1fa9e4066Sahrens /* 2fa9e4066Sahrens * CDDL HEADER START 3fa9e4066Sahrens * 4fa9e4066Sahrens * The contents of this file are subject to the terms of the 5033f9833Sek * Common Development and Distribution License (the "License"). 6033f9833Sek * 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 /* 22dc7cd546SMark Shellenbaum * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 23fa9e4066Sahrens */ 24fa9e4066Sahrens 2555da60b9SMark J Musante /* Portions Copyright 2010 Robert Milkowski */ 2655da60b9SMark J Musante 27fa9e4066Sahrens #include <sys/types.h> 28fa9e4066Sahrens #include <sys/param.h> 29fa9e4066Sahrens #include <sys/systm.h> 30fa9e4066Sahrens #include <sys/sysmacros.h> 31fa9e4066Sahrens #include <sys/kmem.h> 32fa9e4066Sahrens #include <sys/pathname.h> 33fa9e4066Sahrens #include <sys/vnode.h> 34fa9e4066Sahrens #include <sys/vfs.h> 35aa59c4cbSrsb #include <sys/vfs_opreg.h> 36fa9e4066Sahrens #include <sys/mntent.h> 37fa9e4066Sahrens #include <sys/mount.h> 38fa9e4066Sahrens #include <sys/cmn_err.h> 39fa9e4066Sahrens #include "fs/fs_subr.h" 40fa9e4066Sahrens #include <sys/zfs_znode.h> 41893a6d32Sahrens #include <sys/zfs_dir.h> 42fa9e4066Sahrens #include <sys/zil.h> 43fa9e4066Sahrens #include <sys/fs/zfs.h> 44fa9e4066Sahrens #include <sys/dmu.h> 45fa9e4066Sahrens #include <sys/dsl_prop.h> 46b1b8ab34Slling #include <sys/dsl_dataset.h> 47ecd6cf80Smarks #include <sys/dsl_deleg.h> 48fa9e4066Sahrens #include <sys/spa.h> 49fa9e4066Sahrens #include <sys/zap.h> 500a586ceaSMark Shellenbaum #include <sys/sa.h> 51fa9e4066Sahrens #include <sys/varargs.h> 52fa9e4066Sahrens #include <sys/policy.h> 53fa9e4066Sahrens #include <sys/atomic.h> 54fa9e4066Sahrens #include <sys/mkdev.h> 55fa9e4066Sahrens #include <sys/modctl.h> 56ecd6cf80Smarks #include <sys/refstr.h> 57fa9e4066Sahrens #include <sys/zfs_ioctl.h> 58fa9e4066Sahrens #include <sys/zfs_ctldir.h> 59da6c28aaSamw #include <sys/zfs_fuid.h> 60ea8dc4b6Seschrock #include <sys/bootconf.h> 61a0965f35Sbonwick #include <sys/sunddi.h> 62033f9833Sek #include <sys/dnlc.h> 63f18faf3fSek #include <sys/dmu_objset.h> 64e7cbe64fSgw #include <sys/spa_boot.h> 650a586ceaSMark Shellenbaum #include <sys/sa.h> 660a586ceaSMark Shellenbaum #include "zfs_comutil.h" 67fa9e4066Sahrens 68fa9e4066Sahrens int zfsfstype; 69fa9e4066Sahrens vfsops_t *zfs_vfsops = NULL; 70a0965f35Sbonwick static major_t zfs_major; 71fa9e4066Sahrens static minor_t zfs_minor; 72fa9e4066Sahrens static kmutex_t zfs_dev_mtx; 73fa9e4066Sahrens 7454d692b7SGeorge Wilson extern int sys_shutdown; 7554d692b7SGeorge Wilson 76fa9e4066Sahrens static int zfs_mount(vfs_t *vfsp, vnode_t *mvp, struct mounta *uap, cred_t *cr); 77fa9e4066Sahrens static int zfs_umount(vfs_t *vfsp, int fflag, cred_t *cr); 78ea8dc4b6Seschrock static int zfs_mountroot(vfs_t *vfsp, enum whymountroot); 79fa9e4066Sahrens static int zfs_root(vfs_t *vfsp, vnode_t **vpp); 80fa9e4066Sahrens static int zfs_statvfs(vfs_t *vfsp, struct statvfs64 *statp); 81fa9e4066Sahrens static int zfs_vget(vfs_t *vfsp, vnode_t **vpp, fid_t *fidp); 82fa9e4066Sahrens static void zfs_freevfs(vfs_t *vfsp); 83fa9e4066Sahrens 84fa9e4066Sahrens static const fs_operation_def_t zfs_vfsops_template[] = { 85aa59c4cbSrsb VFSNAME_MOUNT, { .vfs_mount = zfs_mount }, 86aa59c4cbSrsb VFSNAME_MOUNTROOT, { .vfs_mountroot = zfs_mountroot }, 87aa59c4cbSrsb VFSNAME_UNMOUNT, { .vfs_unmount = zfs_umount }, 88aa59c4cbSrsb VFSNAME_ROOT, { .vfs_root = zfs_root }, 89aa59c4cbSrsb VFSNAME_STATVFS, { .vfs_statvfs = zfs_statvfs }, 90aa59c4cbSrsb VFSNAME_SYNC, { .vfs_sync = zfs_sync }, 91aa59c4cbSrsb VFSNAME_VGET, { .vfs_vget = zfs_vget }, 92aa59c4cbSrsb VFSNAME_FREEVFS, { .vfs_freevfs = zfs_freevfs }, 93aa59c4cbSrsb NULL, NULL 94fa9e4066Sahrens }; 95fa9e4066Sahrens 96fa9e4066Sahrens static const fs_operation_def_t zfs_vfsops_eio_template[] = { 97aa59c4cbSrsb VFSNAME_FREEVFS, { .vfs_freevfs = zfs_freevfs }, 98aa59c4cbSrsb NULL, NULL 99fa9e4066Sahrens }; 100fa9e4066Sahrens 101fa9e4066Sahrens /* 102fa9e4066Sahrens * We need to keep a count of active fs's. 103fa9e4066Sahrens * This is necessary to prevent our module 104fa9e4066Sahrens * from being unloaded after a umount -f 105fa9e4066Sahrens */ 106fa9e4066Sahrens static uint32_t zfs_active_fs_count = 0; 107fa9e4066Sahrens 108fa9e4066Sahrens static char *noatime_cancel[] = { MNTOPT_ATIME, NULL }; 109fa9e4066Sahrens static char *atime_cancel[] = { MNTOPT_NOATIME, NULL }; 1107b55fa8eSck static char *noxattr_cancel[] = { MNTOPT_XATTR, NULL }; 1117b55fa8eSck static char *xattr_cancel[] = { MNTOPT_NOXATTR, NULL }; 112fa9e4066Sahrens 1137b55fa8eSck /* 114b510d378Slling * MO_DEFAULT is not used since the default value is determined 115b510d378Slling * by the equivalent property. 1167b55fa8eSck */ 117fa9e4066Sahrens static mntopt_t mntopts[] = { 1187b55fa8eSck { MNTOPT_NOXATTR, noxattr_cancel, NULL, 0, NULL }, 1197b55fa8eSck { MNTOPT_XATTR, xattr_cancel, NULL, 0, NULL }, 120b510d378Slling { MNTOPT_NOATIME, noatime_cancel, NULL, 0, NULL }, 121fa9e4066Sahrens { MNTOPT_ATIME, atime_cancel, NULL, 0, NULL } 122fa9e4066Sahrens }; 123fa9e4066Sahrens 124fa9e4066Sahrens static mntopts_t zfs_mntopts = { 125fa9e4066Sahrens sizeof (mntopts) / sizeof (mntopt_t), 126fa9e4066Sahrens mntopts 127fa9e4066Sahrens }; 128fa9e4066Sahrens 129fa9e4066Sahrens /*ARGSUSED*/ 130fa9e4066Sahrens int 131fa9e4066Sahrens zfs_sync(vfs_t *vfsp, short flag, cred_t *cr) 132fa9e4066Sahrens { 133fa9e4066Sahrens /* 134fa9e4066Sahrens * Data integrity is job one. We don't want a compromised kernel 135fa9e4066Sahrens * writing to the storage pool, so we never sync during panic. 136fa9e4066Sahrens */ 137fa9e4066Sahrens if (panicstr) 138fa9e4066Sahrens return (0); 139fa9e4066Sahrens 140fa9e4066Sahrens /* 141fa9e4066Sahrens * SYNC_ATTR is used by fsflush() to force old filesystems like UFS 142fa9e4066Sahrens * to sync metadata, which they would otherwise cache indefinitely. 143fa9e4066Sahrens * Semantically, the only requirement is that the sync be initiated. 144fa9e4066Sahrens * The DMU syncs out txgs frequently, so there's nothing to do. 145fa9e4066Sahrens */ 146fa9e4066Sahrens if (flag & SYNC_ATTR) 147fa9e4066Sahrens return (0); 148fa9e4066Sahrens 149fa9e4066Sahrens if (vfsp != NULL) { 150fa9e4066Sahrens /* 151fa9e4066Sahrens * Sync a specific filesystem. 152fa9e4066Sahrens */ 153fa9e4066Sahrens zfsvfs_t *zfsvfs = vfsp->vfs_data; 15454d692b7SGeorge Wilson dsl_pool_t *dp; 155fa9e4066Sahrens 156fa9e4066Sahrens ZFS_ENTER(zfsvfs); 15754d692b7SGeorge Wilson dp = dmu_objset_pool(zfsvfs->z_os); 15854d692b7SGeorge Wilson 15954d692b7SGeorge Wilson /* 16054d692b7SGeorge Wilson * If the system is shutting down, then skip any 16154d692b7SGeorge Wilson * filesystems which may exist on a suspended pool. 16254d692b7SGeorge Wilson */ 16354d692b7SGeorge Wilson if (sys_shutdown && spa_suspended(dp->dp_spa)) { 16454d692b7SGeorge Wilson ZFS_EXIT(zfsvfs); 16554d692b7SGeorge Wilson return (0); 16654d692b7SGeorge Wilson } 16754d692b7SGeorge Wilson 168fa9e4066Sahrens if (zfsvfs->z_log != NULL) 1695002558fSNeil Perrin zil_commit(zfsvfs->z_log, 0); 17055da60b9SMark J Musante 171fa9e4066Sahrens ZFS_EXIT(zfsvfs); 172fa9e4066Sahrens } else { 173fa9e4066Sahrens /* 174fa9e4066Sahrens * Sync all ZFS filesystems. This is what happens when you 175fa9e4066Sahrens * run sync(1M). Unlike other filesystems, ZFS honors the 176fa9e4066Sahrens * request by waiting for all pools to commit all dirty data. 177fa9e4066Sahrens */ 178fa9e4066Sahrens spa_sync_allpools(); 179fa9e4066Sahrens } 180fa9e4066Sahrens 181fa9e4066Sahrens return (0); 182fa9e4066Sahrens } 183fa9e4066Sahrens 184ea8dc4b6Seschrock static int 185ea8dc4b6Seschrock zfs_create_unique_device(dev_t *dev) 186ea8dc4b6Seschrock { 187ea8dc4b6Seschrock major_t new_major; 188ea8dc4b6Seschrock 189ea8dc4b6Seschrock do { 190ea8dc4b6Seschrock ASSERT3U(zfs_minor, <=, MAXMIN32); 191ea8dc4b6Seschrock minor_t start = zfs_minor; 192ea8dc4b6Seschrock do { 193ea8dc4b6Seschrock mutex_enter(&zfs_dev_mtx); 194ea8dc4b6Seschrock if (zfs_minor >= MAXMIN32) { 195ea8dc4b6Seschrock /* 196ea8dc4b6Seschrock * If we're still using the real major 197ea8dc4b6Seschrock * keep out of /dev/zfs and /dev/zvol minor 198ea8dc4b6Seschrock * number space. If we're using a getudev()'ed 199ea8dc4b6Seschrock * major number, we can use all of its minors. 200ea8dc4b6Seschrock */ 201ea8dc4b6Seschrock if (zfs_major == ddi_name_to_major(ZFS_DRIVER)) 202ea8dc4b6Seschrock zfs_minor = ZFS_MIN_MINOR; 203ea8dc4b6Seschrock else 204ea8dc4b6Seschrock zfs_minor = 0; 205ea8dc4b6Seschrock } else { 206ea8dc4b6Seschrock zfs_minor++; 207ea8dc4b6Seschrock } 208ea8dc4b6Seschrock *dev = makedevice(zfs_major, zfs_minor); 209ea8dc4b6Seschrock mutex_exit(&zfs_dev_mtx); 210ea8dc4b6Seschrock } while (vfs_devismounted(*dev) && zfs_minor != start); 211ea8dc4b6Seschrock if (zfs_minor == start) { 212ea8dc4b6Seschrock /* 213ea8dc4b6Seschrock * We are using all ~262,000 minor numbers for the 214ea8dc4b6Seschrock * current major number. Create a new major number. 215ea8dc4b6Seschrock */ 216ea8dc4b6Seschrock if ((new_major = getudev()) == (major_t)-1) { 217ea8dc4b6Seschrock cmn_err(CE_WARN, 218ea8dc4b6Seschrock "zfs_mount: Can't get unique major " 219ea8dc4b6Seschrock "device number."); 220ea8dc4b6Seschrock return (-1); 221ea8dc4b6Seschrock } 222ea8dc4b6Seschrock mutex_enter(&zfs_dev_mtx); 223ea8dc4b6Seschrock zfs_major = new_major; 224ea8dc4b6Seschrock zfs_minor = 0; 225ea8dc4b6Seschrock 226ea8dc4b6Seschrock mutex_exit(&zfs_dev_mtx); 227ea8dc4b6Seschrock } else { 228ea8dc4b6Seschrock break; 229ea8dc4b6Seschrock } 230ea8dc4b6Seschrock /* CONSTANTCONDITION */ 231ea8dc4b6Seschrock } while (1); 232ea8dc4b6Seschrock 233ea8dc4b6Seschrock return (0); 234ea8dc4b6Seschrock } 235ea8dc4b6Seschrock 236fa9e4066Sahrens static void 237fa9e4066Sahrens atime_changed_cb(void *arg, uint64_t newval) 238fa9e4066Sahrens { 239fa9e4066Sahrens zfsvfs_t *zfsvfs = arg; 240fa9e4066Sahrens 241fa9e4066Sahrens if (newval == TRUE) { 242fa9e4066Sahrens zfsvfs->z_atime = TRUE; 243fa9e4066Sahrens vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NOATIME); 244fa9e4066Sahrens vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_ATIME, NULL, 0); 245fa9e4066Sahrens } else { 246fa9e4066Sahrens zfsvfs->z_atime = FALSE; 247fa9e4066Sahrens vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_ATIME); 248fa9e4066Sahrens vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NOATIME, NULL, 0); 249fa9e4066Sahrens } 250fa9e4066Sahrens } 251fa9e4066Sahrens 2527b55fa8eSck static void 2537b55fa8eSck xattr_changed_cb(void *arg, uint64_t newval) 2547b55fa8eSck { 2557b55fa8eSck zfsvfs_t *zfsvfs = arg; 2567b55fa8eSck 2577b55fa8eSck if (newval == TRUE) { 2587b55fa8eSck /* XXX locking on vfs_flag? */ 2597b55fa8eSck zfsvfs->z_vfs->vfs_flag |= VFS_XATTR; 2607b55fa8eSck vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NOXATTR); 2617b55fa8eSck vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_XATTR, NULL, 0); 2627b55fa8eSck } else { 2637b55fa8eSck /* XXX locking on vfs_flag? */ 2647b55fa8eSck zfsvfs->z_vfs->vfs_flag &= ~VFS_XATTR; 2657b55fa8eSck vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_XATTR); 2667b55fa8eSck vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NOXATTR, NULL, 0); 2677b55fa8eSck } 2687b55fa8eSck } 2697b55fa8eSck 270fa9e4066Sahrens static void 271fa9e4066Sahrens blksz_changed_cb(void *arg, uint64_t newval) 272fa9e4066Sahrens { 273fa9e4066Sahrens zfsvfs_t *zfsvfs = arg; 274fa9e4066Sahrens 275fa9e4066Sahrens if (newval < SPA_MINBLOCKSIZE || 276fa9e4066Sahrens newval > SPA_MAXBLOCKSIZE || !ISP2(newval)) 277fa9e4066Sahrens newval = SPA_MAXBLOCKSIZE; 278fa9e4066Sahrens 279fa9e4066Sahrens zfsvfs->z_max_blksz = newval; 280fa9e4066Sahrens zfsvfs->z_vfs->vfs_bsize = newval; 281fa9e4066Sahrens } 282fa9e4066Sahrens 283fa9e4066Sahrens static void 284fa9e4066Sahrens readonly_changed_cb(void *arg, uint64_t newval) 285fa9e4066Sahrens { 286fa9e4066Sahrens zfsvfs_t *zfsvfs = arg; 287fa9e4066Sahrens 288fa9e4066Sahrens if (newval) { 289fa9e4066Sahrens /* XXX locking on vfs_flag? */ 290fa9e4066Sahrens zfsvfs->z_vfs->vfs_flag |= VFS_RDONLY; 291fa9e4066Sahrens vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_RW); 292fa9e4066Sahrens vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_RO, NULL, 0); 293fa9e4066Sahrens } else { 294fa9e4066Sahrens /* XXX locking on vfs_flag? */ 295fa9e4066Sahrens zfsvfs->z_vfs->vfs_flag &= ~VFS_RDONLY; 296fa9e4066Sahrens vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_RO); 297fa9e4066Sahrens vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_RW, NULL, 0); 298fa9e4066Sahrens } 299fa9e4066Sahrens } 300fa9e4066Sahrens 301fa9e4066Sahrens static void 302fa9e4066Sahrens devices_changed_cb(void *arg, uint64_t newval) 303fa9e4066Sahrens { 304fa9e4066Sahrens zfsvfs_t *zfsvfs = arg; 305fa9e4066Sahrens 306fa9e4066Sahrens if (newval == FALSE) { 307fa9e4066Sahrens zfsvfs->z_vfs->vfs_flag |= VFS_NODEVICES; 308fa9e4066Sahrens vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_DEVICES); 309fa9e4066Sahrens vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NODEVICES, NULL, 0); 310fa9e4066Sahrens } else { 311fa9e4066Sahrens zfsvfs->z_vfs->vfs_flag &= ~VFS_NODEVICES; 312fa9e4066Sahrens vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NODEVICES); 313fa9e4066Sahrens vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_DEVICES, NULL, 0); 314fa9e4066Sahrens } 315fa9e4066Sahrens } 316fa9e4066Sahrens 317fa9e4066Sahrens static void 318fa9e4066Sahrens setuid_changed_cb(void *arg, uint64_t newval) 319fa9e4066Sahrens { 320fa9e4066Sahrens zfsvfs_t *zfsvfs = arg; 321fa9e4066Sahrens 322fa9e4066Sahrens if (newval == FALSE) { 323fa9e4066Sahrens zfsvfs->z_vfs->vfs_flag |= VFS_NOSETUID; 324fa9e4066Sahrens vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_SETUID); 325fa9e4066Sahrens vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NOSETUID, NULL, 0); 326fa9e4066Sahrens } else { 327fa9e4066Sahrens zfsvfs->z_vfs->vfs_flag &= ~VFS_NOSETUID; 328fa9e4066Sahrens vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NOSETUID); 329fa9e4066Sahrens vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_SETUID, NULL, 0); 330fa9e4066Sahrens } 331fa9e4066Sahrens } 332fa9e4066Sahrens 333fa9e4066Sahrens static void 334fa9e4066Sahrens exec_changed_cb(void *arg, uint64_t newval) 335fa9e4066Sahrens { 336fa9e4066Sahrens zfsvfs_t *zfsvfs = arg; 337fa9e4066Sahrens 338fa9e4066Sahrens if (newval == FALSE) { 339fa9e4066Sahrens zfsvfs->z_vfs->vfs_flag |= VFS_NOEXEC; 340fa9e4066Sahrens vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_EXEC); 341fa9e4066Sahrens vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NOEXEC, NULL, 0); 342fa9e4066Sahrens } else { 343fa9e4066Sahrens zfsvfs->z_vfs->vfs_flag &= ~VFS_NOEXEC; 344fa9e4066Sahrens vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NOEXEC); 345fa9e4066Sahrens vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_EXEC, NULL, 0); 346fa9e4066Sahrens } 347fa9e4066Sahrens } 348fa9e4066Sahrens 349da6c28aaSamw /* 350da6c28aaSamw * The nbmand mount option can be changed at mount time. 351da6c28aaSamw * We can't allow it to be toggled on live file systems or incorrect 352da6c28aaSamw * behavior may be seen from cifs clients 353da6c28aaSamw * 354da6c28aaSamw * This property isn't registered via dsl_prop_register(), but this callback 355da6c28aaSamw * will be called when a file system is first mounted 356da6c28aaSamw */ 357da6c28aaSamw static void 358da6c28aaSamw nbmand_changed_cb(void *arg, uint64_t newval) 359da6c28aaSamw { 360da6c28aaSamw zfsvfs_t *zfsvfs = arg; 361da6c28aaSamw if (newval == FALSE) { 362da6c28aaSamw vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NBMAND); 363da6c28aaSamw vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NONBMAND, NULL, 0); 364da6c28aaSamw } else { 365da6c28aaSamw vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NONBMAND); 366da6c28aaSamw vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NBMAND, NULL, 0); 367da6c28aaSamw } 368da6c28aaSamw } 369da6c28aaSamw 370fa9e4066Sahrens static void 371fa9e4066Sahrens snapdir_changed_cb(void *arg, uint64_t newval) 372fa9e4066Sahrens { 373fa9e4066Sahrens zfsvfs_t *zfsvfs = arg; 374fa9e4066Sahrens 375fa9e4066Sahrens zfsvfs->z_show_ctldir = newval; 376fa9e4066Sahrens } 377fa9e4066Sahrens 378da6c28aaSamw static void 379da6c28aaSamw vscan_changed_cb(void *arg, uint64_t newval) 380da6c28aaSamw { 381da6c28aaSamw zfsvfs_t *zfsvfs = arg; 382da6c28aaSamw 383da6c28aaSamw zfsvfs->z_vscan = newval; 384da6c28aaSamw } 385da6c28aaSamw 386*a3c49ce1SAlbert Lee static void 387*a3c49ce1SAlbert Lee acl_mode_changed_cb(void *arg, uint64_t newval) 388*a3c49ce1SAlbert Lee { 389*a3c49ce1SAlbert Lee zfsvfs_t *zfsvfs = arg; 390*a3c49ce1SAlbert Lee 391*a3c49ce1SAlbert Lee zfsvfs->z_acl_mode = newval; 392*a3c49ce1SAlbert Lee } 393*a3c49ce1SAlbert Lee 394fa9e4066Sahrens static void 395fa9e4066Sahrens acl_inherit_changed_cb(void *arg, uint64_t newval) 396fa9e4066Sahrens { 397fa9e4066Sahrens zfsvfs_t *zfsvfs = arg; 398fa9e4066Sahrens 399fa9e4066Sahrens zfsvfs->z_acl_inherit = newval; 400fa9e4066Sahrens } 401fa9e4066Sahrens 402ea8dc4b6Seschrock static int 403ea8dc4b6Seschrock zfs_register_callbacks(vfs_t *vfsp) 404ea8dc4b6Seschrock { 405ea8dc4b6Seschrock struct dsl_dataset *ds = NULL; 406ea8dc4b6Seschrock objset_t *os = NULL; 407ea8dc4b6Seschrock zfsvfs_t *zfsvfs = NULL; 408da6c28aaSamw uint64_t nbmand; 409da6c28aaSamw int readonly, do_readonly = B_FALSE; 410da6c28aaSamw int setuid, do_setuid = B_FALSE; 411da6c28aaSamw int exec, do_exec = B_FALSE; 412da6c28aaSamw int devices, do_devices = B_FALSE; 413da6c28aaSamw int xattr, do_xattr = B_FALSE; 414da6c28aaSamw int atime, do_atime = B_FALSE; 415ea8dc4b6Seschrock int error = 0; 416ea8dc4b6Seschrock 417ea8dc4b6Seschrock ASSERT(vfsp); 418ea8dc4b6Seschrock zfsvfs = vfsp->vfs_data; 419ea8dc4b6Seschrock ASSERT(zfsvfs); 420ea8dc4b6Seschrock os = zfsvfs->z_os; 421fa9e4066Sahrens 422fa9e4066Sahrens /* 423ea8dc4b6Seschrock * The act of registering our callbacks will destroy any mount 424ea8dc4b6Seschrock * options we may have. In order to enable temporary overrides 4257b55fa8eSck * of mount options, we stash away the current values and 426ea8dc4b6Seschrock * restore them after we register the callbacks. 427fa9e4066Sahrens */ 428f9af39baSGeorge Wilson if (vfs_optionisset(vfsp, MNTOPT_RO, NULL) || 429f9af39baSGeorge Wilson !spa_writeable(dmu_objset_spa(os))) { 430ea8dc4b6Seschrock readonly = B_TRUE; 431ea8dc4b6Seschrock do_readonly = B_TRUE; 432ea8dc4b6Seschrock } else if (vfs_optionisset(vfsp, MNTOPT_RW, NULL)) { 433ea8dc4b6Seschrock readonly = B_FALSE; 434ea8dc4b6Seschrock do_readonly = B_TRUE; 435ea8dc4b6Seschrock } 436ea8dc4b6Seschrock if (vfs_optionisset(vfsp, MNTOPT_NOSUID, NULL)) { 437ea8dc4b6Seschrock devices = B_FALSE; 438ea8dc4b6Seschrock setuid = B_FALSE; 439ea8dc4b6Seschrock do_devices = B_TRUE; 440ea8dc4b6Seschrock do_setuid = B_TRUE; 441ea8dc4b6Seschrock } else { 442ea8dc4b6Seschrock if (vfs_optionisset(vfsp, MNTOPT_NODEVICES, NULL)) { 443ea8dc4b6Seschrock devices = B_FALSE; 444ea8dc4b6Seschrock do_devices = B_TRUE; 445b1b8ab34Slling } else if (vfs_optionisset(vfsp, MNTOPT_DEVICES, NULL)) { 446ea8dc4b6Seschrock devices = B_TRUE; 447ea8dc4b6Seschrock do_devices = B_TRUE; 448fa9e4066Sahrens } 449fa9e4066Sahrens 450ea8dc4b6Seschrock if (vfs_optionisset(vfsp, MNTOPT_NOSETUID, NULL)) { 451ea8dc4b6Seschrock setuid = B_FALSE; 452ea8dc4b6Seschrock do_setuid = B_TRUE; 453ea8dc4b6Seschrock } else if (vfs_optionisset(vfsp, MNTOPT_SETUID, NULL)) { 454ea8dc4b6Seschrock setuid = B_TRUE; 455ea8dc4b6Seschrock do_setuid = B_TRUE; 456fa9e4066Sahrens } 457ea8dc4b6Seschrock } 458ea8dc4b6Seschrock if (vfs_optionisset(vfsp, MNTOPT_NOEXEC, NULL)) { 459ea8dc4b6Seschrock exec = B_FALSE; 460ea8dc4b6Seschrock do_exec = B_TRUE; 461ea8dc4b6Seschrock } else if (vfs_optionisset(vfsp, MNTOPT_EXEC, NULL)) { 462ea8dc4b6Seschrock exec = B_TRUE; 463ea8dc4b6Seschrock do_exec = B_TRUE; 464fa9e4066Sahrens } 4657b55fa8eSck if (vfs_optionisset(vfsp, MNTOPT_NOXATTR, NULL)) { 4667b55fa8eSck xattr = B_FALSE; 4677b55fa8eSck do_xattr = B_TRUE; 4687b55fa8eSck } else if (vfs_optionisset(vfsp, MNTOPT_XATTR, NULL)) { 4697b55fa8eSck xattr = B_TRUE; 4707b55fa8eSck do_xattr = B_TRUE; 4717b55fa8eSck } 472b510d378Slling if (vfs_optionisset(vfsp, MNTOPT_NOATIME, NULL)) { 473b510d378Slling atime = B_FALSE; 474b510d378Slling do_atime = B_TRUE; 475b510d378Slling } else if (vfs_optionisset(vfsp, MNTOPT_ATIME, NULL)) { 476b510d378Slling atime = B_TRUE; 477b510d378Slling do_atime = B_TRUE; 478b510d378Slling } 479fa9e4066Sahrens 480da6c28aaSamw /* 481da6c28aaSamw * nbmand is a special property. It can only be changed at 482da6c28aaSamw * mount time. 483da6c28aaSamw * 484da6c28aaSamw * This is weird, but it is documented to only be changeable 485da6c28aaSamw * at mount time. 486da6c28aaSamw */ 487da6c28aaSamw if (vfs_optionisset(vfsp, MNTOPT_NONBMAND, NULL)) { 488da6c28aaSamw nbmand = B_FALSE; 489da6c28aaSamw } else if (vfs_optionisset(vfsp, MNTOPT_NBMAND, NULL)) { 490da6c28aaSamw nbmand = B_TRUE; 491da6c28aaSamw } else { 492da6c28aaSamw char osname[MAXNAMELEN]; 493da6c28aaSamw 494da6c28aaSamw dmu_objset_name(os, osname); 495da6c28aaSamw if (error = dsl_prop_get_integer(osname, "nbmand", &nbmand, 496bb0ade09Sahrens NULL)) { 497bb0ade09Sahrens return (error); 498bb0ade09Sahrens } 499da6c28aaSamw } 500da6c28aaSamw 501fa9e4066Sahrens /* 502ea8dc4b6Seschrock * Register property callbacks. 503ea8dc4b6Seschrock * 504ea8dc4b6Seschrock * It would probably be fine to just check for i/o error from 505ea8dc4b6Seschrock * the first prop_register(), but I guess I like to go 506ea8dc4b6Seschrock * overboard... 507fa9e4066Sahrens */ 508ea8dc4b6Seschrock ds = dmu_objset_ds(os); 509ea8dc4b6Seschrock error = dsl_prop_register(ds, "atime", atime_changed_cb, zfsvfs); 5107b55fa8eSck error = error ? error : dsl_prop_register(ds, 5117b55fa8eSck "xattr", xattr_changed_cb, zfsvfs); 512ea8dc4b6Seschrock error = error ? error : dsl_prop_register(ds, 513ea8dc4b6Seschrock "recordsize", blksz_changed_cb, zfsvfs); 514ea8dc4b6Seschrock error = error ? error : dsl_prop_register(ds, 515ea8dc4b6Seschrock "readonly", readonly_changed_cb, zfsvfs); 516ea8dc4b6Seschrock error = error ? error : dsl_prop_register(ds, 517ea8dc4b6Seschrock "devices", devices_changed_cb, zfsvfs); 518ea8dc4b6Seschrock error = error ? error : dsl_prop_register(ds, 519ea8dc4b6Seschrock "setuid", setuid_changed_cb, zfsvfs); 520ea8dc4b6Seschrock error = error ? error : dsl_prop_register(ds, 521ea8dc4b6Seschrock "exec", exec_changed_cb, zfsvfs); 522ea8dc4b6Seschrock error = error ? error : dsl_prop_register(ds, 523ea8dc4b6Seschrock "snapdir", snapdir_changed_cb, zfsvfs); 524*a3c49ce1SAlbert Lee error = error ? error : dsl_prop_register(ds, 525*a3c49ce1SAlbert Lee "aclmode", acl_mode_changed_cb, zfsvfs); 526ea8dc4b6Seschrock error = error ? error : dsl_prop_register(ds, 527ea8dc4b6Seschrock "aclinherit", acl_inherit_changed_cb, zfsvfs); 528da6c28aaSamw error = error ? error : dsl_prop_register(ds, 529da6c28aaSamw "vscan", vscan_changed_cb, zfsvfs); 530ea8dc4b6Seschrock if (error) 531ea8dc4b6Seschrock goto unregister; 532fa9e4066Sahrens 533ea8dc4b6Seschrock /* 534ea8dc4b6Seschrock * Invoke our callbacks to restore temporary mount options. 535ea8dc4b6Seschrock */ 536ea8dc4b6Seschrock if (do_readonly) 537ea8dc4b6Seschrock readonly_changed_cb(zfsvfs, readonly); 538ea8dc4b6Seschrock if (do_setuid) 539ea8dc4b6Seschrock setuid_changed_cb(zfsvfs, setuid); 540ea8dc4b6Seschrock if (do_exec) 541ea8dc4b6Seschrock exec_changed_cb(zfsvfs, exec); 542ea8dc4b6Seschrock if (do_devices) 543ea8dc4b6Seschrock devices_changed_cb(zfsvfs, devices); 5447b55fa8eSck if (do_xattr) 5457b55fa8eSck xattr_changed_cb(zfsvfs, xattr); 546b510d378Slling if (do_atime) 547b510d378Slling atime_changed_cb(zfsvfs, atime); 548fa9e4066Sahrens 549da6c28aaSamw nbmand_changed_cb(zfsvfs, nbmand); 550da6c28aaSamw 551ea8dc4b6Seschrock return (0); 552fa9e4066Sahrens 553ea8dc4b6Seschrock unregister: 554fa9e4066Sahrens /* 555ea8dc4b6Seschrock * We may attempt to unregister some callbacks that are not 556ea8dc4b6Seschrock * registered, but this is OK; it will simply return ENOMSG, 557ea8dc4b6Seschrock * which we will ignore. 558fa9e4066Sahrens */ 559ea8dc4b6Seschrock (void) dsl_prop_unregister(ds, "atime", atime_changed_cb, zfsvfs); 5607b55fa8eSck (void) dsl_prop_unregister(ds, "xattr", xattr_changed_cb, zfsvfs); 561ea8dc4b6Seschrock (void) dsl_prop_unregister(ds, "recordsize", blksz_changed_cb, zfsvfs); 562ea8dc4b6Seschrock (void) dsl_prop_unregister(ds, "readonly", readonly_changed_cb, zfsvfs); 563ea8dc4b6Seschrock (void) dsl_prop_unregister(ds, "devices", devices_changed_cb, zfsvfs); 564ea8dc4b6Seschrock (void) dsl_prop_unregister(ds, "setuid", setuid_changed_cb, zfsvfs); 565ea8dc4b6Seschrock (void) dsl_prop_unregister(ds, "exec", exec_changed_cb, zfsvfs); 566ea8dc4b6Seschrock (void) dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb, zfsvfs); 567*a3c49ce1SAlbert Lee (void) dsl_prop_unregister(ds, "aclmode", acl_mode_changed_cb, zfsvfs); 568ea8dc4b6Seschrock (void) dsl_prop_unregister(ds, "aclinherit", acl_inherit_changed_cb, 569ea8dc4b6Seschrock zfsvfs); 570da6c28aaSamw (void) dsl_prop_unregister(ds, "vscan", vscan_changed_cb, zfsvfs); 571ea8dc4b6Seschrock return (error); 572ea8dc4b6Seschrock 573ea8dc4b6Seschrock } 574ea8dc4b6Seschrock 5759966ca11SMatthew Ahrens static int 5760a586ceaSMark Shellenbaum zfs_space_delta_cb(dmu_object_type_t bonustype, void *data, 5779966ca11SMatthew Ahrens uint64_t *userp, uint64_t *groupp) 57814843421SMatthew Ahrens { 5790a586ceaSMark Shellenbaum znode_phys_t *znp = data; 5800a586ceaSMark Shellenbaum int error = 0; 58114843421SMatthew Ahrens 58206e0070dSMark Shellenbaum /* 58306e0070dSMark Shellenbaum * Is it a valid type of object to track? 58406e0070dSMark Shellenbaum */ 5850a586ceaSMark Shellenbaum if (bonustype != DMU_OT_ZNODE && bonustype != DMU_OT_SA) 5869966ca11SMatthew Ahrens return (ENOENT); 58714843421SMatthew Ahrens 58806e0070dSMark Shellenbaum /* 58906e0070dSMark Shellenbaum * If we have a NULL data pointer 59006e0070dSMark Shellenbaum * then assume the id's aren't changing and 59106e0070dSMark Shellenbaum * return EEXIST to the dmu to let it know to 59206e0070dSMark Shellenbaum * use the same ids 59306e0070dSMark Shellenbaum */ 59406e0070dSMark Shellenbaum if (data == NULL) 59506e0070dSMark Shellenbaum return (EEXIST); 59606e0070dSMark Shellenbaum 5970a586ceaSMark Shellenbaum if (bonustype == DMU_OT_ZNODE) { 5980a586ceaSMark Shellenbaum *userp = znp->zp_uid; 5990a586ceaSMark Shellenbaum *groupp = znp->zp_gid; 6000a586ceaSMark Shellenbaum } else { 6010a586ceaSMark Shellenbaum int hdrsize; 6020a586ceaSMark Shellenbaum 6030a586ceaSMark Shellenbaum ASSERT(bonustype == DMU_OT_SA); 6040a586ceaSMark Shellenbaum hdrsize = sa_hdrsize(data); 6050a586ceaSMark Shellenbaum 6060a586ceaSMark Shellenbaum if (hdrsize != 0) { 6070a586ceaSMark Shellenbaum *userp = *((uint64_t *)((uintptr_t)data + hdrsize + 6080a586ceaSMark Shellenbaum SA_UID_OFFSET)); 6090a586ceaSMark Shellenbaum *groupp = *((uint64_t *)((uintptr_t)data + hdrsize + 6100a586ceaSMark Shellenbaum SA_GID_OFFSET)); 6110a586ceaSMark Shellenbaum } else { 61206e0070dSMark Shellenbaum /* 61306e0070dSMark Shellenbaum * This should only happen for newly created 61406e0070dSMark Shellenbaum * files that haven't had the znode data filled 61506e0070dSMark Shellenbaum * in yet. 61606e0070dSMark Shellenbaum */ 61706e0070dSMark Shellenbaum *userp = 0; 61806e0070dSMark Shellenbaum *groupp = 0; 6190a586ceaSMark Shellenbaum } 6200a586ceaSMark Shellenbaum } 6210a586ceaSMark Shellenbaum return (error); 62214843421SMatthew Ahrens } 62314843421SMatthew Ahrens 62414843421SMatthew Ahrens static void 62514843421SMatthew Ahrens fuidstr_to_sid(zfsvfs_t *zfsvfs, const char *fuidstr, 62614843421SMatthew Ahrens char *domainbuf, int buflen, uid_t *ridp) 62714843421SMatthew Ahrens { 62814843421SMatthew Ahrens uint64_t fuid; 62914843421SMatthew Ahrens const char *domain; 63014843421SMatthew Ahrens 63114843421SMatthew Ahrens fuid = strtonum(fuidstr, NULL); 63214843421SMatthew Ahrens 63314843421SMatthew Ahrens domain = zfs_fuid_find_by_idx(zfsvfs, FUID_INDEX(fuid)); 63414843421SMatthew Ahrens if (domain) 63514843421SMatthew Ahrens (void) strlcpy(domainbuf, domain, buflen); 63614843421SMatthew Ahrens else 63714843421SMatthew Ahrens domainbuf[0] = '\0'; 63814843421SMatthew Ahrens *ridp = FUID_RID(fuid); 63914843421SMatthew Ahrens } 64014843421SMatthew Ahrens 64114843421SMatthew Ahrens static uint64_t 64214843421SMatthew Ahrens zfs_userquota_prop_to_obj(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type) 64314843421SMatthew Ahrens { 64414843421SMatthew Ahrens switch (type) { 64514843421SMatthew Ahrens case ZFS_PROP_USERUSED: 64614843421SMatthew Ahrens return (DMU_USERUSED_OBJECT); 64714843421SMatthew Ahrens case ZFS_PROP_GROUPUSED: 64814843421SMatthew Ahrens return (DMU_GROUPUSED_OBJECT); 64914843421SMatthew Ahrens case ZFS_PROP_USERQUOTA: 65014843421SMatthew Ahrens return (zfsvfs->z_userquota_obj); 65114843421SMatthew Ahrens case ZFS_PROP_GROUPQUOTA: 65214843421SMatthew Ahrens return (zfsvfs->z_groupquota_obj); 65314843421SMatthew Ahrens } 65414843421SMatthew Ahrens return (0); 65514843421SMatthew Ahrens } 65614843421SMatthew Ahrens 65714843421SMatthew Ahrens int 65814843421SMatthew Ahrens zfs_userspace_many(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type, 65914843421SMatthew Ahrens uint64_t *cookiep, void *vbuf, uint64_t *bufsizep) 66014843421SMatthew Ahrens { 66114843421SMatthew Ahrens int error; 66214843421SMatthew Ahrens zap_cursor_t zc; 66314843421SMatthew Ahrens zap_attribute_t za; 66414843421SMatthew Ahrens zfs_useracct_t *buf = vbuf; 66514843421SMatthew Ahrens uint64_t obj; 66614843421SMatthew Ahrens 66714843421SMatthew Ahrens if (!dmu_objset_userspace_present(zfsvfs->z_os)) 66814843421SMatthew Ahrens return (ENOTSUP); 66914843421SMatthew Ahrens 67014843421SMatthew Ahrens obj = zfs_userquota_prop_to_obj(zfsvfs, type); 67114843421SMatthew Ahrens if (obj == 0) { 67214843421SMatthew Ahrens *bufsizep = 0; 67314843421SMatthew Ahrens return (0); 67414843421SMatthew Ahrens } 67514843421SMatthew Ahrens 67614843421SMatthew Ahrens for (zap_cursor_init_serialized(&zc, zfsvfs->z_os, obj, *cookiep); 67714843421SMatthew Ahrens (error = zap_cursor_retrieve(&zc, &za)) == 0; 67814843421SMatthew Ahrens zap_cursor_advance(&zc)) { 67914843421SMatthew Ahrens if ((uintptr_t)buf - (uintptr_t)vbuf + sizeof (zfs_useracct_t) > 68014843421SMatthew Ahrens *bufsizep) 68114843421SMatthew Ahrens break; 68214843421SMatthew Ahrens 68314843421SMatthew Ahrens fuidstr_to_sid(zfsvfs, za.za_name, 68414843421SMatthew Ahrens buf->zu_domain, sizeof (buf->zu_domain), &buf->zu_rid); 68514843421SMatthew Ahrens 68614843421SMatthew Ahrens buf->zu_space = za.za_first_integer; 68714843421SMatthew Ahrens buf++; 68814843421SMatthew Ahrens } 68914843421SMatthew Ahrens if (error == ENOENT) 69014843421SMatthew Ahrens error = 0; 69114843421SMatthew Ahrens 69214843421SMatthew Ahrens ASSERT3U((uintptr_t)buf - (uintptr_t)vbuf, <=, *bufsizep); 69314843421SMatthew Ahrens *bufsizep = (uintptr_t)buf - (uintptr_t)vbuf; 69414843421SMatthew Ahrens *cookiep = zap_cursor_serialize(&zc); 69514843421SMatthew Ahrens zap_cursor_fini(&zc); 69614843421SMatthew Ahrens return (error); 69714843421SMatthew Ahrens } 69814843421SMatthew Ahrens 69914843421SMatthew Ahrens /* 70014843421SMatthew Ahrens * buf must be big enough (eg, 32 bytes) 70114843421SMatthew Ahrens */ 70214843421SMatthew Ahrens static int 70314843421SMatthew Ahrens id_to_fuidstr(zfsvfs_t *zfsvfs, const char *domain, uid_t rid, 70414843421SMatthew Ahrens char *buf, boolean_t addok) 70514843421SMatthew Ahrens { 70614843421SMatthew Ahrens uint64_t fuid; 70714843421SMatthew Ahrens int domainid = 0; 70814843421SMatthew Ahrens 70914843421SMatthew Ahrens if (domain && domain[0]) { 71014843421SMatthew Ahrens domainid = zfs_fuid_find_by_domain(zfsvfs, domain, NULL, addok); 71114843421SMatthew Ahrens if (domainid == -1) 71214843421SMatthew Ahrens return (ENOENT); 71314843421SMatthew Ahrens } 71414843421SMatthew Ahrens fuid = FUID_ENCODE(domainid, rid); 71514843421SMatthew Ahrens (void) sprintf(buf, "%llx", (longlong_t)fuid); 71614843421SMatthew Ahrens return (0); 71714843421SMatthew Ahrens } 71814843421SMatthew Ahrens 71914843421SMatthew Ahrens int 72014843421SMatthew Ahrens zfs_userspace_one(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type, 72114843421SMatthew Ahrens const char *domain, uint64_t rid, uint64_t *valp) 72214843421SMatthew Ahrens { 72314843421SMatthew Ahrens char buf[32]; 72414843421SMatthew Ahrens int err; 72514843421SMatthew Ahrens uint64_t obj; 72614843421SMatthew Ahrens 72714843421SMatthew Ahrens *valp = 0; 72814843421SMatthew Ahrens 72914843421SMatthew Ahrens if (!dmu_objset_userspace_present(zfsvfs->z_os)) 73014843421SMatthew Ahrens return (ENOTSUP); 73114843421SMatthew Ahrens 73214843421SMatthew Ahrens obj = zfs_userquota_prop_to_obj(zfsvfs, type); 73314843421SMatthew Ahrens if (obj == 0) 73414843421SMatthew Ahrens return (0); 73514843421SMatthew Ahrens 73614843421SMatthew Ahrens err = id_to_fuidstr(zfsvfs, domain, rid, buf, B_FALSE); 73714843421SMatthew Ahrens if (err) 73814843421SMatthew Ahrens return (err); 73914843421SMatthew Ahrens 74014843421SMatthew Ahrens err = zap_lookup(zfsvfs->z_os, obj, buf, 8, 1, valp); 74114843421SMatthew Ahrens if (err == ENOENT) 74214843421SMatthew Ahrens err = 0; 74314843421SMatthew Ahrens return (err); 74414843421SMatthew Ahrens } 74514843421SMatthew Ahrens 74614843421SMatthew Ahrens int 74714843421SMatthew Ahrens zfs_set_userquota(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type, 74814843421SMatthew Ahrens const char *domain, uint64_t rid, uint64_t quota) 74914843421SMatthew Ahrens { 75014843421SMatthew Ahrens char buf[32]; 75114843421SMatthew Ahrens int err; 75214843421SMatthew Ahrens dmu_tx_t *tx; 75314843421SMatthew Ahrens uint64_t *objp; 75414843421SMatthew Ahrens boolean_t fuid_dirtied; 75514843421SMatthew Ahrens 75614843421SMatthew Ahrens if (type != ZFS_PROP_USERQUOTA && type != ZFS_PROP_GROUPQUOTA) 75714843421SMatthew Ahrens return (EINVAL); 75814843421SMatthew Ahrens 75914843421SMatthew Ahrens if (zfsvfs->z_version < ZPL_VERSION_USERSPACE) 76014843421SMatthew Ahrens return (ENOTSUP); 76114843421SMatthew Ahrens 76214843421SMatthew Ahrens objp = (type == ZFS_PROP_USERQUOTA) ? &zfsvfs->z_userquota_obj : 76314843421SMatthew Ahrens &zfsvfs->z_groupquota_obj; 76414843421SMatthew Ahrens 76514843421SMatthew Ahrens err = id_to_fuidstr(zfsvfs, domain, rid, buf, B_TRUE); 76614843421SMatthew Ahrens if (err) 76714843421SMatthew Ahrens return (err); 76814843421SMatthew Ahrens fuid_dirtied = zfsvfs->z_fuid_dirty; 76914843421SMatthew Ahrens 77014843421SMatthew Ahrens tx = dmu_tx_create(zfsvfs->z_os); 77114843421SMatthew Ahrens dmu_tx_hold_zap(tx, *objp ? *objp : DMU_NEW_OBJECT, B_TRUE, NULL); 77214843421SMatthew Ahrens if (*objp == 0) { 77314843421SMatthew Ahrens dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE, 77414843421SMatthew Ahrens zfs_userquota_prop_prefixes[type]); 77514843421SMatthew Ahrens } 77614843421SMatthew Ahrens if (fuid_dirtied) 77714843421SMatthew Ahrens zfs_fuid_txhold(zfsvfs, tx); 77814843421SMatthew Ahrens err = dmu_tx_assign(tx, TXG_WAIT); 77914843421SMatthew Ahrens if (err) { 78014843421SMatthew Ahrens dmu_tx_abort(tx); 78114843421SMatthew Ahrens return (err); 78214843421SMatthew Ahrens } 78314843421SMatthew Ahrens 78414843421SMatthew Ahrens mutex_enter(&zfsvfs->z_lock); 78514843421SMatthew Ahrens if (*objp == 0) { 78614843421SMatthew Ahrens *objp = zap_create(zfsvfs->z_os, DMU_OT_USERGROUP_QUOTA, 78714843421SMatthew Ahrens DMU_OT_NONE, 0, tx); 78814843421SMatthew Ahrens VERIFY(0 == zap_add(zfsvfs->z_os, MASTER_NODE_OBJ, 78914843421SMatthew Ahrens zfs_userquota_prop_prefixes[type], 8, 1, objp, tx)); 79014843421SMatthew Ahrens } 79114843421SMatthew Ahrens mutex_exit(&zfsvfs->z_lock); 79214843421SMatthew Ahrens 79314843421SMatthew Ahrens if (quota == 0) { 79414843421SMatthew Ahrens err = zap_remove(zfsvfs->z_os, *objp, buf, tx); 79514843421SMatthew Ahrens if (err == ENOENT) 79614843421SMatthew Ahrens err = 0; 79714843421SMatthew Ahrens } else { 79814843421SMatthew Ahrens err = zap_update(zfsvfs->z_os, *objp, buf, 8, 1, "a, tx); 79914843421SMatthew Ahrens } 80014843421SMatthew Ahrens ASSERT(err == 0); 80114843421SMatthew Ahrens if (fuid_dirtied) 80214843421SMatthew Ahrens zfs_fuid_sync(zfsvfs, tx); 80314843421SMatthew Ahrens dmu_tx_commit(tx); 80414843421SMatthew Ahrens return (err); 80514843421SMatthew Ahrens } 80614843421SMatthew Ahrens 80714843421SMatthew Ahrens boolean_t 8080a586ceaSMark Shellenbaum zfs_fuid_overquota(zfsvfs_t *zfsvfs, boolean_t isgroup, uint64_t fuid) 80914843421SMatthew Ahrens { 81014843421SMatthew Ahrens char buf[32]; 81114843421SMatthew Ahrens uint64_t used, quota, usedobj, quotaobj; 81214843421SMatthew Ahrens int err; 81314843421SMatthew Ahrens 81414843421SMatthew Ahrens usedobj = isgroup ? DMU_GROUPUSED_OBJECT : DMU_USERUSED_OBJECT; 81514843421SMatthew Ahrens quotaobj = isgroup ? zfsvfs->z_groupquota_obj : zfsvfs->z_userquota_obj; 81614843421SMatthew Ahrens 81714843421SMatthew Ahrens if (quotaobj == 0 || zfsvfs->z_replay) 81814843421SMatthew Ahrens return (B_FALSE); 81914843421SMatthew Ahrens 82014843421SMatthew Ahrens (void) sprintf(buf, "%llx", (longlong_t)fuid); 82114843421SMatthew Ahrens err = zap_lookup(zfsvfs->z_os, quotaobj, buf, 8, 1, "a); 82214843421SMatthew Ahrens if (err != 0) 82314843421SMatthew Ahrens return (B_FALSE); 82414843421SMatthew Ahrens 82514843421SMatthew Ahrens err = zap_lookup(zfsvfs->z_os, usedobj, buf, 8, 1, &used); 82614843421SMatthew Ahrens if (err != 0) 82714843421SMatthew Ahrens return (B_FALSE); 82814843421SMatthew Ahrens return (used >= quota); 82914843421SMatthew Ahrens } 83014843421SMatthew Ahrens 8310a586ceaSMark Shellenbaum boolean_t 8320a586ceaSMark Shellenbaum zfs_owner_overquota(zfsvfs_t *zfsvfs, znode_t *zp, boolean_t isgroup) 8330a586ceaSMark Shellenbaum { 8340a586ceaSMark Shellenbaum uint64_t fuid; 8350a586ceaSMark Shellenbaum uint64_t quotaobj; 8360a586ceaSMark Shellenbaum 8370a586ceaSMark Shellenbaum quotaobj = isgroup ? zfsvfs->z_groupquota_obj : zfsvfs->z_userquota_obj; 8380a586ceaSMark Shellenbaum 839f1696b23SMark Shellenbaum fuid = isgroup ? zp->z_gid : zp->z_uid; 8400a586ceaSMark Shellenbaum 8410a586ceaSMark Shellenbaum if (quotaobj == 0 || zfsvfs->z_replay) 8420a586ceaSMark Shellenbaum return (B_FALSE); 8430a586ceaSMark Shellenbaum 8440a586ceaSMark Shellenbaum return (zfs_fuid_overquota(zfsvfs, isgroup, fuid)); 8450a586ceaSMark Shellenbaum } 8460a586ceaSMark Shellenbaum 84714843421SMatthew Ahrens int 848af4c679fSSean McEnroe zfsvfs_create(const char *osname, zfsvfs_t **zfvp) 84914843421SMatthew Ahrens { 85014843421SMatthew Ahrens objset_t *os; 85114843421SMatthew Ahrens zfsvfs_t *zfsvfs; 85214843421SMatthew Ahrens uint64_t zval; 85314843421SMatthew Ahrens int i, error; 8540a586ceaSMark Shellenbaum uint64_t sa_obj; 85514843421SMatthew Ahrens 856503ad85cSMatthew Ahrens zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP); 85714843421SMatthew Ahrens 858503ad85cSMatthew Ahrens /* 859503ad85cSMatthew Ahrens * We claim to always be readonly so we can open snapshots; 860503ad85cSMatthew Ahrens * other ZPL code will prevent us from writing to snapshots. 861503ad85cSMatthew Ahrens */ 862503ad85cSMatthew Ahrens error = dmu_objset_own(osname, DMU_OST_ZFS, B_TRUE, zfsvfs, &os); 863503ad85cSMatthew Ahrens if (error) { 864503ad85cSMatthew Ahrens kmem_free(zfsvfs, sizeof (zfsvfs_t)); 86514843421SMatthew Ahrens return (error); 866503ad85cSMatthew Ahrens } 86714843421SMatthew Ahrens 86814843421SMatthew Ahrens /* 86914843421SMatthew Ahrens * Initialize the zfs-specific filesystem structure. 87014843421SMatthew Ahrens * Should probably make this a kmem cache, shuffle fields, 87114843421SMatthew Ahrens * and just bzero up to z_hold_mtx[]. 87214843421SMatthew Ahrens */ 87314843421SMatthew Ahrens zfsvfs->z_vfs = NULL; 87414843421SMatthew Ahrens zfsvfs->z_parent = zfsvfs; 87514843421SMatthew Ahrens zfsvfs->z_max_blksz = SPA_MAXBLOCKSIZE; 87614843421SMatthew Ahrens zfsvfs->z_show_ctldir = ZFS_SNAPDIR_VISIBLE; 87714843421SMatthew Ahrens zfsvfs->z_os = os; 87814843421SMatthew Ahrens 87914843421SMatthew Ahrens error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zfsvfs->z_version); 88014843421SMatthew Ahrens if (error) { 88114843421SMatthew Ahrens goto out; 882dc7cd546SMark Shellenbaum } else if (zfsvfs->z_version > 883dc7cd546SMark Shellenbaum zfs_zpl_version_map(spa_version(dmu_objset_spa(os)))) { 884dc7cd546SMark Shellenbaum (void) printf("Can't mount a version %lld file system " 885dc7cd546SMark Shellenbaum "on a version %lld pool\n. Pool must be upgraded to mount " 886dc7cd546SMark Shellenbaum "this file system.", (u_longlong_t)zfsvfs->z_version, 887dc7cd546SMark Shellenbaum (u_longlong_t)spa_version(dmu_objset_spa(os))); 88814843421SMatthew Ahrens error = ENOTSUP; 88914843421SMatthew Ahrens goto out; 89014843421SMatthew Ahrens } 89114843421SMatthew Ahrens if ((error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &zval)) != 0) 89214843421SMatthew Ahrens goto out; 89314843421SMatthew Ahrens zfsvfs->z_norm = (int)zval; 89414843421SMatthew Ahrens 89514843421SMatthew Ahrens if ((error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &zval)) != 0) 89614843421SMatthew Ahrens goto out; 89714843421SMatthew Ahrens zfsvfs->z_utf8 = (zval != 0); 89814843421SMatthew Ahrens 89914843421SMatthew Ahrens if ((error = zfs_get_zplprop(os, ZFS_PROP_CASE, &zval)) != 0) 90014843421SMatthew Ahrens goto out; 90114843421SMatthew Ahrens zfsvfs->z_case = (uint_t)zval; 90214843421SMatthew Ahrens 90314843421SMatthew Ahrens /* 90414843421SMatthew Ahrens * Fold case on file systems that are always or sometimes case 90514843421SMatthew Ahrens * insensitive. 90614843421SMatthew Ahrens */ 90714843421SMatthew Ahrens if (zfsvfs->z_case == ZFS_CASE_INSENSITIVE || 90814843421SMatthew Ahrens zfsvfs->z_case == ZFS_CASE_MIXED) 90914843421SMatthew Ahrens zfsvfs->z_norm |= U8_TEXTPREP_TOUPPER; 91014843421SMatthew Ahrens 91114843421SMatthew Ahrens zfsvfs->z_use_fuids = USE_FUIDS(zfsvfs->z_version, zfsvfs->z_os); 9120a586ceaSMark Shellenbaum zfsvfs->z_use_sa = USE_SA(zfsvfs->z_version, zfsvfs->z_os); 9130a586ceaSMark Shellenbaum 9140a586ceaSMark Shellenbaum if (zfsvfs->z_use_sa) { 9150a586ceaSMark Shellenbaum /* should either have both of these objects or none */ 9160a586ceaSMark Shellenbaum error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, 9170a586ceaSMark Shellenbaum &sa_obj); 9180a586ceaSMark Shellenbaum if (error) 9190a586ceaSMark Shellenbaum return (error); 9200a586ceaSMark Shellenbaum } else { 9210a586ceaSMark Shellenbaum /* 9220a586ceaSMark Shellenbaum * Pre SA versions file systems should never touch 9230a586ceaSMark Shellenbaum * either the attribute registration or layout objects. 9240a586ceaSMark Shellenbaum */ 9250a586ceaSMark Shellenbaum sa_obj = 0; 9260a586ceaSMark Shellenbaum } 9270a586ceaSMark Shellenbaum 9281d8ccc7bSMark Shellenbaum error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END, 9291d8ccc7bSMark Shellenbaum &zfsvfs->z_attr_table); 9301d8ccc7bSMark Shellenbaum if (error) 9311d8ccc7bSMark Shellenbaum goto out; 9320a586ceaSMark Shellenbaum 9330a586ceaSMark Shellenbaum if (zfsvfs->z_version >= ZPL_VERSION_SA) 9340a586ceaSMark Shellenbaum sa_register_update_callback(os, zfs_sa_upgrade); 93514843421SMatthew Ahrens 93614843421SMatthew Ahrens error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1, 93714843421SMatthew Ahrens &zfsvfs->z_root); 93814843421SMatthew Ahrens if (error) 93914843421SMatthew Ahrens goto out; 94014843421SMatthew Ahrens ASSERT(zfsvfs->z_root != 0); 94114843421SMatthew Ahrens 94214843421SMatthew Ahrens error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1, 94314843421SMatthew Ahrens &zfsvfs->z_unlinkedobj); 94414843421SMatthew Ahrens if (error) 94514843421SMatthew Ahrens goto out; 94614843421SMatthew Ahrens 94714843421SMatthew Ahrens error = zap_lookup(os, MASTER_NODE_OBJ, 94814843421SMatthew Ahrens zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA], 94914843421SMatthew Ahrens 8, 1, &zfsvfs->z_userquota_obj); 95014843421SMatthew Ahrens if (error && error != ENOENT) 95114843421SMatthew Ahrens goto out; 95214843421SMatthew Ahrens 95314843421SMatthew Ahrens error = zap_lookup(os, MASTER_NODE_OBJ, 95414843421SMatthew Ahrens zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA], 95514843421SMatthew Ahrens 8, 1, &zfsvfs->z_groupquota_obj); 95614843421SMatthew Ahrens if (error && error != ENOENT) 95714843421SMatthew Ahrens goto out; 95814843421SMatthew Ahrens 95914843421SMatthew Ahrens error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1, 96014843421SMatthew Ahrens &zfsvfs->z_fuid_obj); 96114843421SMatthew Ahrens if (error && error != ENOENT) 96214843421SMatthew Ahrens goto out; 96314843421SMatthew Ahrens 96414843421SMatthew Ahrens error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SHARES_DIR, 8, 1, 96514843421SMatthew Ahrens &zfsvfs->z_shares_dir); 96614843421SMatthew Ahrens if (error && error != ENOENT) 96714843421SMatthew Ahrens goto out; 96814843421SMatthew Ahrens 96914843421SMatthew Ahrens mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); 97014843421SMatthew Ahrens mutex_init(&zfsvfs->z_lock, NULL, MUTEX_DEFAULT, NULL); 97114843421SMatthew Ahrens list_create(&zfsvfs->z_all_znodes, sizeof (znode_t), 97214843421SMatthew Ahrens offsetof(znode_t, z_link_node)); 97314843421SMatthew Ahrens rrw_init(&zfsvfs->z_teardown_lock); 97414843421SMatthew Ahrens rw_init(&zfsvfs->z_teardown_inactive_lock, NULL, RW_DEFAULT, NULL); 97514843421SMatthew Ahrens rw_init(&zfsvfs->z_fuid_lock, NULL, RW_DEFAULT, NULL); 97614843421SMatthew Ahrens for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) 97714843421SMatthew Ahrens mutex_init(&zfsvfs->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL); 97814843421SMatthew Ahrens 979af4c679fSSean McEnroe *zfvp = zfsvfs; 98014843421SMatthew Ahrens return (0); 98114843421SMatthew Ahrens 98214843421SMatthew Ahrens out: 983503ad85cSMatthew Ahrens dmu_objset_disown(os, zfsvfs); 984af4c679fSSean McEnroe *zfvp = NULL; 98514843421SMatthew Ahrens kmem_free(zfsvfs, sizeof (zfsvfs_t)); 98614843421SMatthew Ahrens return (error); 98714843421SMatthew Ahrens } 98814843421SMatthew Ahrens 989f18faf3fSek static int 990f18faf3fSek zfsvfs_setup(zfsvfs_t *zfsvfs, boolean_t mounting) 991f18faf3fSek { 992f18faf3fSek int error; 993f18faf3fSek 994f18faf3fSek error = zfs_register_callbacks(zfsvfs->z_vfs); 995f18faf3fSek if (error) 996f18faf3fSek return (error); 997f18faf3fSek 998f18faf3fSek /* 999f18faf3fSek * Set the objset user_ptr to track its zfsvfs. 1000f18faf3fSek */ 1001503ad85cSMatthew Ahrens mutex_enter(&zfsvfs->z_os->os_user_ptr_lock); 1002f18faf3fSek dmu_objset_set_user(zfsvfs->z_os, zfsvfs); 1003503ad85cSMatthew Ahrens mutex_exit(&zfsvfs->z_os->os_user_ptr_lock); 1004f18faf3fSek 1005377c02aaSNeil Perrin zfsvfs->z_log = zil_open(zfsvfs->z_os, zfs_get_data); 1006377c02aaSNeil Perrin 1007f18faf3fSek /* 1008f18faf3fSek * If we are not mounting (ie: online recv), then we don't 1009f18faf3fSek * have to worry about replaying the log as we blocked all 1010f18faf3fSek * operations out since we closed the ZIL. 1011f18faf3fSek */ 1012f18faf3fSek if (mounting) { 1013a6e57bd4SNeil Perrin boolean_t readonly; 1014a6e57bd4SNeil Perrin 1015f18faf3fSek /* 1016f18faf3fSek * During replay we remove the read only flag to 1017f18faf3fSek * allow replays to succeed. 1018f18faf3fSek */ 1019f18faf3fSek readonly = zfsvfs->z_vfs->vfs_flag & VFS_RDONLY; 10201209a471SNeil Perrin if (readonly != 0) 10211209a471SNeil Perrin zfsvfs->z_vfs->vfs_flag &= ~VFS_RDONLY; 10221209a471SNeil Perrin else 10231209a471SNeil Perrin zfs_unlinked_drain(zfsvfs); 1024f18faf3fSek 102555da60b9SMark J Musante /* 102655da60b9SMark J Musante * Parse and replay the intent log. 102755da60b9SMark J Musante * 102855da60b9SMark J Musante * Because of ziltest, this must be done after 102955da60b9SMark J Musante * zfs_unlinked_drain(). (Further note: ziltest 103055da60b9SMark J Musante * doesn't use readonly mounts, where 103155da60b9SMark J Musante * zfs_unlinked_drain() isn't called.) This is because 103255da60b9SMark J Musante * ziltest causes spa_sync() to think it's committed, 103355da60b9SMark J Musante * but actually it is not, so the intent log contains 103455da60b9SMark J Musante * many txg's worth of changes. 103555da60b9SMark J Musante * 103655da60b9SMark J Musante * In particular, if object N is in the unlinked set in 103755da60b9SMark J Musante * the last txg to actually sync, then it could be 103855da60b9SMark J Musante * actually freed in a later txg and then reallocated 103955da60b9SMark J Musante * in a yet later txg. This would write a "create 104055da60b9SMark J Musante * object N" record to the intent log. Normally, this 104155da60b9SMark J Musante * would be fine because the spa_sync() would have 104255da60b9SMark J Musante * written out the fact that object N is free, before 104355da60b9SMark J Musante * we could write the "create object N" intent log 104455da60b9SMark J Musante * record. 104555da60b9SMark J Musante * 104655da60b9SMark J Musante * But when we are in ziltest mode, we advance the "open 104755da60b9SMark J Musante * txg" without actually spa_sync()-ing the changes to 104855da60b9SMark J Musante * disk. So we would see that object N is still 104955da60b9SMark J Musante * allocated and in the unlinked set, and there is an 105055da60b9SMark J Musante * intent log record saying to allocate it. 105155da60b9SMark J Musante */ 1052f9af39baSGeorge Wilson if (spa_writeable(dmu_objset_spa(zfsvfs->z_os))) { 1053f9af39baSGeorge Wilson if (zil_replay_disable) { 1054f9af39baSGeorge Wilson zil_destroy(zfsvfs->z_log, B_FALSE); 1055f9af39baSGeorge Wilson } else { 1056f9af39baSGeorge Wilson zfsvfs->z_replay = B_TRUE; 1057f9af39baSGeorge Wilson zil_replay(zfsvfs->z_os, zfsvfs, 1058f9af39baSGeorge Wilson zfs_replay_vector); 1059f9af39baSGeorge Wilson zfsvfs->z_replay = B_FALSE; 1060f9af39baSGeorge Wilson } 10611209a471SNeil Perrin } 1062f18faf3fSek zfsvfs->z_vfs->vfs_flag |= readonly; /* restore readonly bit */ 1063f18faf3fSek } 1064f18faf3fSek 1065f18faf3fSek return (0); 1066f18faf3fSek } 1067f18faf3fSek 106814843421SMatthew Ahrens void 106914843421SMatthew Ahrens zfsvfs_free(zfsvfs_t *zfsvfs) 107047f263f4Sek { 107114843421SMatthew Ahrens int i; 10724e9583b2STom Erickson extern krwlock_t zfsvfs_lock; /* in zfs_znode.c */ 10734e9583b2STom Erickson 10744e9583b2STom Erickson /* 10754e9583b2STom Erickson * This is a barrier to prevent the filesystem from going away in 10764e9583b2STom Erickson * zfs_znode_move() until we can safely ensure that the filesystem is 10774e9583b2STom Erickson * not unmounted. We consider the filesystem valid before the barrier 10784e9583b2STom Erickson * and invalid after the barrier. 10794e9583b2STom Erickson */ 10804e9583b2STom Erickson rw_enter(&zfsvfs_lock, RW_READER); 10814e9583b2STom Erickson rw_exit(&zfsvfs_lock); 108214843421SMatthew Ahrens 108314843421SMatthew Ahrens zfs_fuid_destroy(zfsvfs); 108414843421SMatthew Ahrens 108547f263f4Sek mutex_destroy(&zfsvfs->z_znodes_lock); 10869e1320c0SMark Shellenbaum mutex_destroy(&zfsvfs->z_lock); 108747f263f4Sek list_destroy(&zfsvfs->z_all_znodes); 108847f263f4Sek rrw_destroy(&zfsvfs->z_teardown_lock); 108947f263f4Sek rw_destroy(&zfsvfs->z_teardown_inactive_lock); 109047f263f4Sek rw_destroy(&zfsvfs->z_fuid_lock); 109114843421SMatthew Ahrens for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) 109214843421SMatthew Ahrens mutex_destroy(&zfsvfs->z_hold_mtx[i]); 109347f263f4Sek kmem_free(zfsvfs, sizeof (zfsvfs_t)); 109447f263f4Sek } 109547f263f4Sek 109614843421SMatthew Ahrens static void 109714843421SMatthew Ahrens zfs_set_fuid_feature(zfsvfs_t *zfsvfs) 109814843421SMatthew Ahrens { 109914843421SMatthew Ahrens zfsvfs->z_use_fuids = USE_FUIDS(zfsvfs->z_version, zfsvfs->z_os); 110044bffe01SMark Shellenbaum if (zfsvfs->z_vfs) { 110144bffe01SMark Shellenbaum if (zfsvfs->z_use_fuids) { 110244bffe01SMark Shellenbaum vfs_set_feature(zfsvfs->z_vfs, VFSFT_XVATTR); 110344bffe01SMark Shellenbaum vfs_set_feature(zfsvfs->z_vfs, VFSFT_SYSATTR_VIEWS); 110444bffe01SMark Shellenbaum vfs_set_feature(zfsvfs->z_vfs, VFSFT_ACEMASKONACCESS); 110544bffe01SMark Shellenbaum vfs_set_feature(zfsvfs->z_vfs, VFSFT_ACLONCREATE); 110644bffe01SMark Shellenbaum vfs_set_feature(zfsvfs->z_vfs, VFSFT_ACCESS_FILTER); 110744bffe01SMark Shellenbaum vfs_set_feature(zfsvfs->z_vfs, VFSFT_REPARSE); 110844bffe01SMark Shellenbaum } else { 110944bffe01SMark Shellenbaum vfs_clear_feature(zfsvfs->z_vfs, VFSFT_XVATTR); 111044bffe01SMark Shellenbaum vfs_clear_feature(zfsvfs->z_vfs, VFSFT_SYSATTR_VIEWS); 111144bffe01SMark Shellenbaum vfs_clear_feature(zfsvfs->z_vfs, VFSFT_ACEMASKONACCESS); 111244bffe01SMark Shellenbaum vfs_clear_feature(zfsvfs->z_vfs, VFSFT_ACLONCREATE); 111344bffe01SMark Shellenbaum vfs_clear_feature(zfsvfs->z_vfs, VFSFT_ACCESS_FILTER); 111444bffe01SMark Shellenbaum vfs_clear_feature(zfsvfs->z_vfs, VFSFT_REPARSE); 111544bffe01SMark Shellenbaum } 111614843421SMatthew Ahrens } 11170a586ceaSMark Shellenbaum zfsvfs->z_use_sa = USE_SA(zfsvfs->z_version, zfsvfs->z_os); 111814843421SMatthew Ahrens } 111914843421SMatthew Ahrens 1120ea8dc4b6Seschrock static int 1121088f3894Sahrens zfs_domount(vfs_t *vfsp, char *osname) 1122ea8dc4b6Seschrock { 1123ea8dc4b6Seschrock dev_t mount_dev; 112414843421SMatthew Ahrens uint64_t recordsize, fsid_guid; 1125ea8dc4b6Seschrock int error = 0; 1126ea8dc4b6Seschrock zfsvfs_t *zfsvfs; 1127ea8dc4b6Seschrock 1128ea8dc4b6Seschrock ASSERT(vfsp); 1129ea8dc4b6Seschrock ASSERT(osname); 1130fa9e4066Sahrens 1131503ad85cSMatthew Ahrens error = zfsvfs_create(osname, &zfsvfs); 113214843421SMatthew Ahrens if (error) 113314843421SMatthew Ahrens return (error); 1134fa9e4066Sahrens zfsvfs->z_vfs = vfsp; 1135fa9e4066Sahrens 1136ea8dc4b6Seschrock /* Initialize the generic filesystem structure. */ 1137fa9e4066Sahrens vfsp->vfs_bcount = 0; 1138fa9e4066Sahrens vfsp->vfs_data = NULL; 1139fa9e4066Sahrens 1140ea8dc4b6Seschrock if (zfs_create_unique_device(&mount_dev) == -1) { 1141ea8dc4b6Seschrock error = ENODEV; 1142ea8dc4b6Seschrock goto out; 1143ea8dc4b6Seschrock } 1144fa9e4066Sahrens ASSERT(vfs_devismounted(mount_dev) == 0); 1145fa9e4066Sahrens 1146ea8dc4b6Seschrock if (error = dsl_prop_get_integer(osname, "recordsize", &recordsize, 1147ea8dc4b6Seschrock NULL)) 1148ea8dc4b6Seschrock goto out; 1149fa9e4066Sahrens 1150fa9e4066Sahrens vfsp->vfs_dev = mount_dev; 1151fa9e4066Sahrens vfsp->vfs_fstype = zfsfstype; 1152fa9e4066Sahrens vfsp->vfs_bsize = recordsize; 1153fa9e4066Sahrens vfsp->vfs_flag |= VFS_NOTRUNC; 1154fa9e4066Sahrens vfsp->vfs_data = zfsvfs; 1155fa9e4066Sahrens 115614843421SMatthew Ahrens /* 115714843421SMatthew Ahrens * The fsid is 64 bits, composed of an 8-bit fs type, which 115814843421SMatthew Ahrens * separates our fsid from any other filesystem types, and a 115914843421SMatthew Ahrens * 56-bit objset unique ID. The objset unique ID is unique to 116014843421SMatthew Ahrens * all objsets open on this system, provided by unique_create(). 116114843421SMatthew Ahrens * The 8-bit fs type must be put in the low bits of fsid[1] 116214843421SMatthew Ahrens * because that's where other Solaris filesystems put it. 116314843421SMatthew Ahrens */ 116414843421SMatthew Ahrens fsid_guid = dmu_objset_fsid_guid(zfsvfs->z_os); 116514843421SMatthew Ahrens ASSERT((fsid_guid & ~((1ULL<<56)-1)) == 0); 116614843421SMatthew Ahrens vfsp->vfs_fsid.val[0] = fsid_guid; 116714843421SMatthew Ahrens vfsp->vfs_fsid.val[1] = ((fsid_guid>>32) << 8) | 116814843421SMatthew Ahrens zfsfstype & 0xFF; 1169ea8dc4b6Seschrock 1170da6c28aaSamw /* 1171da6c28aaSamw * Set features for file system. 1172da6c28aaSamw */ 117314843421SMatthew Ahrens zfs_set_fuid_feature(zfsvfs); 1174de8267e0Stimh if (zfsvfs->z_case == ZFS_CASE_INSENSITIVE) { 1175de8267e0Stimh vfs_set_feature(vfsp, VFSFT_DIRENTFLAGS); 1176de8267e0Stimh vfs_set_feature(vfsp, VFSFT_CASEINSENSITIVE); 1177de8267e0Stimh vfs_set_feature(vfsp, VFSFT_NOCASESENSITIVE); 1178de8267e0Stimh } else if (zfsvfs->z_case == ZFS_CASE_MIXED) { 1179de8267e0Stimh vfs_set_feature(vfsp, VFSFT_DIRENTFLAGS); 1180de8267e0Stimh vfs_set_feature(vfsp, VFSFT_CASEINSENSITIVE); 1181de8267e0Stimh } 1182c242f9a0Schunli zhang - Sun Microsystems - Irvine United States vfs_set_feature(vfsp, VFSFT_ZEROCOPY_SUPPORTED); 1183da6c28aaSamw 1184ea8dc4b6Seschrock if (dmu_objset_is_snapshot(zfsvfs->z_os)) { 1185da6c28aaSamw uint64_t pval; 11867b55fa8eSck 1187fa9e4066Sahrens atime_changed_cb(zfsvfs, B_FALSE); 1188fa9e4066Sahrens readonly_changed_cb(zfsvfs, B_TRUE); 1189da6c28aaSamw if (error = dsl_prop_get_integer(osname, "xattr", &pval, NULL)) 11907b55fa8eSck goto out; 1191da6c28aaSamw xattr_changed_cb(zfsvfs, pval); 1192fa9e4066Sahrens zfsvfs->z_issnap = B_TRUE; 1193b9deb9cbSMark J Musante zfsvfs->z_os->os_sync = ZFS_SYNC_DISABLED; 1194777badbaSMatthew Ahrens 1195503ad85cSMatthew Ahrens mutex_enter(&zfsvfs->z_os->os_user_ptr_lock); 1196777badbaSMatthew Ahrens dmu_objset_set_user(zfsvfs->z_os, zfsvfs); 1197503ad85cSMatthew Ahrens mutex_exit(&zfsvfs->z_os->os_user_ptr_lock); 1198fa9e4066Sahrens } else { 1199f18faf3fSek error = zfsvfs_setup(zfsvfs, B_TRUE); 1200ea8dc4b6Seschrock } 1201fa9e4066Sahrens 1202ea8dc4b6Seschrock if (!zfsvfs->z_issnap) 1203ea8dc4b6Seschrock zfsctl_create(zfsvfs); 1204ea8dc4b6Seschrock out: 1205ea8dc4b6Seschrock if (error) { 1206503ad85cSMatthew Ahrens dmu_objset_disown(zfsvfs->z_os, zfsvfs); 120714843421SMatthew Ahrens zfsvfs_free(zfsvfs); 1208ea8dc4b6Seschrock } else { 1209ea8dc4b6Seschrock atomic_add_32(&zfs_active_fs_count, 1); 1210ea8dc4b6Seschrock } 1211fa9e4066Sahrens 1212ea8dc4b6Seschrock return (error); 1213ea8dc4b6Seschrock } 1214ea8dc4b6Seschrock 1215ea8dc4b6Seschrock void 1216ea8dc4b6Seschrock zfs_unregister_callbacks(zfsvfs_t *zfsvfs) 1217ea8dc4b6Seschrock { 1218ea8dc4b6Seschrock objset_t *os = zfsvfs->z_os; 1219ea8dc4b6Seschrock struct dsl_dataset *ds; 1220ea8dc4b6Seschrock 1221ea8dc4b6Seschrock /* 1222ea8dc4b6Seschrock * Unregister properties. 1223ea8dc4b6Seschrock */ 1224ea8dc4b6Seschrock if (!dmu_objset_is_snapshot(os)) { 1225fa9e4066Sahrens ds = dmu_objset_ds(os); 1226ea8dc4b6Seschrock VERIFY(dsl_prop_unregister(ds, "atime", atime_changed_cb, 1227fa9e4066Sahrens zfsvfs) == 0); 1228fa9e4066Sahrens 12297b55fa8eSck VERIFY(dsl_prop_unregister(ds, "xattr", xattr_changed_cb, 12307b55fa8eSck zfsvfs) == 0); 12317b55fa8eSck 1232ea8dc4b6Seschrock VERIFY(dsl_prop_unregister(ds, "recordsize", blksz_changed_cb, 1233fa9e4066Sahrens zfsvfs) == 0); 1234fa9e4066Sahrens 1235ea8dc4b6Seschrock VERIFY(dsl_prop_unregister(ds, "readonly", readonly_changed_cb, 1236fa9e4066Sahrens zfsvfs) == 0); 1237fa9e4066Sahrens 1238ea8dc4b6Seschrock VERIFY(dsl_prop_unregister(ds, "devices", devices_changed_cb, 1239fa9e4066Sahrens zfsvfs) == 0); 1240fa9e4066Sahrens 1241ea8dc4b6Seschrock VERIFY(dsl_prop_unregister(ds, "setuid", setuid_changed_cb, 1242fa9e4066Sahrens zfsvfs) == 0); 1243fa9e4066Sahrens 1244ea8dc4b6Seschrock VERIFY(dsl_prop_unregister(ds, "exec", exec_changed_cb, 1245fa9e4066Sahrens zfsvfs) == 0); 1246fa9e4066Sahrens 1247ea8dc4b6Seschrock VERIFY(dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb, 1248fa9e4066Sahrens zfsvfs) == 0); 1249fa9e4066Sahrens 1250*a3c49ce1SAlbert Lee VERIFY(dsl_prop_unregister(ds, "aclmode", acl_mode_changed_cb, 1251*a3c49ce1SAlbert Lee zfsvfs) == 0); 1252*a3c49ce1SAlbert Lee 1253ea8dc4b6Seschrock VERIFY(dsl_prop_unregister(ds, "aclinherit", 1254fa9e4066Sahrens acl_inherit_changed_cb, zfsvfs) == 0); 1255da6c28aaSamw 1256da6c28aaSamw VERIFY(dsl_prop_unregister(ds, "vscan", 1257da6c28aaSamw vscan_changed_cb, zfsvfs) == 0); 1258ea8dc4b6Seschrock } 1259ea8dc4b6Seschrock } 1260fa9e4066Sahrens 1261b1b8ab34Slling /* 1262b1b8ab34Slling * Convert a decimal digit string to a uint64_t integer. 1263b1b8ab34Slling */ 1264b1b8ab34Slling static int 1265b1b8ab34Slling str_to_uint64(char *str, uint64_t *objnum) 1266b1b8ab34Slling { 1267b1b8ab34Slling uint64_t num = 0; 1268b1b8ab34Slling 1269b1b8ab34Slling while (*str) { 1270b1b8ab34Slling if (*str < '0' || *str > '9') 1271b1b8ab34Slling return (EINVAL); 1272b1b8ab34Slling 1273b1b8ab34Slling num = num*10 + *str++ - '0'; 1274b1b8ab34Slling } 1275b1b8ab34Slling 1276b1b8ab34Slling *objnum = num; 1277b1b8ab34Slling return (0); 1278b1b8ab34Slling } 1279b1b8ab34Slling 1280b1b8ab34Slling /* 1281b1b8ab34Slling * The boot path passed from the boot loader is in the form of 1282b1b8ab34Slling * "rootpool-name/root-filesystem-object-number'. Convert this 1283b1b8ab34Slling * string to a dataset name: "rootpool-name/root-filesystem-name". 1284b1b8ab34Slling */ 1285b1b8ab34Slling static int 1286e7cbe64fSgw zfs_parse_bootfs(char *bpath, char *outpath) 1287b1b8ab34Slling { 1288b1b8ab34Slling char *slashp; 1289b1b8ab34Slling uint64_t objnum; 1290b1b8ab34Slling int error; 1291b1b8ab34Slling 1292b1b8ab34Slling if (*bpath == 0 || *bpath == '/') 1293b1b8ab34Slling return (EINVAL); 1294b1b8ab34Slling 129519397407SSherry Moore (void) strcpy(outpath, bpath); 129619397407SSherry Moore 1297b1b8ab34Slling slashp = strchr(bpath, '/'); 1298b1b8ab34Slling 1299b1b8ab34Slling /* if no '/', just return the pool name */ 1300b1b8ab34Slling if (slashp == NULL) { 1301b1b8ab34Slling return (0); 1302b1b8ab34Slling } 1303b1b8ab34Slling 130419397407SSherry Moore /* if not a number, just return the root dataset name */ 130519397407SSherry Moore if (str_to_uint64(slashp+1, &objnum)) { 130619397407SSherry Moore return (0); 130719397407SSherry Moore } 1308b1b8ab34Slling 1309b1b8ab34Slling *slashp = '\0'; 1310b1b8ab34Slling error = dsl_dsobj_to_dsname(bpath, objnum, outpath); 1311b1b8ab34Slling *slashp = '/'; 1312b1b8ab34Slling 1313b1b8ab34Slling return (error); 1314b1b8ab34Slling } 1315b1b8ab34Slling 13164201a95eSRic Aleshire /* 13174201a95eSRic Aleshire * zfs_check_global_label: 13184201a95eSRic Aleshire * Check that the hex label string is appropriate for the dataset 13194201a95eSRic Aleshire * being mounted into the global_zone proper. 13204201a95eSRic Aleshire * 13214201a95eSRic Aleshire * Return an error if the hex label string is not default or 13224201a95eSRic Aleshire * admin_low/admin_high. For admin_low labels, the corresponding 13234201a95eSRic Aleshire * dataset must be readonly. 13244201a95eSRic Aleshire */ 13254201a95eSRic Aleshire int 13264201a95eSRic Aleshire zfs_check_global_label(const char *dsname, const char *hexsl) 13274201a95eSRic Aleshire { 13284201a95eSRic Aleshire if (strcasecmp(hexsl, ZFS_MLSLABEL_DEFAULT) == 0) 13294201a95eSRic Aleshire return (0); 13304201a95eSRic Aleshire if (strcasecmp(hexsl, ADMIN_HIGH) == 0) 13314201a95eSRic Aleshire return (0); 13324201a95eSRic Aleshire if (strcasecmp(hexsl, ADMIN_LOW) == 0) { 13334201a95eSRic Aleshire /* must be readonly */ 13344201a95eSRic Aleshire uint64_t rdonly; 13354201a95eSRic Aleshire 13364201a95eSRic Aleshire if (dsl_prop_get_integer(dsname, 13374201a95eSRic Aleshire zfs_prop_to_name(ZFS_PROP_READONLY), &rdonly, NULL)) 13384201a95eSRic Aleshire return (EACCES); 13394201a95eSRic Aleshire return (rdonly ? 0 : EACCES); 13404201a95eSRic Aleshire } 13414201a95eSRic Aleshire return (EACCES); 13424201a95eSRic Aleshire } 13434201a95eSRic Aleshire 13444201a95eSRic Aleshire /* 13454201a95eSRic Aleshire * zfs_mount_label_policy: 13464201a95eSRic Aleshire * Determine whether the mount is allowed according to MAC check. 13474201a95eSRic Aleshire * by comparing (where appropriate) label of the dataset against 13484201a95eSRic Aleshire * the label of the zone being mounted into. If the dataset has 13494201a95eSRic Aleshire * no label, create one. 13504201a95eSRic Aleshire * 13514201a95eSRic Aleshire * Returns: 13524201a95eSRic Aleshire * 0 : access allowed 13534201a95eSRic Aleshire * >0 : error code, such as EACCES 13544201a95eSRic Aleshire */ 13554201a95eSRic Aleshire static int 13564201a95eSRic Aleshire zfs_mount_label_policy(vfs_t *vfsp, char *osname) 13574201a95eSRic Aleshire { 13584201a95eSRic Aleshire int error, retv; 13594201a95eSRic Aleshire zone_t *mntzone = NULL; 13604201a95eSRic Aleshire ts_label_t *mnt_tsl; 13614201a95eSRic Aleshire bslabel_t *mnt_sl; 13624201a95eSRic Aleshire bslabel_t ds_sl; 13634201a95eSRic Aleshire char ds_hexsl[MAXNAMELEN]; 13644201a95eSRic Aleshire 13654201a95eSRic Aleshire retv = EACCES; /* assume the worst */ 13664201a95eSRic Aleshire 13674201a95eSRic Aleshire /* 13684201a95eSRic Aleshire * Start by getting the dataset label if it exists. 13694201a95eSRic Aleshire */ 13704201a95eSRic Aleshire error = dsl_prop_get(osname, zfs_prop_to_name(ZFS_PROP_MLSLABEL), 13714201a95eSRic Aleshire 1, sizeof (ds_hexsl), &ds_hexsl, NULL); 13724201a95eSRic Aleshire if (error) 13734201a95eSRic Aleshire return (EACCES); 13744201a95eSRic Aleshire 13754201a95eSRic Aleshire /* 13764201a95eSRic Aleshire * If labeling is NOT enabled, then disallow the mount of datasets 13774201a95eSRic Aleshire * which have a non-default label already. No other label checks 13784201a95eSRic Aleshire * are needed. 13794201a95eSRic Aleshire */ 13804201a95eSRic Aleshire if (!is_system_labeled()) { 13814201a95eSRic Aleshire if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) == 0) 13824201a95eSRic Aleshire return (0); 13834201a95eSRic Aleshire return (EACCES); 13844201a95eSRic Aleshire } 13854201a95eSRic Aleshire 13864201a95eSRic Aleshire /* 13874201a95eSRic Aleshire * Get the label of the mountpoint. If mounting into the global 13884201a95eSRic Aleshire * zone (i.e. mountpoint is not within an active zone and the 13894201a95eSRic Aleshire * zoned property is off), the label must be default or 13904201a95eSRic Aleshire * admin_low/admin_high only; no other checks are needed. 13914201a95eSRic Aleshire */ 13924201a95eSRic Aleshire mntzone = zone_find_by_any_path(refstr_value(vfsp->vfs_mntpt), B_FALSE); 13934201a95eSRic Aleshire if (mntzone->zone_id == GLOBAL_ZONEID) { 13944201a95eSRic Aleshire uint64_t zoned; 13954201a95eSRic Aleshire 13964201a95eSRic Aleshire zone_rele(mntzone); 13974201a95eSRic Aleshire 13984201a95eSRic Aleshire if (dsl_prop_get_integer(osname, 13994201a95eSRic Aleshire zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL)) 14004201a95eSRic Aleshire return (EACCES); 14014201a95eSRic Aleshire if (!zoned) 14024201a95eSRic Aleshire return (zfs_check_global_label(osname, ds_hexsl)); 14034201a95eSRic Aleshire else 14044201a95eSRic Aleshire /* 14054201a95eSRic Aleshire * This is the case of a zone dataset being mounted 14064201a95eSRic Aleshire * initially, before the zone has been fully created; 14074201a95eSRic Aleshire * allow this mount into global zone. 14084201a95eSRic Aleshire */ 14094201a95eSRic Aleshire return (0); 14104201a95eSRic Aleshire } 14114201a95eSRic Aleshire 14124201a95eSRic Aleshire mnt_tsl = mntzone->zone_slabel; 14134201a95eSRic Aleshire ASSERT(mnt_tsl != NULL); 14144201a95eSRic Aleshire label_hold(mnt_tsl); 14154201a95eSRic Aleshire mnt_sl = label2bslabel(mnt_tsl); 14164201a95eSRic Aleshire 14174201a95eSRic Aleshire if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) == 0) { 14184201a95eSRic Aleshire /* 14194201a95eSRic Aleshire * The dataset doesn't have a real label, so fabricate one. 14204201a95eSRic Aleshire */ 14214201a95eSRic Aleshire char *str = NULL; 14224201a95eSRic Aleshire 14234201a95eSRic Aleshire if (l_to_str_internal(mnt_sl, &str) == 0 && 14244201a95eSRic Aleshire dsl_prop_set(osname, zfs_prop_to_name(ZFS_PROP_MLSLABEL), 142592241e0bSTom Erickson ZPROP_SRC_LOCAL, 1, strlen(str) + 1, str) == 0) 14264201a95eSRic Aleshire retv = 0; 14274201a95eSRic Aleshire if (str != NULL) 14284201a95eSRic Aleshire kmem_free(str, strlen(str) + 1); 14294201a95eSRic Aleshire } else if (hexstr_to_label(ds_hexsl, &ds_sl) == 0) { 14304201a95eSRic Aleshire /* 14314201a95eSRic Aleshire * Now compare labels to complete the MAC check. If the 14324201a95eSRic Aleshire * labels are equal then allow access. If the mountpoint 14334201a95eSRic Aleshire * label dominates the dataset label, allow readonly access. 14344201a95eSRic Aleshire * Otherwise, access is denied. 14354201a95eSRic Aleshire */ 14364201a95eSRic Aleshire if (blequal(mnt_sl, &ds_sl)) 14374201a95eSRic Aleshire retv = 0; 14384201a95eSRic Aleshire else if (bldominates(mnt_sl, &ds_sl)) { 14394201a95eSRic Aleshire vfs_setmntopt(vfsp, MNTOPT_RO, NULL, 0); 14404201a95eSRic Aleshire retv = 0; 14414201a95eSRic Aleshire } 14424201a95eSRic Aleshire } 14434201a95eSRic Aleshire 14444201a95eSRic Aleshire label_rele(mnt_tsl); 14454201a95eSRic Aleshire zone_rele(mntzone); 14464201a95eSRic Aleshire return (retv); 14474201a95eSRic Aleshire } 14484201a95eSRic Aleshire 1449ea8dc4b6Seschrock static int 1450ea8dc4b6Seschrock zfs_mountroot(vfs_t *vfsp, enum whymountroot why) 1451ea8dc4b6Seschrock { 1452ea8dc4b6Seschrock int error = 0; 1453ea8dc4b6Seschrock static int zfsrootdone = 0; 1454ea8dc4b6Seschrock zfsvfs_t *zfsvfs = NULL; 1455ea8dc4b6Seschrock znode_t *zp = NULL; 1456ea8dc4b6Seschrock vnode_t *vp = NULL; 1457e7cbe64fSgw char *zfs_bootfs; 1458051aabe6Staylor char *zfs_devid; 1459ea8dc4b6Seschrock 1460ea8dc4b6Seschrock ASSERT(vfsp); 1461ea8dc4b6Seschrock 1462ea8dc4b6Seschrock /* 1463b1b8ab34Slling * The filesystem that we mount as root is defined in the 1464e7cbe64fSgw * boot property "zfs-bootfs" with a format of 1465e7cbe64fSgw * "poolname/root-dataset-objnum". 1466ea8dc4b6Seschrock */ 1467ea8dc4b6Seschrock if (why == ROOT_INIT) { 1468ea8dc4b6Seschrock if (zfsrootdone++) 1469ea8dc4b6Seschrock return (EBUSY); 1470e7cbe64fSgw /* 1471e7cbe64fSgw * the process of doing a spa_load will require the 1472e7cbe64fSgw * clock to be set before we could (for example) do 1473e7cbe64fSgw * something better by looking at the timestamp on 1474e7cbe64fSgw * an uberblock, so just set it to -1. 1475e7cbe64fSgw */ 1476e7cbe64fSgw clkset(-1); 1477fa9e4066Sahrens 1478051aabe6Staylor if ((zfs_bootfs = spa_get_bootprop("zfs-bootfs")) == NULL) { 1479051aabe6Staylor cmn_err(CE_NOTE, "spa_get_bootfs: can not get " 1480051aabe6Staylor "bootfs name"); 1481e7cbe64fSgw return (EINVAL); 1482986fd29aSsetje } 1483051aabe6Staylor zfs_devid = spa_get_bootprop("diskdevid"); 1484051aabe6Staylor error = spa_import_rootpool(rootfs.bo_name, zfs_devid); 1485051aabe6Staylor if (zfs_devid) 1486051aabe6Staylor spa_free_bootprop(zfs_devid); 1487051aabe6Staylor if (error) { 1488051aabe6Staylor spa_free_bootprop(zfs_bootfs); 1489051aabe6Staylor cmn_err(CE_NOTE, "spa_import_rootpool: error %d", 1490e7cbe64fSgw error); 1491e7cbe64fSgw return (error); 1492e7cbe64fSgw } 1493e7cbe64fSgw if (error = zfs_parse_bootfs(zfs_bootfs, rootfs.bo_name)) { 1494051aabe6Staylor spa_free_bootprop(zfs_bootfs); 1495051aabe6Staylor cmn_err(CE_NOTE, "zfs_parse_bootfs: error %d", 1496e7cbe64fSgw error); 1497b1b8ab34Slling return (error); 1498e7cbe64fSgw } 1499e7cbe64fSgw 1500051aabe6Staylor spa_free_bootprop(zfs_bootfs); 1501fa9e4066Sahrens 1502ea8dc4b6Seschrock if (error = vfs_lock(vfsp)) 1503ea8dc4b6Seschrock return (error); 1504fa9e4066Sahrens 1505088f3894Sahrens if (error = zfs_domount(vfsp, rootfs.bo_name)) { 1506051aabe6Staylor cmn_err(CE_NOTE, "zfs_domount: error %d", error); 1507ea8dc4b6Seschrock goto out; 1508e7cbe64fSgw } 1509ea8dc4b6Seschrock 1510ea8dc4b6Seschrock zfsvfs = (zfsvfs_t *)vfsp->vfs_data; 1511ea8dc4b6Seschrock ASSERT(zfsvfs); 1512e7cbe64fSgw if (error = zfs_zget(zfsvfs, zfsvfs->z_root, &zp)) { 1513051aabe6Staylor cmn_err(CE_NOTE, "zfs_zget: error %d", error); 1514ea8dc4b6Seschrock goto out; 1515e7cbe64fSgw } 1516ea8dc4b6Seschrock 1517ea8dc4b6Seschrock vp = ZTOV(zp); 1518ea8dc4b6Seschrock mutex_enter(&vp->v_lock); 1519ea8dc4b6Seschrock vp->v_flag |= VROOT; 1520ea8dc4b6Seschrock mutex_exit(&vp->v_lock); 1521ea8dc4b6Seschrock rootvp = vp; 1522ea8dc4b6Seschrock 1523ea8dc4b6Seschrock /* 152440d3dfe1Smarks * Leave rootvp held. The root file system is never unmounted. 1525ea8dc4b6Seschrock */ 1526ea8dc4b6Seschrock 1527ea8dc4b6Seschrock vfs_add((struct vnode *)0, vfsp, 1528ea8dc4b6Seschrock (vfsp->vfs_flag & VFS_RDONLY) ? MS_RDONLY : 0); 1529ea8dc4b6Seschrock out: 1530ea8dc4b6Seschrock vfs_unlock(vfsp); 1531e7cbe64fSgw return (error); 1532ea8dc4b6Seschrock } else if (why == ROOT_REMOUNT) { 1533ea8dc4b6Seschrock readonly_changed_cb(vfsp->vfs_data, B_FALSE); 1534ea8dc4b6Seschrock vfsp->vfs_flag |= VFS_REMOUNT; 1535b510d378Slling 1536b510d378Slling /* refresh mount options */ 1537b510d378Slling zfs_unregister_callbacks(vfsp->vfs_data); 1538b510d378Slling return (zfs_register_callbacks(vfsp)); 1539b510d378Slling 1540ea8dc4b6Seschrock } else if (why == ROOT_UNMOUNT) { 1541ea8dc4b6Seschrock zfs_unregister_callbacks((zfsvfs_t *)vfsp->vfs_data); 1542ea8dc4b6Seschrock (void) zfs_sync(vfsp, 0, 0); 1543ea8dc4b6Seschrock return (0); 1544ea8dc4b6Seschrock } 1545ea8dc4b6Seschrock 1546ea8dc4b6Seschrock /* 1547ea8dc4b6Seschrock * if "why" is equal to anything else other than ROOT_INIT, 1548ea8dc4b6Seschrock * ROOT_REMOUNT, or ROOT_UNMOUNT, we do not support it. 1549ea8dc4b6Seschrock */ 1550ea8dc4b6Seschrock return (ENOTSUP); 1551ea8dc4b6Seschrock } 1552ea8dc4b6Seschrock 1553ea8dc4b6Seschrock /*ARGSUSED*/ 1554ea8dc4b6Seschrock static int 1555ea8dc4b6Seschrock zfs_mount(vfs_t *vfsp, vnode_t *mvp, struct mounta *uap, cred_t *cr) 1556ea8dc4b6Seschrock { 1557ea8dc4b6Seschrock char *osname; 1558ea8dc4b6Seschrock pathname_t spn; 1559ea8dc4b6Seschrock int error = 0; 1560ea8dc4b6Seschrock uio_seg_t fromspace = (uap->flags & MS_SYSSPACE) ? 1561b1b8ab34Slling UIO_SYSSPACE : UIO_USERSPACE; 1562ea8dc4b6Seschrock int canwrite; 1563ea8dc4b6Seschrock 1564ea8dc4b6Seschrock if (mvp->v_type != VDIR) 1565ea8dc4b6Seschrock return (ENOTDIR); 1566ea8dc4b6Seschrock 1567ea8dc4b6Seschrock mutex_enter(&mvp->v_lock); 1568ea8dc4b6Seschrock if ((uap->flags & MS_REMOUNT) == 0 && 1569ea8dc4b6Seschrock (uap->flags & MS_OVERLAY) == 0 && 1570ea8dc4b6Seschrock (mvp->v_count != 1 || (mvp->v_flag & VROOT))) { 1571ea8dc4b6Seschrock mutex_exit(&mvp->v_lock); 1572ea8dc4b6Seschrock return (EBUSY); 1573ea8dc4b6Seschrock } 1574ea8dc4b6Seschrock mutex_exit(&mvp->v_lock); 1575ea8dc4b6Seschrock 1576ea8dc4b6Seschrock /* 1577ea8dc4b6Seschrock * ZFS does not support passing unparsed data in via MS_DATA. 1578ea8dc4b6Seschrock * Users should use the MS_OPTIONSTR interface; this means 1579ea8dc4b6Seschrock * that all option parsing is already done and the options struct 1580ea8dc4b6Seschrock * can be interrogated. 1581ea8dc4b6Seschrock */ 1582ea8dc4b6Seschrock if ((uap->flags & MS_DATA) && uap->datalen > 0) 1583ea8dc4b6Seschrock return (EINVAL); 1584ea8dc4b6Seschrock 1585ea8dc4b6Seschrock /* 1586ea8dc4b6Seschrock * Get the objset name (the "special" mount argument). 1587ea8dc4b6Seschrock */ 1588ea8dc4b6Seschrock if (error = pn_get(uap->spec, fromspace, &spn)) 1589ea8dc4b6Seschrock return (error); 1590ea8dc4b6Seschrock 1591ea8dc4b6Seschrock osname = spn.pn_path; 1592ea8dc4b6Seschrock 1593ecd6cf80Smarks /* 1594ecd6cf80Smarks * Check for mount privilege? 1595ecd6cf80Smarks * 1596ecd6cf80Smarks * If we don't have privilege then see if 1597ecd6cf80Smarks * we have local permission to allow it 1598ecd6cf80Smarks */ 1599ecd6cf80Smarks error = secpolicy_fs_mount(cr, mvp, vfsp); 1600ecd6cf80Smarks if (error) { 160198679b56SMark Shellenbaum if (dsl_deleg_access(osname, ZFS_DELEG_PERM_MOUNT, cr) == 0) { 1602ecd6cf80Smarks vattr_t vattr; 1603ecd6cf80Smarks 1604ecd6cf80Smarks /* 1605ecd6cf80Smarks * Make sure user is the owner of the mount point 1606ecd6cf80Smarks * or has sufficient privileges. 1607ecd6cf80Smarks */ 1608ecd6cf80Smarks 1609ecd6cf80Smarks vattr.va_mask = AT_UID; 1610ecd6cf80Smarks 161198679b56SMark Shellenbaum if (VOP_GETATTR(mvp, &vattr, 0, cr, NULL)) { 1612ecd6cf80Smarks goto out; 1613ecd6cf80Smarks } 1614ecd6cf80Smarks 16152459a9eaSmarks if (secpolicy_vnode_owner(cr, vattr.va_uid) != 0 && 16162459a9eaSmarks VOP_ACCESS(mvp, VWRITE, 0, cr, NULL) != 0) { 1617ecd6cf80Smarks goto out; 1618ecd6cf80Smarks } 1619ecd6cf80Smarks secpolicy_fs_mount_clearopts(cr, vfsp); 1620ecd6cf80Smarks } else { 1621ecd6cf80Smarks goto out; 1622ecd6cf80Smarks } 1623ecd6cf80Smarks } 1624ea8dc4b6Seschrock 1625ea8dc4b6Seschrock /* 1626ea8dc4b6Seschrock * Refuse to mount a filesystem if we are in a local zone and the 1627ea8dc4b6Seschrock * dataset is not visible. 1628ea8dc4b6Seschrock */ 1629ea8dc4b6Seschrock if (!INGLOBALZONE(curproc) && 1630ea8dc4b6Seschrock (!zone_dataset_visible(osname, &canwrite) || !canwrite)) { 1631ea8dc4b6Seschrock error = EPERM; 1632ea8dc4b6Seschrock goto out; 1633ea8dc4b6Seschrock } 1634ea8dc4b6Seschrock 16354201a95eSRic Aleshire error = zfs_mount_label_policy(vfsp, osname); 16364201a95eSRic Aleshire if (error) 16374201a95eSRic Aleshire goto out; 16384201a95eSRic Aleshire 1639b510d378Slling /* 1640b510d378Slling * When doing a remount, we simply refresh our temporary properties 1641b510d378Slling * according to those options set in the current VFS options. 1642b510d378Slling */ 1643b510d378Slling if (uap->flags & MS_REMOUNT) { 1644b510d378Slling /* refresh mount options */ 1645b510d378Slling zfs_unregister_callbacks(vfsp->vfs_data); 1646b510d378Slling error = zfs_register_callbacks(vfsp); 1647b510d378Slling goto out; 1648b510d378Slling } 1649b510d378Slling 1650088f3894Sahrens error = zfs_domount(vfsp, osname); 1651ea8dc4b6Seschrock 1652142ae85dSChris Kirby /* 1653142ae85dSChris Kirby * Add an extra VFS_HOLD on our parent vfs so that it can't 1654142ae85dSChris Kirby * disappear due to a forced unmount. 1655142ae85dSChris Kirby */ 1656984a131bSChris Kirby if (error == 0 && ((zfsvfs_t *)vfsp->vfs_data)->z_issnap) 1657142ae85dSChris Kirby VFS_HOLD(mvp->v_vfsp); 1658142ae85dSChris Kirby 1659ea8dc4b6Seschrock out: 1660fa9e4066Sahrens pn_free(&spn); 1661fa9e4066Sahrens return (error); 1662fa9e4066Sahrens } 1663fa9e4066Sahrens 1664fa9e4066Sahrens static int 1665fa9e4066Sahrens zfs_statvfs(vfs_t *vfsp, struct statvfs64 *statp) 1666fa9e4066Sahrens { 1667fa9e4066Sahrens zfsvfs_t *zfsvfs = vfsp->vfs_data; 1668fa9e4066Sahrens dev32_t d32; 1669a2eea2e1Sahrens uint64_t refdbytes, availbytes, usedobjs, availobjs; 1670fa9e4066Sahrens 1671fa9e4066Sahrens ZFS_ENTER(zfsvfs); 1672fa9e4066Sahrens 1673a2eea2e1Sahrens dmu_objset_space(zfsvfs->z_os, 1674a2eea2e1Sahrens &refdbytes, &availbytes, &usedobjs, &availobjs); 1675fa9e4066Sahrens 1676fa9e4066Sahrens /* 1677fa9e4066Sahrens * The underlying storage pool actually uses multiple block sizes. 1678fa9e4066Sahrens * We report the fragsize as the smallest block size we support, 1679fa9e4066Sahrens * and we report our blocksize as the filesystem's maximum blocksize. 1680fa9e4066Sahrens */ 1681fa9e4066Sahrens statp->f_frsize = 1UL << SPA_MINBLOCKSHIFT; 1682fa9e4066Sahrens statp->f_bsize = zfsvfs->z_max_blksz; 1683fa9e4066Sahrens 1684fa9e4066Sahrens /* 1685fa9e4066Sahrens * The following report "total" blocks of various kinds in the 1686fa9e4066Sahrens * file system, but reported in terms of f_frsize - the 1687fa9e4066Sahrens * "fragment" size. 1688fa9e4066Sahrens */ 1689fa9e4066Sahrens 1690a2eea2e1Sahrens statp->f_blocks = (refdbytes + availbytes) >> SPA_MINBLOCKSHIFT; 1691a2eea2e1Sahrens statp->f_bfree = availbytes >> SPA_MINBLOCKSHIFT; 1692fa9e4066Sahrens statp->f_bavail = statp->f_bfree; /* no root reservation */ 1693fa9e4066Sahrens 1694fa9e4066Sahrens /* 1695fa9e4066Sahrens * statvfs() should really be called statufs(), because it assumes 1696fa9e4066Sahrens * static metadata. ZFS doesn't preallocate files, so the best 1697fa9e4066Sahrens * we can do is report the max that could possibly fit in f_files, 1698fa9e4066Sahrens * and that minus the number actually used in f_ffree. 1699fa9e4066Sahrens * For f_ffree, report the smaller of the number of object available 1700fa9e4066Sahrens * and the number of blocks (each object will take at least a block). 1701fa9e4066Sahrens */ 1702a2eea2e1Sahrens statp->f_ffree = MIN(availobjs, statp->f_bfree); 1703fa9e4066Sahrens statp->f_favail = statp->f_ffree; /* no "root reservation" */ 1704a2eea2e1Sahrens statp->f_files = statp->f_ffree + usedobjs; 1705fa9e4066Sahrens 1706fa9e4066Sahrens (void) cmpldev(&d32, vfsp->vfs_dev); 1707fa9e4066Sahrens statp->f_fsid = d32; 1708fa9e4066Sahrens 1709fa9e4066Sahrens /* 1710fa9e4066Sahrens * We're a zfs filesystem. 1711fa9e4066Sahrens */ 1712fa9e4066Sahrens (void) strcpy(statp->f_basetype, vfssw[vfsp->vfs_fstype].vsw_name); 1713fa9e4066Sahrens 1714a5be7ebbSmarks statp->f_flag = vf_to_stf(vfsp->vfs_flag); 1715fa9e4066Sahrens 1716fa9e4066Sahrens statp->f_namemax = ZFS_MAXNAMELEN; 1717fa9e4066Sahrens 1718fa9e4066Sahrens /* 1719fa9e4066Sahrens * We have all of 32 characters to stuff a string here. 1720fa9e4066Sahrens * Is there anything useful we could/should provide? 1721fa9e4066Sahrens */ 1722fa9e4066Sahrens bzero(statp->f_fstr, sizeof (statp->f_fstr)); 1723fa9e4066Sahrens 1724fa9e4066Sahrens ZFS_EXIT(zfsvfs); 1725fa9e4066Sahrens return (0); 1726fa9e4066Sahrens } 1727fa9e4066Sahrens 1728fa9e4066Sahrens static int 1729fa9e4066Sahrens zfs_root(vfs_t *vfsp, vnode_t **vpp) 1730fa9e4066Sahrens { 1731fa9e4066Sahrens zfsvfs_t *zfsvfs = vfsp->vfs_data; 1732fa9e4066Sahrens znode_t *rootzp; 1733fa9e4066Sahrens int error; 1734fa9e4066Sahrens 1735fa9e4066Sahrens ZFS_ENTER(zfsvfs); 1736fa9e4066Sahrens 1737fa9e4066Sahrens error = zfs_zget(zfsvfs, zfsvfs->z_root, &rootzp); 1738fa9e4066Sahrens if (error == 0) 1739fa9e4066Sahrens *vpp = ZTOV(rootzp); 1740fa9e4066Sahrens 1741fa9e4066Sahrens ZFS_EXIT(zfsvfs); 1742fa9e4066Sahrens return (error); 1743fa9e4066Sahrens } 1744fa9e4066Sahrens 1745f18faf3fSek /* 1746f18faf3fSek * Teardown the zfsvfs::z_os. 1747f18faf3fSek * 1748f18faf3fSek * Note, if 'unmounting' if FALSE, we return with the 'z_teardown_lock' 1749f18faf3fSek * and 'z_teardown_inactive_lock' held. 1750f18faf3fSek */ 1751f18faf3fSek static int 1752f18faf3fSek zfsvfs_teardown(zfsvfs_t *zfsvfs, boolean_t unmounting) 1753f18faf3fSek { 1754874395d5Smaybee znode_t *zp; 1755f18faf3fSek 1756f18faf3fSek rrw_enter(&zfsvfs->z_teardown_lock, RW_WRITER, FTAG); 1757f18faf3fSek 1758f18faf3fSek if (!unmounting) { 1759f18faf3fSek /* 1760f18faf3fSek * We purge the parent filesystem's vfsp as the parent 1761f18faf3fSek * filesystem and all of its snapshots have their vnode's 1762f18faf3fSek * v_vfsp set to the parent's filesystem's vfsp. Note, 1763f18faf3fSek * 'z_parent' is self referential for non-snapshots. 1764f18faf3fSek */ 1765f18faf3fSek (void) dnlc_purge_vfsp(zfsvfs->z_parent->z_vfs, 0); 1766f18faf3fSek } 1767f18faf3fSek 1768f18faf3fSek /* 1769f18faf3fSek * Close the zil. NB: Can't close the zil while zfs_inactive 1770f18faf3fSek * threads are blocked as zil_close can call zfs_inactive. 1771f18faf3fSek */ 1772f18faf3fSek if (zfsvfs->z_log) { 1773f18faf3fSek zil_close(zfsvfs->z_log); 1774f18faf3fSek zfsvfs->z_log = NULL; 1775f18faf3fSek } 1776f18faf3fSek 1777f18faf3fSek rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_WRITER); 1778f18faf3fSek 1779f18faf3fSek /* 1780f18faf3fSek * If we are not unmounting (ie: online recv) and someone already 1781f18faf3fSek * unmounted this file system while we were doing the switcheroo, 1782f18faf3fSek * or a reopen of z_os failed then just bail out now. 1783f18faf3fSek */ 1784f18faf3fSek if (!unmounting && (zfsvfs->z_unmounted || zfsvfs->z_os == NULL)) { 1785f18faf3fSek rw_exit(&zfsvfs->z_teardown_inactive_lock); 1786f18faf3fSek rrw_exit(&zfsvfs->z_teardown_lock, FTAG); 1787f18faf3fSek return (EIO); 1788f18faf3fSek } 1789f18faf3fSek 1790f18faf3fSek /* 1791f18faf3fSek * At this point there are no vops active, and any new vops will 1792f18faf3fSek * fail with EIO since we have z_teardown_lock for writer (only 1793f18faf3fSek * relavent for forced unmount). 1794f18faf3fSek * 1795f18faf3fSek * Release all holds on dbufs. 1796f18faf3fSek */ 1797f18faf3fSek mutex_enter(&zfsvfs->z_znodes_lock); 1798874395d5Smaybee for (zp = list_head(&zfsvfs->z_all_znodes); zp != NULL; 1799874395d5Smaybee zp = list_next(&zfsvfs->z_all_znodes, zp)) 18000a586ceaSMark Shellenbaum if (zp->z_sa_hdl) { 1801874395d5Smaybee ASSERT(ZTOV(zp)->v_count > 0); 1802874395d5Smaybee zfs_znode_dmu_fini(zp); 1803f18faf3fSek } 1804f18faf3fSek mutex_exit(&zfsvfs->z_znodes_lock); 1805f18faf3fSek 1806f18faf3fSek /* 1807f18faf3fSek * If we are unmounting, set the unmounted flag and let new vops 1808f18faf3fSek * unblock. zfs_inactive will have the unmounted behavior, and all 1809f18faf3fSek * other vops will fail with EIO. 1810f18faf3fSek */ 1811f18faf3fSek if (unmounting) { 1812f18faf3fSek zfsvfs->z_unmounted = B_TRUE; 1813f18faf3fSek rrw_exit(&zfsvfs->z_teardown_lock, FTAG); 1814f18faf3fSek rw_exit(&zfsvfs->z_teardown_inactive_lock); 1815f18faf3fSek } 1816f18faf3fSek 1817f18faf3fSek /* 1818f18faf3fSek * z_os will be NULL if there was an error in attempting to reopen 1819f18faf3fSek * zfsvfs, so just return as the properties had already been 1820f18faf3fSek * unregistered and cached data had been evicted before. 1821f18faf3fSek */ 1822f18faf3fSek if (zfsvfs->z_os == NULL) 1823f18faf3fSek return (0); 1824f18faf3fSek 1825f18faf3fSek /* 1826f18faf3fSek * Unregister properties. 1827f18faf3fSek */ 1828f18faf3fSek zfs_unregister_callbacks(zfsvfs); 1829f18faf3fSek 1830f18faf3fSek /* 1831f18faf3fSek * Evict cached data 1832f18faf3fSek */ 18335aaeed5cSSanjeev Bagewadi if (dmu_objset_is_dirty_anywhere(zfsvfs->z_os)) 1834f9af39baSGeorge Wilson if (!(zfsvfs->z_vfs->vfs_flag & VFS_RDONLY)) 1835f9af39baSGeorge Wilson txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), 0); 18365aaeed5cSSanjeev Bagewadi (void) dmu_objset_evict_dbufs(zfsvfs->z_os); 1837f18faf3fSek 1838f18faf3fSek return (0); 1839f18faf3fSek } 1840f18faf3fSek 1841fa9e4066Sahrens /*ARGSUSED*/ 1842fa9e4066Sahrens static int 1843fa9e4066Sahrens zfs_umount(vfs_t *vfsp, int fflag, cred_t *cr) 1844fa9e4066Sahrens { 1845fa9e4066Sahrens zfsvfs_t *zfsvfs = vfsp->vfs_data; 1846f18faf3fSek objset_t *os; 1847fa9e4066Sahrens int ret; 1848fa9e4066Sahrens 1849ecd6cf80Smarks ret = secpolicy_fs_unmount(cr, vfsp); 1850ecd6cf80Smarks if (ret) { 185198679b56SMark Shellenbaum if (dsl_deleg_access((char *)refstr_value(vfsp->vfs_resource), 185298679b56SMark Shellenbaum ZFS_DELEG_PERM_MOUNT, cr)) 1853ecd6cf80Smarks return (ret); 1854ecd6cf80Smarks } 1855033f9833Sek 1856ed097989Sek /* 1857ed097989Sek * We purge the parent filesystem's vfsp as the parent filesystem 1858ed097989Sek * and all of its snapshots have their vnode's v_vfsp set to the 1859ed097989Sek * parent's filesystem's vfsp. Note, 'z_parent' is self 1860ed097989Sek * referential for non-snapshots. 1861ed097989Sek */ 1862ed097989Sek (void) dnlc_purge_vfsp(zfsvfs->z_parent->z_vfs, 0); 1863033f9833Sek 1864fa9e4066Sahrens /* 1865fa9e4066Sahrens * Unmount any snapshots mounted under .zfs before unmounting the 1866fa9e4066Sahrens * dataset itself. 1867fa9e4066Sahrens */ 1868fa9e4066Sahrens if (zfsvfs->z_ctldir != NULL && 1869ecd6cf80Smarks (ret = zfsctl_umount_snapshots(vfsp, fflag, cr)) != 0) { 1870fa9e4066Sahrens return (ret); 1871ecd6cf80Smarks } 1872fa9e4066Sahrens 187391ebeef5Sahrens if (!(fflag & MS_FORCE)) { 1874fa9e4066Sahrens /* 187591ebeef5Sahrens * Check the number of active vnodes in the file system. 187691ebeef5Sahrens * Our count is maintained in the vfs structure, but the 187791ebeef5Sahrens * number is off by 1 to indicate a hold on the vfs 187891ebeef5Sahrens * structure itself. 187991ebeef5Sahrens * 188091ebeef5Sahrens * The '.zfs' directory maintains a reference of its 188191ebeef5Sahrens * own, and any active references underneath are 188291ebeef5Sahrens * reflected in the vnode count. 1883fa9e4066Sahrens */ 188491ebeef5Sahrens if (zfsvfs->z_ctldir == NULL) { 188591ebeef5Sahrens if (vfsp->vfs_count > 1) 188691ebeef5Sahrens return (EBUSY); 188791ebeef5Sahrens } else { 188891ebeef5Sahrens if (vfsp->vfs_count > 2 || 1889f18faf3fSek zfsvfs->z_ctldir->v_count > 1) 189091ebeef5Sahrens return (EBUSY); 1891fa9e4066Sahrens } 189291ebeef5Sahrens } 1893fa9e4066Sahrens 189491ebeef5Sahrens vfsp->vfs_flag |= VFS_UNMOUNTED; 189591ebeef5Sahrens 1896f18faf3fSek VERIFY(zfsvfs_teardown(zfsvfs, B_TRUE) == 0); 1897f18faf3fSek os = zfsvfs->z_os; 189891ebeef5Sahrens 189991ebeef5Sahrens /* 1900f18faf3fSek * z_os will be NULL if there was an error in 1901f18faf3fSek * attempting to reopen zfsvfs. 190291ebeef5Sahrens */ 1903f18faf3fSek if (os != NULL) { 1904f18faf3fSek /* 1905f18faf3fSek * Unset the objset user_ptr. 1906f18faf3fSek */ 1907503ad85cSMatthew Ahrens mutex_enter(&os->os_user_ptr_lock); 1908f18faf3fSek dmu_objset_set_user(os, NULL); 1909503ad85cSMatthew Ahrens mutex_exit(&os->os_user_ptr_lock); 191091ebeef5Sahrens 1911f18faf3fSek /* 1912745cd3c5Smaybee * Finally release the objset 1913f18faf3fSek */ 1914503ad85cSMatthew Ahrens dmu_objset_disown(os, zfsvfs); 191591ebeef5Sahrens } 191691ebeef5Sahrens 191791ebeef5Sahrens /* 191891ebeef5Sahrens * We can now safely destroy the '.zfs' directory node. 191991ebeef5Sahrens */ 192091ebeef5Sahrens if (zfsvfs->z_ctldir != NULL) 192191ebeef5Sahrens zfsctl_destroy(zfsvfs); 1922fa9e4066Sahrens 1923fa9e4066Sahrens return (0); 1924fa9e4066Sahrens } 1925fa9e4066Sahrens 1926fa9e4066Sahrens static int 1927fa9e4066Sahrens zfs_vget(vfs_t *vfsp, vnode_t **vpp, fid_t *fidp) 1928fa9e4066Sahrens { 1929fa9e4066Sahrens zfsvfs_t *zfsvfs = vfsp->vfs_data; 1930fa9e4066Sahrens znode_t *zp; 1931fa9e4066Sahrens uint64_t object = 0; 1932fa9e4066Sahrens uint64_t fid_gen = 0; 1933fa9e4066Sahrens uint64_t gen_mask; 1934fa9e4066Sahrens uint64_t zp_gen; 1935fa9e4066Sahrens int i, err; 1936fa9e4066Sahrens 1937fa9e4066Sahrens *vpp = NULL; 1938fa9e4066Sahrens 1939fa9e4066Sahrens ZFS_ENTER(zfsvfs); 1940fa9e4066Sahrens 1941fa9e4066Sahrens if (fidp->fid_len == LONG_FID_LEN) { 1942fa9e4066Sahrens zfid_long_t *zlfid = (zfid_long_t *)fidp; 1943fa9e4066Sahrens uint64_t objsetid = 0; 1944fa9e4066Sahrens uint64_t setgen = 0; 1945fa9e4066Sahrens 1946fa9e4066Sahrens for (i = 0; i < sizeof (zlfid->zf_setid); i++) 1947fa9e4066Sahrens objsetid |= ((uint64_t)zlfid->zf_setid[i]) << (8 * i); 1948fa9e4066Sahrens 1949fa9e4066Sahrens for (i = 0; i < sizeof (zlfid->zf_setgen); i++) 1950fa9e4066Sahrens setgen |= ((uint64_t)zlfid->zf_setgen[i]) << (8 * i); 1951fa9e4066Sahrens 1952fa9e4066Sahrens ZFS_EXIT(zfsvfs); 1953fa9e4066Sahrens 1954fa9e4066Sahrens err = zfsctl_lookup_objset(vfsp, objsetid, &zfsvfs); 1955fa9e4066Sahrens if (err) 1956fa9e4066Sahrens return (EINVAL); 1957fa9e4066Sahrens ZFS_ENTER(zfsvfs); 1958fa9e4066Sahrens } 1959fa9e4066Sahrens 1960fa9e4066Sahrens if (fidp->fid_len == SHORT_FID_LEN || fidp->fid_len == LONG_FID_LEN) { 1961fa9e4066Sahrens zfid_short_t *zfid = (zfid_short_t *)fidp; 1962fa9e4066Sahrens 1963fa9e4066Sahrens for (i = 0; i < sizeof (zfid->zf_object); i++) 1964fa9e4066Sahrens object |= ((uint64_t)zfid->zf_object[i]) << (8 * i); 1965fa9e4066Sahrens 1966fa9e4066Sahrens for (i = 0; i < sizeof (zfid->zf_gen); i++) 1967fa9e4066Sahrens fid_gen |= ((uint64_t)zfid->zf_gen[i]) << (8 * i); 1968fa9e4066Sahrens } else { 1969fa9e4066Sahrens ZFS_EXIT(zfsvfs); 1970fa9e4066Sahrens return (EINVAL); 1971fa9e4066Sahrens } 1972fa9e4066Sahrens 1973fa9e4066Sahrens /* A zero fid_gen means we are in the .zfs control directories */ 1974fa9e4066Sahrens if (fid_gen == 0 && 1975fa9e4066Sahrens (object == ZFSCTL_INO_ROOT || object == ZFSCTL_INO_SNAPDIR)) { 1976fa9e4066Sahrens *vpp = zfsvfs->z_ctldir; 1977fa9e4066Sahrens ASSERT(*vpp != NULL); 1978fa9e4066Sahrens if (object == ZFSCTL_INO_SNAPDIR) { 1979fa9e4066Sahrens VERIFY(zfsctl_root_lookup(*vpp, "snapshot", vpp, NULL, 1980da6c28aaSamw 0, NULL, NULL, NULL, NULL, NULL) == 0); 1981fa9e4066Sahrens } else { 1982fa9e4066Sahrens VN_HOLD(*vpp); 1983fa9e4066Sahrens } 1984fa9e4066Sahrens ZFS_EXIT(zfsvfs); 1985fa9e4066Sahrens return (0); 1986fa9e4066Sahrens } 1987fa9e4066Sahrens 1988fa9e4066Sahrens gen_mask = -1ULL >> (64 - 8 * i); 1989fa9e4066Sahrens 1990fa9e4066Sahrens dprintf("getting %llu [%u mask %llx]\n", object, fid_gen, gen_mask); 1991fa9e4066Sahrens if (err = zfs_zget(zfsvfs, object, &zp)) { 1992fa9e4066Sahrens ZFS_EXIT(zfsvfs); 1993fa9e4066Sahrens return (err); 1994fa9e4066Sahrens } 19950a586ceaSMark Shellenbaum (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zfsvfs), &zp_gen, 19960a586ceaSMark Shellenbaum sizeof (uint64_t)); 19970a586ceaSMark Shellenbaum zp_gen = zp_gen & gen_mask; 1998fa9e4066Sahrens if (zp_gen == 0) 1999fa9e4066Sahrens zp_gen = 1; 2000893a6d32Sahrens if (zp->z_unlinked || zp_gen != fid_gen) { 2001fa9e4066Sahrens dprintf("znode gen (%u) != fid gen (%u)\n", zp_gen, fid_gen); 2002fa9e4066Sahrens VN_RELE(ZTOV(zp)); 2003fa9e4066Sahrens ZFS_EXIT(zfsvfs); 2004fa9e4066Sahrens return (EINVAL); 2005fa9e4066Sahrens } 2006fa9e4066Sahrens 2007fa9e4066Sahrens *vpp = ZTOV(zp); 2008fa9e4066Sahrens ZFS_EXIT(zfsvfs); 2009fa9e4066Sahrens return (0); 2010fa9e4066Sahrens } 2011fa9e4066Sahrens 2012f18faf3fSek /* 2013f18faf3fSek * Block out VOPs and close zfsvfs_t::z_os 2014f18faf3fSek * 2015f18faf3fSek * Note, if successful, then we return with the 'z_teardown_lock' and 2016f18faf3fSek * 'z_teardown_inactive_lock' write held. 2017f18faf3fSek */ 2018f18faf3fSek int 2019503ad85cSMatthew Ahrens zfs_suspend_fs(zfsvfs_t *zfsvfs) 2020f18faf3fSek { 2021f18faf3fSek int error; 2022f18faf3fSek 2023f18faf3fSek if ((error = zfsvfs_teardown(zfsvfs, B_FALSE)) != 0) 2024f18faf3fSek return (error); 2025503ad85cSMatthew Ahrens dmu_objset_disown(zfsvfs->z_os, zfsvfs); 2026f18faf3fSek 2027f18faf3fSek return (0); 2028f18faf3fSek } 2029f18faf3fSek 2030f18faf3fSek /* 2031f18faf3fSek * Reopen zfsvfs_t::z_os and release VOPs. 2032f18faf3fSek */ 2033f18faf3fSek int 2034503ad85cSMatthew Ahrens zfs_resume_fs(zfsvfs_t *zfsvfs, const char *osname) 2035f18faf3fSek { 203644bffe01SMark Shellenbaum int err; 2037f18faf3fSek 2038f18faf3fSek ASSERT(RRW_WRITE_HELD(&zfsvfs->z_teardown_lock)); 2039f18faf3fSek ASSERT(RW_WRITE_HELD(&zfsvfs->z_teardown_inactive_lock)); 2040f18faf3fSek 2041503ad85cSMatthew Ahrens err = dmu_objset_own(osname, DMU_OST_ZFS, B_FALSE, zfsvfs, 2042503ad85cSMatthew Ahrens &zfsvfs->z_os); 2043f18faf3fSek if (err) { 2044f18faf3fSek zfsvfs->z_os = NULL; 2045f18faf3fSek } else { 2046f18faf3fSek znode_t *zp; 20470a586ceaSMark Shellenbaum uint64_t sa_obj = 0; 20480a586ceaSMark Shellenbaum 204944bffe01SMark Shellenbaum /* 205044bffe01SMark Shellenbaum * Make sure version hasn't changed 205144bffe01SMark Shellenbaum */ 205244bffe01SMark Shellenbaum 205344bffe01SMark Shellenbaum err = zfs_get_zplprop(zfsvfs->z_os, ZFS_PROP_VERSION, 205444bffe01SMark Shellenbaum &zfsvfs->z_version); 20550a586ceaSMark Shellenbaum 205644bffe01SMark Shellenbaum if (err) 20570a586ceaSMark Shellenbaum goto bail; 20580a586ceaSMark Shellenbaum 205944bffe01SMark Shellenbaum err = zap_lookup(zfsvfs->z_os, MASTER_NODE_OBJ, 206044bffe01SMark Shellenbaum ZFS_SA_ATTRS, 8, 1, &sa_obj); 206144bffe01SMark Shellenbaum 206244bffe01SMark Shellenbaum if (err && zfsvfs->z_version >= ZPL_VERSION_SA) 206344bffe01SMark Shellenbaum goto bail; 20640a586ceaSMark Shellenbaum 20651d8ccc7bSMark Shellenbaum if ((err = sa_setup(zfsvfs->z_os, sa_obj, 20661d8ccc7bSMark Shellenbaum zfs_attr_table, ZPL_END, &zfsvfs->z_attr_table)) != 0) 20671d8ccc7bSMark Shellenbaum goto bail; 2068f18faf3fSek 206944bffe01SMark Shellenbaum if (zfsvfs->z_version >= ZPL_VERSION_SA) 207044bffe01SMark Shellenbaum sa_register_update_callback(zfsvfs->z_os, 207144bffe01SMark Shellenbaum zfs_sa_upgrade); 207244bffe01SMark Shellenbaum 2073f18faf3fSek VERIFY(zfsvfs_setup(zfsvfs, B_FALSE) == 0); 2074f18faf3fSek 207544bffe01SMark Shellenbaum zfs_set_fuid_feature(zfsvfs); 207644bffe01SMark Shellenbaum 2077f18faf3fSek /* 2078f18faf3fSek * Attempt to re-establish all the active znodes with 2079f18faf3fSek * their dbufs. If a zfs_rezget() fails, then we'll let 2080f18faf3fSek * any potential callers discover that via ZFS_ENTER_VERIFY_VP 2081f18faf3fSek * when they try to use their znode. 2082f18faf3fSek */ 2083f18faf3fSek mutex_enter(&zfsvfs->z_znodes_lock); 2084f18faf3fSek for (zp = list_head(&zfsvfs->z_all_znodes); zp; 2085f18faf3fSek zp = list_next(&zfsvfs->z_all_znodes, zp)) { 2086f18faf3fSek (void) zfs_rezget(zp); 2087f18faf3fSek } 2088f18faf3fSek mutex_exit(&zfsvfs->z_znodes_lock); 2089f18faf3fSek } 2090f18faf3fSek 20910a586ceaSMark Shellenbaum bail: 2092f18faf3fSek /* release the VOPs */ 2093f18faf3fSek rw_exit(&zfsvfs->z_teardown_inactive_lock); 2094f18faf3fSek rrw_exit(&zfsvfs->z_teardown_lock, FTAG); 2095f18faf3fSek 2096f18faf3fSek if (err) { 2097f18faf3fSek /* 209844bffe01SMark Shellenbaum * Since we couldn't reopen zfsvfs::z_os, or 209944bffe01SMark Shellenbaum * setup the sa framework force unmount this file system. 2100f18faf3fSek */ 2101f18faf3fSek if (vn_vfswlock(zfsvfs->z_vfs->vfs_vnodecovered) == 0) 2102f18faf3fSek (void) dounmount(zfsvfs->z_vfs, MS_FORCE, CRED()); 2103f18faf3fSek } 2104f18faf3fSek return (err); 2105f18faf3fSek } 2106f18faf3fSek 2107fa9e4066Sahrens static void 2108fa9e4066Sahrens zfs_freevfs(vfs_t *vfsp) 2109fa9e4066Sahrens { 2110fa9e4066Sahrens zfsvfs_t *zfsvfs = vfsp->vfs_data; 2111142ae85dSChris Kirby 2112142ae85dSChris Kirby /* 2113142ae85dSChris Kirby * If this is a snapshot, we have an extra VFS_HOLD on our parent 2114f80ce222SChris Kirby * from zfs_mount(). Release it here. If we came through 2115f80ce222SChris Kirby * zfs_mountroot() instead, we didn't grab an extra hold, so 2116f80ce222SChris Kirby * skip the VFS_RELE for rootvfs. 2117142ae85dSChris Kirby */ 2118f80ce222SChris Kirby if (zfsvfs->z_issnap && (vfsp != rootvfs)) 2119142ae85dSChris Kirby VFS_RELE(zfsvfs->z_parent->z_vfs); 2120142ae85dSChris Kirby 212114843421SMatthew Ahrens zfsvfs_free(zfsvfs); 2122fa9e4066Sahrens 2123fa9e4066Sahrens atomic_add_32(&zfs_active_fs_count, -1); 2124fa9e4066Sahrens } 2125fa9e4066Sahrens 2126fa9e4066Sahrens /* 2127fa9e4066Sahrens * VFS_INIT() initialization. Note that there is no VFS_FINI(), 2128fa9e4066Sahrens * so we can't safely do any non-idempotent initialization here. 2129fa9e4066Sahrens * Leave that to zfs_init() and zfs_fini(), which are called 2130fa9e4066Sahrens * from the module's _init() and _fini() entry points. 2131fa9e4066Sahrens */ 2132fa9e4066Sahrens /*ARGSUSED*/ 2133fa9e4066Sahrens static int 2134fa9e4066Sahrens zfs_vfsinit(int fstype, char *name) 2135fa9e4066Sahrens { 2136fa9e4066Sahrens int error; 2137fa9e4066Sahrens 2138fa9e4066Sahrens zfsfstype = fstype; 2139fa9e4066Sahrens 2140fa9e4066Sahrens /* 2141fa9e4066Sahrens * Setup vfsops and vnodeops tables. 2142fa9e4066Sahrens */ 2143fa9e4066Sahrens error = vfs_setfsops(fstype, zfs_vfsops_template, &zfs_vfsops); 2144fa9e4066Sahrens if (error != 0) { 2145fa9e4066Sahrens cmn_err(CE_WARN, "zfs: bad vfs ops template"); 2146fa9e4066Sahrens } 2147fa9e4066Sahrens 2148fa9e4066Sahrens error = zfs_create_op_tables(); 2149fa9e4066Sahrens if (error) { 2150fa9e4066Sahrens zfs_remove_op_tables(); 2151fa9e4066Sahrens cmn_err(CE_WARN, "zfs: bad vnode ops template"); 2152fa9e4066Sahrens (void) vfs_freevfsops_by_type(zfsfstype); 2153fa9e4066Sahrens return (error); 2154fa9e4066Sahrens } 2155fa9e4066Sahrens 2156fa9e4066Sahrens mutex_init(&zfs_dev_mtx, NULL, MUTEX_DEFAULT, NULL); 2157fa9e4066Sahrens 2158fa9e4066Sahrens /* 2159a0965f35Sbonwick * Unique major number for all zfs mounts. 2160a0965f35Sbonwick * If we run out of 32-bit minors, we'll getudev() another major. 2161fa9e4066Sahrens */ 2162a0965f35Sbonwick zfs_major = ddi_name_to_major(ZFS_DRIVER); 2163a0965f35Sbonwick zfs_minor = ZFS_MIN_MINOR; 2164fa9e4066Sahrens 2165fa9e4066Sahrens return (0); 2166fa9e4066Sahrens } 2167fa9e4066Sahrens 2168fa9e4066Sahrens void 2169fa9e4066Sahrens zfs_init(void) 2170fa9e4066Sahrens { 2171fa9e4066Sahrens /* 2172fa9e4066Sahrens * Initialize .zfs directory structures 2173fa9e4066Sahrens */ 2174fa9e4066Sahrens zfsctl_init(); 2175fa9e4066Sahrens 2176fa9e4066Sahrens /* 2177fa9e4066Sahrens * Initialize znode cache, vnode ops, etc... 2178fa9e4066Sahrens */ 2179fa9e4066Sahrens zfs_znode_init(); 218014843421SMatthew Ahrens 218114843421SMatthew Ahrens dmu_objset_register_type(DMU_OST_ZFS, zfs_space_delta_cb); 2182fa9e4066Sahrens } 2183fa9e4066Sahrens 2184fa9e4066Sahrens void 2185fa9e4066Sahrens zfs_fini(void) 2186fa9e4066Sahrens { 2187fa9e4066Sahrens zfsctl_fini(); 2188fa9e4066Sahrens zfs_znode_fini(); 2189fa9e4066Sahrens } 2190fa9e4066Sahrens 2191fa9e4066Sahrens int 2192fa9e4066Sahrens zfs_busy(void) 2193fa9e4066Sahrens { 2194fa9e4066Sahrens return (zfs_active_fs_count != 0); 2195fa9e4066Sahrens } 2196fa9e4066Sahrens 2197e7437265Sahrens int 219814843421SMatthew Ahrens zfs_set_version(zfsvfs_t *zfsvfs, uint64_t newvers) 2199e7437265Sahrens { 2200e7437265Sahrens int error; 220114843421SMatthew Ahrens objset_t *os = zfsvfs->z_os; 2202e7437265Sahrens dmu_tx_t *tx; 2203e7437265Sahrens 2204e7437265Sahrens if (newvers < ZPL_VERSION_INITIAL || newvers > ZPL_VERSION) 2205e7437265Sahrens return (EINVAL); 2206e7437265Sahrens 220714843421SMatthew Ahrens if (newvers < zfsvfs->z_version) 220814843421SMatthew Ahrens return (EINVAL); 2209e7437265Sahrens 22100a586ceaSMark Shellenbaum if (zfs_spa_version_map(newvers) > 22110a586ceaSMark Shellenbaum spa_version(dmu_objset_spa(zfsvfs->z_os))) 22120a586ceaSMark Shellenbaum return (ENOTSUP); 22130a586ceaSMark Shellenbaum 2214e7437265Sahrens tx = dmu_tx_create(os); 221514843421SMatthew Ahrens dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_FALSE, ZPL_VERSION_STR); 22160a586ceaSMark Shellenbaum if (newvers >= ZPL_VERSION_SA && !zfsvfs->z_use_sa) { 22170a586ceaSMark Shellenbaum dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE, 22180a586ceaSMark Shellenbaum ZFS_SA_ATTRS); 22190a586ceaSMark Shellenbaum dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); 22200a586ceaSMark Shellenbaum } 2221e7437265Sahrens error = dmu_tx_assign(tx, TXG_WAIT); 2222e7437265Sahrens if (error) { 2223e7437265Sahrens dmu_tx_abort(tx); 222414843421SMatthew Ahrens return (error); 222514843421SMatthew Ahrens } 22260a586ceaSMark Shellenbaum 222714843421SMatthew Ahrens error = zap_update(os, MASTER_NODE_OBJ, ZPL_VERSION_STR, 222814843421SMatthew Ahrens 8, 1, &newvers, tx); 222914843421SMatthew Ahrens 223014843421SMatthew Ahrens if (error) { 223114843421SMatthew Ahrens dmu_tx_commit(tx); 223214843421SMatthew Ahrens return (error); 2233e7437265Sahrens } 2234e7437265Sahrens 22350a586ceaSMark Shellenbaum if (newvers >= ZPL_VERSION_SA && !zfsvfs->z_use_sa) { 22360a586ceaSMark Shellenbaum uint64_t sa_obj; 22370a586ceaSMark Shellenbaum 22380a586ceaSMark Shellenbaum ASSERT3U(spa_version(dmu_objset_spa(zfsvfs->z_os)), >=, 22390a586ceaSMark Shellenbaum SPA_VERSION_SA); 22400a586ceaSMark Shellenbaum sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE, 22410a586ceaSMark Shellenbaum DMU_OT_NONE, 0, tx); 22420a586ceaSMark Shellenbaum 22430a586ceaSMark Shellenbaum error = zap_add(os, MASTER_NODE_OBJ, 22440a586ceaSMark Shellenbaum ZFS_SA_ATTRS, 8, 1, &sa_obj, tx); 22450a586ceaSMark Shellenbaum ASSERT3U(error, ==, 0); 22460a586ceaSMark Shellenbaum 22470a586ceaSMark Shellenbaum VERIFY(0 == sa_set_sa_object(os, sa_obj)); 22480a586ceaSMark Shellenbaum sa_register_update_callback(os, zfs_sa_upgrade); 22490a586ceaSMark Shellenbaum } 22500a586ceaSMark Shellenbaum 22513f9d6ad7SLin Ling spa_history_log_internal(LOG_DS_UPGRADE, 22523f9d6ad7SLin Ling dmu_objset_spa(os), tx, "oldver=%llu newver=%llu dataset = %llu", 225314843421SMatthew Ahrens zfsvfs->z_version, newvers, dmu_objset_id(os)); 225414843421SMatthew Ahrens 2255e7437265Sahrens dmu_tx_commit(tx); 2256e7437265Sahrens 225714843421SMatthew Ahrens zfsvfs->z_version = newvers; 225814843421SMatthew Ahrens 225944bffe01SMark Shellenbaum zfs_set_fuid_feature(zfsvfs); 226014843421SMatthew Ahrens 226114843421SMatthew Ahrens return (0); 2262e7437265Sahrens } 2263e7437265Sahrens 2264de8267e0Stimh /* 2265de8267e0Stimh * Read a property stored within the master node. 2266de8267e0Stimh */ 2267de8267e0Stimh int 2268de8267e0Stimh zfs_get_zplprop(objset_t *os, zfs_prop_t prop, uint64_t *value) 2269de8267e0Stimh { 2270de8267e0Stimh const char *pname; 22710a48a24eStimh int error = ENOENT; 2272de8267e0Stimh 2273de8267e0Stimh /* 2274de8267e0Stimh * Look up the file system's value for the property. For the 2275de8267e0Stimh * version property, we look up a slightly different string. 2276de8267e0Stimh */ 2277de8267e0Stimh if (prop == ZFS_PROP_VERSION) 2278de8267e0Stimh pname = ZPL_VERSION_STR; 2279de8267e0Stimh else 2280de8267e0Stimh pname = zfs_prop_to_name(prop); 2281de8267e0Stimh 22820a48a24eStimh if (os != NULL) 22830a48a24eStimh error = zap_lookup(os, MASTER_NODE_OBJ, pname, 8, 1, value); 2284de8267e0Stimh 2285aa60ed0eSmaybee if (error == ENOENT) { 2286de8267e0Stimh /* No value set, use the default value */ 2287de8267e0Stimh switch (prop) { 2288aa60ed0eSmaybee case ZFS_PROP_VERSION: 2289aa60ed0eSmaybee *value = ZPL_VERSION; 2290aa60ed0eSmaybee break; 2291de8267e0Stimh case ZFS_PROP_NORMALIZE: 2292de8267e0Stimh case ZFS_PROP_UTF8ONLY: 2293de8267e0Stimh *value = 0; 2294de8267e0Stimh break; 2295de8267e0Stimh case ZFS_PROP_CASE: 2296de8267e0Stimh *value = ZFS_CASE_SENSITIVE; 2297de8267e0Stimh break; 2298de8267e0Stimh default: 2299aa60ed0eSmaybee return (error); 2300de8267e0Stimh } 2301aa60ed0eSmaybee error = 0; 2302de8267e0Stimh } 2303aa60ed0eSmaybee return (error); 2304de8267e0Stimh } 2305de8267e0Stimh 2306fa9e4066Sahrens static vfsdef_t vfw = { 2307fa9e4066Sahrens VFSDEF_VERSION, 2308fa9e4066Sahrens MNTTYPE_ZFS, 2309fa9e4066Sahrens zfs_vfsinit, 2310da6c28aaSamw VSW_HASPROTO|VSW_CANRWRO|VSW_CANREMOUNT|VSW_VOLATILEDEV|VSW_STATS| 23110fbb751dSJohn Levon VSW_XID|VSW_ZMOUNT, 2312fa9e4066Sahrens &zfs_mntopts 2313fa9e4066Sahrens }; 2314fa9e4066Sahrens 2315fa9e4066Sahrens struct modlfs zfs_modlfs = { 2316e7437265Sahrens &mod_fsops, "ZFS filesystem version " SPA_VERSION_STRING, &vfw 2317fa9e4066Sahrens }; 2318