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 /* 22*dcba9f3fSGeorge Wilson * Copyright 2010 Sun Microsystems, Inc. All rights reserved. 23fa9e4066Sahrens * Use is subject to license terms. 24fa9e4066Sahrens */ 25fa9e4066Sahrens 26fa9e4066Sahrens #include <sys/zfs_context.h> 27*dcba9f3fSGeorge Wilson #include <sys/spa_impl.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 47*dcba9f3fSGeorge Wilson static void 48*dcba9f3fSGeorge Wilson vdev_disk_hold(vdev_t *vd) 49*dcba9f3fSGeorge Wilson { 50*dcba9f3fSGeorge Wilson ddi_devid_t devid; 51*dcba9f3fSGeorge Wilson char *minor; 52*dcba9f3fSGeorge Wilson 53*dcba9f3fSGeorge Wilson ASSERT(spa_config_held(vd->vdev_spa, SCL_STATE, RW_WRITER)); 54*dcba9f3fSGeorge Wilson 55*dcba9f3fSGeorge Wilson /* 56*dcba9f3fSGeorge Wilson * We must have a pathname, and it must be absolute. 57*dcba9f3fSGeorge Wilson */ 58*dcba9f3fSGeorge Wilson if (vd->vdev_path == NULL || vd->vdev_path[0] != '/') 59*dcba9f3fSGeorge Wilson return; 60*dcba9f3fSGeorge Wilson 61*dcba9f3fSGeorge Wilson /* 62*dcba9f3fSGeorge Wilson * Only prefetch path and devid info if the device has 63*dcba9f3fSGeorge Wilson * never been opened. 64*dcba9f3fSGeorge Wilson */ 65*dcba9f3fSGeorge Wilson if (vd->vdev_tsd != NULL) 66*dcba9f3fSGeorge Wilson return; 67*dcba9f3fSGeorge Wilson 68*dcba9f3fSGeorge Wilson if (vd->vdev_wholedisk == -1ULL) { 69*dcba9f3fSGeorge Wilson size_t len = strlen(vd->vdev_path) + 3; 70*dcba9f3fSGeorge Wilson char *buf = kmem_alloc(len, KM_SLEEP); 71*dcba9f3fSGeorge Wilson 72*dcba9f3fSGeorge Wilson (void) snprintf(buf, len, "%ss0", vd->vdev_path); 73*dcba9f3fSGeorge Wilson 74*dcba9f3fSGeorge Wilson (void) ldi_vp_from_name(buf, &vd->vdev_name_vp); 75*dcba9f3fSGeorge Wilson kmem_free(buf, len); 76*dcba9f3fSGeorge Wilson } 77*dcba9f3fSGeorge Wilson 78*dcba9f3fSGeorge Wilson if (vd->vdev_name_vp == NULL) 79*dcba9f3fSGeorge Wilson (void) ldi_vp_from_name(vd->vdev_path, &vd->vdev_name_vp); 80*dcba9f3fSGeorge Wilson 81*dcba9f3fSGeorge Wilson if (vd->vdev_devid != NULL && 82*dcba9f3fSGeorge Wilson ddi_devid_str_decode(vd->vdev_devid, &devid, &minor) == 0) { 83*dcba9f3fSGeorge Wilson (void) ldi_vp_from_devid(devid, minor, &vd->vdev_devid_vp); 84*dcba9f3fSGeorge Wilson ddi_devid_str_free(minor); 85*dcba9f3fSGeorge Wilson ddi_devid_free(devid); 86*dcba9f3fSGeorge Wilson } 87*dcba9f3fSGeorge Wilson } 88*dcba9f3fSGeorge Wilson 89*dcba9f3fSGeorge Wilson static void 90*dcba9f3fSGeorge Wilson vdev_disk_rele(vdev_t *vd) 91*dcba9f3fSGeorge Wilson { 92*dcba9f3fSGeorge Wilson ASSERT(spa_config_held(vd->vdev_spa, SCL_STATE, RW_WRITER)); 93*dcba9f3fSGeorge Wilson 94*dcba9f3fSGeorge Wilson if (vd->vdev_name_vp) { 95*dcba9f3fSGeorge Wilson VN_RELE_ASYNC(vd->vdev_name_vp, 96*dcba9f3fSGeorge Wilson dsl_pool_vnrele_taskq(vd->vdev_spa->spa_dsl_pool)); 97*dcba9f3fSGeorge Wilson vd->vdev_name_vp = NULL; 98*dcba9f3fSGeorge Wilson } 99*dcba9f3fSGeorge Wilson if (vd->vdev_devid_vp) { 100*dcba9f3fSGeorge Wilson VN_RELE_ASYNC(vd->vdev_devid_vp, 101*dcba9f3fSGeorge Wilson dsl_pool_vnrele_taskq(vd->vdev_spa->spa_dsl_pool)); 102*dcba9f3fSGeorge Wilson vd->vdev_devid_vp = NULL; 103*dcba9f3fSGeorge Wilson } 104*dcba9f3fSGeorge Wilson } 105*dcba9f3fSGeorge Wilson 106fa9e4066Sahrens static int 107e14bb325SJeff Bonwick vdev_disk_open(vdev_t *vd, uint64_t *psize, uint64_t *ashift) 108fa9e4066Sahrens { 1098ad4d6ddSJeff Bonwick spa_t *spa = vd->vdev_spa; 110fa9e4066Sahrens vdev_disk_t *dvd; 111e14bb325SJeff Bonwick struct dk_minfo dkm; 1120a4e9518Sgw int error; 113e14bb325SJeff Bonwick dev_t dev; 114e14bb325SJeff Bonwick int otyp; 115fa9e4066Sahrens 116fa9e4066Sahrens /* 117fa9e4066Sahrens * We must have a pathname, and it must be absolute. 118fa9e4066Sahrens */ 119fa9e4066Sahrens if (vd->vdev_path == NULL || vd->vdev_path[0] != '/') { 120fa9e4066Sahrens vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL; 121fa9e4066Sahrens return (EINVAL); 122fa9e4066Sahrens } 123fa9e4066Sahrens 124095bcd66SGeorge Wilson /* 125095bcd66SGeorge Wilson * Reopen the device if it's not currently open. Otherwise, 126095bcd66SGeorge Wilson * just update the physical size of the device. 127095bcd66SGeorge Wilson */ 128095bcd66SGeorge Wilson if (vd->vdev_tsd != NULL) { 129095bcd66SGeorge Wilson ASSERT(vd->vdev_reopening); 130095bcd66SGeorge Wilson dvd = vd->vdev_tsd; 131095bcd66SGeorge Wilson goto skip_open; 132095bcd66SGeorge Wilson } 133095bcd66SGeorge Wilson 134fa9e4066Sahrens dvd = vd->vdev_tsd = kmem_zalloc(sizeof (vdev_disk_t), KM_SLEEP); 135fa9e4066Sahrens 136fa9e4066Sahrens /* 137fa9e4066Sahrens * When opening a disk device, we want to preserve the user's original 138fa9e4066Sahrens * intent. We always want to open the device by the path the user gave 139fa9e4066Sahrens * us, even if it is one of multiple paths to the save device. But we 140fa9e4066Sahrens * also want to be able to survive disks being removed/recabled. 141fa9e4066Sahrens * Therefore the sequence of opening devices is: 142fa9e4066Sahrens * 143afefbcddSeschrock * 1. Try opening the device by path. For legacy pools without the 144afefbcddSeschrock * 'whole_disk' property, attempt to fix the path by appending 's0'. 145fa9e4066Sahrens * 146fa9e4066Sahrens * 2. If the devid of the device matches the stored value, return 147fa9e4066Sahrens * success. 148fa9e4066Sahrens * 149fa9e4066Sahrens * 3. Otherwise, the device may have moved. Try opening the device 150fa9e4066Sahrens * by the devid instead. 151fa9e4066Sahrens */ 152fa9e4066Sahrens if (vd->vdev_devid != NULL) { 153fa9e4066Sahrens if (ddi_devid_str_decode(vd->vdev_devid, &dvd->vd_devid, 154fa9e4066Sahrens &dvd->vd_minor) != 0) { 155fa9e4066Sahrens vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL; 156fa9e4066Sahrens return (EINVAL); 157fa9e4066Sahrens } 158fa9e4066Sahrens } 159fa9e4066Sahrens 160fa9e4066Sahrens error = EINVAL; /* presume failure */ 161fa9e4066Sahrens 162095bcd66SGeorge Wilson if (vd->vdev_path != NULL) { 163fa9e4066Sahrens ddi_devid_t devid; 164fa9e4066Sahrens 165afefbcddSeschrock if (vd->vdev_wholedisk == -1ULL) { 166afefbcddSeschrock size_t len = strlen(vd->vdev_path) + 3; 167afefbcddSeschrock char *buf = kmem_alloc(len, KM_SLEEP); 168afefbcddSeschrock ldi_handle_t lh; 169afefbcddSeschrock 170afefbcddSeschrock (void) snprintf(buf, len, "%ss0", vd->vdev_path); 171afefbcddSeschrock 1728ad4d6ddSJeff Bonwick if (ldi_open_by_name(buf, spa_mode(spa), kcred, 173afefbcddSeschrock &lh, zfs_li) == 0) { 174afefbcddSeschrock spa_strfree(vd->vdev_path); 175afefbcddSeschrock vd->vdev_path = buf; 176afefbcddSeschrock vd->vdev_wholedisk = 1ULL; 1778ad4d6ddSJeff Bonwick (void) ldi_close(lh, spa_mode(spa), kcred); 178afefbcddSeschrock } else { 179afefbcddSeschrock kmem_free(buf, len); 180afefbcddSeschrock } 181afefbcddSeschrock } 182fa9e4066Sahrens 1838ad4d6ddSJeff Bonwick error = ldi_open_by_name(vd->vdev_path, spa_mode(spa), kcred, 184afefbcddSeschrock &dvd->vd_lh, zfs_li); 185fa9e4066Sahrens 186fa9e4066Sahrens /* 187fa9e4066Sahrens * Compare the devid to the stored value. 188fa9e4066Sahrens */ 189fa9e4066Sahrens if (error == 0 && vd->vdev_devid != NULL && 190fa9e4066Sahrens ldi_get_devid(dvd->vd_lh, &devid) == 0) { 191fa9e4066Sahrens if (ddi_devid_compare(devid, dvd->vd_devid) != 0) { 192fa9e4066Sahrens error = EINVAL; 1938ad4d6ddSJeff Bonwick (void) ldi_close(dvd->vd_lh, spa_mode(spa), 1948ad4d6ddSJeff Bonwick kcred); 195fa9e4066Sahrens dvd->vd_lh = NULL; 196fa9e4066Sahrens } 197fa9e4066Sahrens ddi_devid_free(devid); 198fa9e4066Sahrens } 199afefbcddSeschrock 200afefbcddSeschrock /* 201afefbcddSeschrock * If we succeeded in opening the device, but 'vdev_wholedisk' 202afefbcddSeschrock * is not yet set, then this must be a slice. 203afefbcddSeschrock */ 204afefbcddSeschrock if (error == 0 && vd->vdev_wholedisk == -1ULL) 205afefbcddSeschrock vd->vdev_wholedisk = 0; 206fa9e4066Sahrens } 207fa9e4066Sahrens 208fa9e4066Sahrens /* 209fa9e4066Sahrens * If we were unable to open by path, or the devid check fails, open by 210fa9e4066Sahrens * devid instead. 211fa9e4066Sahrens */ 212fa9e4066Sahrens if (error != 0 && vd->vdev_devid != NULL) 213fa9e4066Sahrens error = ldi_open_by_devid(dvd->vd_devid, dvd->vd_minor, 2148ad4d6ddSJeff Bonwick spa_mode(spa), kcred, &dvd->vd_lh, zfs_li); 215fa9e4066Sahrens 2163d7072f8Seschrock /* 2173d7072f8Seschrock * If all else fails, then try opening by physical path (if available) 2183d7072f8Seschrock * or the logical path (if we failed due to the devid check). While not 2193d7072f8Seschrock * as reliable as the devid, this will give us something, and the higher 2203d7072f8Seschrock * level vdev validation will prevent us from opening the wrong device. 2213d7072f8Seschrock */ 2223d7072f8Seschrock if (error) { 2233d7072f8Seschrock if (vd->vdev_physpath != NULL && 224deb8317bSMark J Musante (dev = ddi_pathname_to_dev_t(vd->vdev_physpath)) != NODEV) 2258ad4d6ddSJeff Bonwick error = ldi_open_by_dev(&dev, OTYP_BLK, spa_mode(spa), 2263d7072f8Seschrock kcred, &dvd->vd_lh, zfs_li); 2273d7072f8Seschrock 2283d7072f8Seschrock /* 2293d7072f8Seschrock * Note that we don't support the legacy auto-wholedisk support 2303d7072f8Seschrock * as above. This hasn't been used in a very long time and we 2313d7072f8Seschrock * don't need to propagate its oddities to this edge condition. 2323d7072f8Seschrock */ 233095bcd66SGeorge Wilson if (error && vd->vdev_path != NULL) 2348ad4d6ddSJeff Bonwick error = ldi_open_by_name(vd->vdev_path, spa_mode(spa), 2358ad4d6ddSJeff Bonwick kcred, &dvd->vd_lh, zfs_li); 2363d7072f8Seschrock } 2373d7072f8Seschrock 238e14bb325SJeff Bonwick if (error) { 239fa9e4066Sahrens vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED; 240fa9e4066Sahrens return (error); 241e14bb325SJeff Bonwick } 242fa9e4066Sahrens 2433d7072f8Seschrock /* 2443d7072f8Seschrock * Once a device is opened, verify that the physical device path (if 2453d7072f8Seschrock * available) is up to date. 2463d7072f8Seschrock */ 2473d7072f8Seschrock if (ldi_get_dev(dvd->vd_lh, &dev) == 0 && 2483d7072f8Seschrock ldi_get_otyp(dvd->vd_lh, &otyp) == 0) { 2490a4e9518Sgw char *physpath, *minorname; 2500a4e9518Sgw 2513d7072f8Seschrock physpath = kmem_alloc(MAXPATHLEN, KM_SLEEP); 2523d7072f8Seschrock minorname = NULL; 2533d7072f8Seschrock if (ddi_dev_pathname(dev, otyp, physpath) == 0 && 2543d7072f8Seschrock ldi_get_minor_name(dvd->vd_lh, &minorname) == 0 && 2553d7072f8Seschrock (vd->vdev_physpath == NULL || 2563d7072f8Seschrock strcmp(vd->vdev_physpath, physpath) != 0)) { 2573d7072f8Seschrock if (vd->vdev_physpath) 2583d7072f8Seschrock spa_strfree(vd->vdev_physpath); 2593d7072f8Seschrock (void) strlcat(physpath, ":", MAXPATHLEN); 2603d7072f8Seschrock (void) strlcat(physpath, minorname, MAXPATHLEN); 2613d7072f8Seschrock vd->vdev_physpath = spa_strdup(physpath); 2623d7072f8Seschrock } 2633d7072f8Seschrock if (minorname) 2643d7072f8Seschrock kmem_free(minorname, strlen(minorname) + 1); 2653d7072f8Seschrock kmem_free(physpath, MAXPATHLEN); 2663d7072f8Seschrock } 2673d7072f8Seschrock 268095bcd66SGeorge Wilson skip_open: 269fa9e4066Sahrens /* 270fa9e4066Sahrens * Determine the actual size of the device. 271fa9e4066Sahrens */ 272fa9e4066Sahrens if (ldi_get_size(dvd->vd_lh, psize) != 0) { 273fa9e4066Sahrens vd->vdev_stat.vs_aux = VDEV_AUX_OPEN_FAILED; 274fa9e4066Sahrens return (EINVAL); 275fa9e4066Sahrens } 276fa9e4066Sahrens 277ecc2d604Sbonwick /* 278ecc2d604Sbonwick * If we own the whole disk, try to enable disk write caching. 279ecc2d604Sbonwick * We ignore errors because it's OK if we can't do it. 280ecc2d604Sbonwick */ 281bef6b7d2Swebaker if (vd->vdev_wholedisk == 1) { 282ecc2d604Sbonwick int wce = 1; 283ecc2d604Sbonwick (void) ldi_ioctl(dvd->vd_lh, DKIOCSETWCE, (intptr_t)&wce, 284ecc2d604Sbonwick FKIOCTL, kcred, NULL); 285ecc2d604Sbonwick } 286bef6b7d2Swebaker 287ecc2d604Sbonwick /* 288ecc2d604Sbonwick * Determine the device's minimum transfer size. 289ecc2d604Sbonwick * If the ioctl isn't supported, assume DEV_BSIZE. 290ecc2d604Sbonwick */ 291ecc2d604Sbonwick if (ldi_ioctl(dvd->vd_lh, DKIOCGMEDIAINFO, (intptr_t)&dkm, 292ecc2d604Sbonwick FKIOCTL, kcred, NULL) != 0) 293ecc2d604Sbonwick dkm.dki_lbsize = DEV_BSIZE; 294bef6b7d2Swebaker 295ecc2d604Sbonwick *ashift = highbit(MAX(dkm.dki_lbsize, SPA_MINBLOCKSIZE)) - 1; 296bef6b7d2Swebaker 297b468a217Seschrock /* 298b468a217Seschrock * Clear the nowritecache bit, so that on a vdev_reopen() we will 299b468a217Seschrock * try again. 300b468a217Seschrock */ 301b468a217Seschrock vd->vdev_nowritecache = B_FALSE; 302b468a217Seschrock 303fa9e4066Sahrens return (0); 304fa9e4066Sahrens } 305fa9e4066Sahrens 306fa9e4066Sahrens static void 307fa9e4066Sahrens vdev_disk_close(vdev_t *vd) 308fa9e4066Sahrens { 309fa9e4066Sahrens vdev_disk_t *dvd = vd->vdev_tsd; 310fa9e4066Sahrens 311095bcd66SGeorge Wilson if (vd->vdev_reopening || dvd == NULL) 312fa9e4066Sahrens return; 313fa9e4066Sahrens 314fa9e4066Sahrens if (dvd->vd_minor != NULL) 315fa9e4066Sahrens ddi_devid_str_free(dvd->vd_minor); 316fa9e4066Sahrens 317fa9e4066Sahrens if (dvd->vd_devid != NULL) 318fa9e4066Sahrens ddi_devid_free(dvd->vd_devid); 319fa9e4066Sahrens 320fa9e4066Sahrens if (dvd->vd_lh != NULL) 3218ad4d6ddSJeff Bonwick (void) ldi_close(dvd->vd_lh, spa_mode(vd->vdev_spa), kcred); 322fa9e4066Sahrens 323fa9e4066Sahrens kmem_free(dvd, sizeof (vdev_disk_t)); 324fa9e4066Sahrens vd->vdev_tsd = NULL; 325fa9e4066Sahrens } 326fa9e4066Sahrens 327e7cbe64fSgw int 328e7cbe64fSgw vdev_disk_physio(ldi_handle_t vd_lh, caddr_t data, size_t size, 329e7cbe64fSgw uint64_t offset, int flags) 330e7cbe64fSgw { 331e7cbe64fSgw buf_t *bp; 332e7cbe64fSgw int error = 0; 333e7cbe64fSgw 334e7cbe64fSgw if (vd_lh == NULL) 335e7cbe64fSgw return (EINVAL); 336e7cbe64fSgw 337e7cbe64fSgw ASSERT(flags & B_READ || flags & B_WRITE); 338e7cbe64fSgw 339e7cbe64fSgw bp = getrbuf(KM_SLEEP); 340e7cbe64fSgw bp->b_flags = flags | B_BUSY | B_NOCACHE | B_FAILFAST; 341e7cbe64fSgw bp->b_bcount = size; 342e7cbe64fSgw bp->b_un.b_addr = (void *)data; 343e7cbe64fSgw bp->b_lblkno = lbtodb(offset); 344e7cbe64fSgw bp->b_bufsize = size; 345e7cbe64fSgw 346e7cbe64fSgw error = ldi_strategy(vd_lh, bp); 347e7cbe64fSgw ASSERT(error == 0); 348e7cbe64fSgw if ((error = biowait(bp)) == 0 && bp->b_resid != 0) 349e7cbe64fSgw error = EIO; 350e7cbe64fSgw freerbuf(bp); 351e7cbe64fSgw 352e7cbe64fSgw return (error); 353e7cbe64fSgw } 354e7cbe64fSgw 355fa9e4066Sahrens static void 356fa9e4066Sahrens vdev_disk_io_intr(buf_t *bp) 357fa9e4066Sahrens { 358fa9e4066Sahrens vdev_disk_buf_t *vdb = (vdev_disk_buf_t *)bp; 359fa9e4066Sahrens zio_t *zio = vdb->vdb_io; 360fa9e4066Sahrens 36151ece835Seschrock /* 36251ece835Seschrock * The rest of the zio stack only deals with EIO, ECKSUM, and ENXIO. 36351ece835Seschrock * Rather than teach the rest of the stack about other error 36451ece835Seschrock * possibilities (EFAULT, etc), we normalize the error value here. 36551ece835Seschrock */ 36651ece835Seschrock zio->io_error = (geterror(bp) != 0 ? EIO : 0); 36751ece835Seschrock 36851ece835Seschrock if (zio->io_error == 0 && bp->b_resid != 0) 369fa9e4066Sahrens zio->io_error = EIO; 370fa9e4066Sahrens 371fa9e4066Sahrens kmem_free(vdb, sizeof (vdev_disk_buf_t)); 372fa9e4066Sahrens 373e05725b1Sbonwick zio_interrupt(zio); 374fa9e4066Sahrens } 375fa9e4066Sahrens 376f4a72450SJeff Bonwick static void 377f4a72450SJeff Bonwick vdev_disk_ioctl_free(zio_t *zio) 378f4a72450SJeff Bonwick { 379f4a72450SJeff Bonwick kmem_free(zio->io_vsd, sizeof (struct dk_callback)); 380f4a72450SJeff Bonwick } 381f4a72450SJeff Bonwick 38222fe2c88SJonathan Adams static const zio_vsd_ops_t vdev_disk_vsd_ops = { 38322fe2c88SJonathan Adams vdev_disk_ioctl_free, 38422fe2c88SJonathan Adams zio_vsd_default_cksum_report 38522fe2c88SJonathan Adams }; 38622fe2c88SJonathan Adams 387fa9e4066Sahrens static void 388fa9e4066Sahrens vdev_disk_ioctl_done(void *zio_arg, int error) 389fa9e4066Sahrens { 390fa9e4066Sahrens zio_t *zio = zio_arg; 391fa9e4066Sahrens 392fa9e4066Sahrens zio->io_error = error; 393fa9e4066Sahrens 394e05725b1Sbonwick zio_interrupt(zio); 395fa9e4066Sahrens } 396fa9e4066Sahrens 397e05725b1Sbonwick static int 398fa9e4066Sahrens vdev_disk_io_start(zio_t *zio) 399fa9e4066Sahrens { 400fa9e4066Sahrens vdev_t *vd = zio->io_vd; 401fa9e4066Sahrens vdev_disk_t *dvd = vd->vdev_tsd; 402fa9e4066Sahrens vdev_disk_buf_t *vdb; 403e14bb325SJeff Bonwick struct dk_callback *dkc; 404fa9e4066Sahrens buf_t *bp; 405e14bb325SJeff Bonwick int error; 406fa9e4066Sahrens 407fa9e4066Sahrens if (zio->io_type == ZIO_TYPE_IOCTL) { 408fa9e4066Sahrens /* XXPOLICY */ 4090a4e9518Sgw if (!vdev_readable(vd)) { 410fa9e4066Sahrens zio->io_error = ENXIO; 411e05725b1Sbonwick return (ZIO_PIPELINE_CONTINUE); 412fa9e4066Sahrens } 413fa9e4066Sahrens 414fa9e4066Sahrens switch (zio->io_cmd) { 415fa9e4066Sahrens 416fa9e4066Sahrens case DKIOCFLUSHWRITECACHE: 417fa9e4066Sahrens 418a2eea2e1Sahrens if (zfs_nocacheflush) 419a2eea2e1Sahrens break; 420a2eea2e1Sahrens 421b468a217Seschrock if (vd->vdev_nowritecache) { 422b468a217Seschrock zio->io_error = ENOTSUP; 423b468a217Seschrock break; 424b468a217Seschrock } 425b468a217Seschrock 426e14bb325SJeff Bonwick zio->io_vsd = dkc = kmem_alloc(sizeof (*dkc), KM_SLEEP); 42722fe2c88SJonathan Adams zio->io_vsd_ops = &vdev_disk_vsd_ops; 428e14bb325SJeff Bonwick 429e14bb325SJeff Bonwick dkc->dkc_callback = vdev_disk_ioctl_done; 430e14bb325SJeff Bonwick dkc->dkc_flag = FLUSH_VOLATILE; 431e14bb325SJeff Bonwick dkc->dkc_cookie = zio; 432fa9e4066Sahrens 433fa9e4066Sahrens error = ldi_ioctl(dvd->vd_lh, zio->io_cmd, 434e14bb325SJeff Bonwick (uintptr_t)dkc, FKIOCTL, kcred, NULL); 435fa9e4066Sahrens 436fa9e4066Sahrens if (error == 0) { 437fa9e4066Sahrens /* 438fa9e4066Sahrens * The ioctl will be done asychronously, 439fa9e4066Sahrens * and will call vdev_disk_ioctl_done() 440fa9e4066Sahrens * upon completion. 441fa9e4066Sahrens */ 442e05725b1Sbonwick return (ZIO_PIPELINE_STOP); 443e05725b1Sbonwick } 444e05725b1Sbonwick 445e05725b1Sbonwick if (error == ENOTSUP || error == ENOTTY) { 446b468a217Seschrock /* 447d5782879Smishra * If we get ENOTSUP or ENOTTY, we know that 448d5782879Smishra * no future attempts will ever succeed. 449d5782879Smishra * In this case we set a persistent bit so 450d5782879Smishra * that we don't bother with the ioctl in the 451d5782879Smishra * future. 452b468a217Seschrock */ 453b468a217Seschrock vd->vdev_nowritecache = B_TRUE; 454fa9e4066Sahrens } 455fa9e4066Sahrens zio->io_error = error; 456b468a217Seschrock 457fa9e4066Sahrens break; 458fa9e4066Sahrens 459fa9e4066Sahrens default: 460fa9e4066Sahrens zio->io_error = ENOTSUP; 461fa9e4066Sahrens } 462fa9e4066Sahrens 463e05725b1Sbonwick return (ZIO_PIPELINE_CONTINUE); 464fa9e4066Sahrens } 465fa9e4066Sahrens 466fa9e4066Sahrens vdb = kmem_alloc(sizeof (vdev_disk_buf_t), KM_SLEEP); 467fa9e4066Sahrens 468fa9e4066Sahrens vdb->vdb_io = zio; 469fa9e4066Sahrens bp = &vdb->vdb_buf; 470fa9e4066Sahrens 471fa9e4066Sahrens bioinit(bp); 472e14bb325SJeff Bonwick bp->b_flags = B_BUSY | B_NOCACHE | 4738956713aSEric Schrock (zio->io_type == ZIO_TYPE_READ ? B_READ : B_WRITE); 4748956713aSEric Schrock if (!(zio->io_flags & (ZIO_FLAG_IO_RETRY | ZIO_FLAG_TRYHARD))) 4758956713aSEric Schrock bp->b_flags |= B_FAILFAST; 476fa9e4066Sahrens bp->b_bcount = zio->io_size; 477fa9e4066Sahrens bp->b_un.b_addr = zio->io_data; 478fa9e4066Sahrens bp->b_lblkno = lbtodb(zio->io_offset); 479fa9e4066Sahrens bp->b_bufsize = zio->io_size; 480fa9e4066Sahrens bp->b_iodone = (int (*)())vdev_disk_io_intr; 481fa9e4066Sahrens 482fa9e4066Sahrens /* ldi_strategy() will return non-zero only on programming errors */ 483e14bb325SJeff Bonwick VERIFY(ldi_strategy(dvd->vd_lh, bp) == 0); 484e05725b1Sbonwick 485e05725b1Sbonwick return (ZIO_PIPELINE_STOP); 486fa9e4066Sahrens } 487fa9e4066Sahrens 488e14bb325SJeff Bonwick static void 489fa9e4066Sahrens vdev_disk_io_done(zio_t *zio) 490fa9e4066Sahrens { 491e14bb325SJeff Bonwick vdev_t *vd = zio->io_vd; 492ea8dc4b6Seschrock 4933d7072f8Seschrock /* 4943d7072f8Seschrock * If the device returned EIO, then attempt a DKIOCSTATE ioctl to see if 4953d7072f8Seschrock * the device has been removed. If this is the case, then we trigger an 4960a4e9518Sgw * asynchronous removal of the device. Otherwise, probe the device and 4971f7ad2e1Sgw * make sure it's still accessible. 4983d7072f8Seschrock */ 4991d713200SEric Schrock if (zio->io_error == EIO && !vd->vdev_remove_wanted) { 5000a4e9518Sgw vdev_disk_t *dvd = vd->vdev_tsd; 501e14bb325SJeff Bonwick int state = DKIO_NONE; 5020a4e9518Sgw 503e14bb325SJeff Bonwick if (ldi_ioctl(dvd->vd_lh, DKIOCSTATE, (intptr_t)&state, 504e14bb325SJeff Bonwick FKIOCTL, kcred, NULL) == 0 && state != DKIO_INSERTED) { 5051d713200SEric Schrock /* 5061d713200SEric Schrock * We post the resource as soon as possible, instead of 5071d713200SEric Schrock * when the async removal actually happens, because the 5081d713200SEric Schrock * DE is using this information to discard previous I/O 5091d713200SEric Schrock * errors. 5101d713200SEric Schrock */ 5111d713200SEric Schrock zfs_post_remove(zio->io_spa, vd); 5123d7072f8Seschrock vd->vdev_remove_wanted = B_TRUE; 5133d7072f8Seschrock spa_async_request(zio->io_spa, SPA_ASYNC_REMOVE); 5143d7072f8Seschrock } 5153d7072f8Seschrock } 516fa9e4066Sahrens } 517fa9e4066Sahrens 518fa9e4066Sahrens vdev_ops_t vdev_disk_ops = { 519fa9e4066Sahrens vdev_disk_open, 520fa9e4066Sahrens vdev_disk_close, 521fa9e4066Sahrens vdev_default_asize, 522fa9e4066Sahrens vdev_disk_io_start, 523fa9e4066Sahrens vdev_disk_io_done, 524fa9e4066Sahrens NULL, 525*dcba9f3fSGeorge Wilson vdev_disk_hold, 526*dcba9f3fSGeorge Wilson vdev_disk_rele, 527fa9e4066Sahrens VDEV_TYPE_DISK, /* name of this vdev type */ 528fa9e4066Sahrens B_TRUE /* leaf vdev */ 529fa9e4066Sahrens }; 530e7cbe64fSgw 531e7cbe64fSgw /* 532051aabe6Staylor * Given the root disk device devid or pathname, read the label from 533051aabe6Staylor * the device, and construct a configuration nvlist. 534e7cbe64fSgw */ 535f940fbb1SLin Ling int 536f940fbb1SLin Ling vdev_disk_read_rootlabel(char *devpath, char *devid, nvlist_t **config) 537e7cbe64fSgw { 538e7cbe64fSgw ldi_handle_t vd_lh; 539e7cbe64fSgw vdev_label_t *label; 540e7cbe64fSgw uint64_t s, size; 541e7cbe64fSgw int l; 542051aabe6Staylor ddi_devid_t tmpdevid; 543f4565e39SLin Ling int error = -1; 544051aabe6Staylor char *minor_name; 545e7cbe64fSgw 546e7cbe64fSgw /* 547e7cbe64fSgw * Read the device label and build the nvlist. 548e7cbe64fSgw */ 549f4565e39SLin Ling if (devid != NULL && ddi_devid_str_decode(devid, &tmpdevid, 550051aabe6Staylor &minor_name) == 0) { 551051aabe6Staylor error = ldi_open_by_devid(tmpdevid, minor_name, 5528ad4d6ddSJeff Bonwick FREAD, kcred, &vd_lh, zfs_li); 553051aabe6Staylor ddi_devid_free(tmpdevid); 554051aabe6Staylor ddi_devid_str_free(minor_name); 555051aabe6Staylor } 556051aabe6Staylor 557f4565e39SLin Ling if (error && (error = ldi_open_by_name(devpath, FREAD, kcred, &vd_lh, 558f4565e39SLin Ling zfs_li))) 559f940fbb1SLin Ling return (error); 560e7cbe64fSgw 561bf82a41bSeschrock if (ldi_get_size(vd_lh, &s)) { 562bf82a41bSeschrock (void) ldi_close(vd_lh, FREAD, kcred); 563f940fbb1SLin Ling return (EIO); 564bf82a41bSeschrock } 565e7cbe64fSgw 566e7cbe64fSgw size = P2ALIGN_TYPED(s, sizeof (vdev_label_t), uint64_t); 567e7cbe64fSgw label = kmem_alloc(sizeof (vdev_label_t), KM_SLEEP); 568e7cbe64fSgw 56917f1e64aSEric Taylor *config = NULL; 570e7cbe64fSgw for (l = 0; l < VDEV_LABELS; l++) { 571e7cbe64fSgw uint64_t offset, state, txg = 0; 572e7cbe64fSgw 573e7cbe64fSgw /* read vdev label */ 574e7cbe64fSgw offset = vdev_label_offset(size, l, 0); 575e7cbe64fSgw if (vdev_disk_physio(vd_lh, (caddr_t)label, 5762264ca7fSLin Ling VDEV_SKIP_SIZE + VDEV_PHYS_SIZE, offset, B_READ) != 0) 577e7cbe64fSgw continue; 578e7cbe64fSgw 579e7cbe64fSgw if (nvlist_unpack(label->vl_vdev_phys.vp_nvlist, 580f940fbb1SLin Ling sizeof (label->vl_vdev_phys.vp_nvlist), config, 0) != 0) { 581f940fbb1SLin Ling *config = NULL; 582e7cbe64fSgw continue; 583e7cbe64fSgw } 584e7cbe64fSgw 585f940fbb1SLin Ling if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_STATE, 586e7cbe64fSgw &state) != 0 || state >= POOL_STATE_DESTROYED) { 587f940fbb1SLin Ling nvlist_free(*config); 588f940fbb1SLin Ling *config = NULL; 589e7cbe64fSgw continue; 590e7cbe64fSgw } 591e7cbe64fSgw 592f940fbb1SLin Ling if (nvlist_lookup_uint64(*config, ZPOOL_CONFIG_POOL_TXG, 593e7cbe64fSgw &txg) != 0 || txg == 0) { 594f940fbb1SLin Ling nvlist_free(*config); 595f940fbb1SLin Ling *config = NULL; 596e7cbe64fSgw continue; 597e7cbe64fSgw } 598e7cbe64fSgw 599e7cbe64fSgw break; 600e7cbe64fSgw } 601e7cbe64fSgw 602e7cbe64fSgw kmem_free(label, sizeof (vdev_label_t)); 603bf82a41bSeschrock (void) ldi_close(vd_lh, FREAD, kcred); 60417f1e64aSEric Taylor if (*config == NULL) 60517f1e64aSEric Taylor error = EIDRM; 606bf82a41bSeschrock 607f940fbb1SLin Ling return (error); 608e7cbe64fSgw } 609