 || state >= POOL_STATE_DESTROYED) { 508f940fbb1SLin Ling nvlist_free(*config); 509f940fbb1SLin Ling *config = NULL; 510e7cbe64fSgw continue; 511e7cbe64fSgw } 512e7cbe64fSgw 513f940fbb1SLin Ling if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_TXG, 514e7cbe64fSgw &txg) != 0 || txg == 0) { 515f940fbb1SLin Ling nvlist_free(*config); 516f940fbb1SLin Ling *config = NULL; 517e7cbe64fSgw continue; 518e7cbe64fSgw } 519e7cbe64fSgw 520e7cbe64fSgw break; 521e7cbe64fSgw } 522e7cbe64fSgw 523e7cbe64fSgw kmem_free(label, sizeof (vdev_label_t)); 524bf82a41bSeschrock (void) ldi_close(vd_lh, FREAD, kcred); 525bf82a41bSeschrock 526f940fbb1SLin Ling return (error); 527e7cbe64fSgw } 528