zfs_vfsops.c revision 243493
1168404Spjd/* 2168404Spjd * CDDL HEADER START 3168404Spjd * 4168404Spjd * The contents of this file are subject to the terms of the 5168404Spjd * Common Development and Distribution License (the "License"). 6168404Spjd * You may not use this file except in compliance with the License. 7168404Spjd * 8168404Spjd * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9168404Spjd * or http://www.opensolaris.org/os/licensing. 10168404Spjd * See the License for the specific language governing permissions 11168404Spjd * and limitations under the License. 12168404Spjd * 13168404Spjd * When distributing Covered Code, include this CDDL HEADER in each 14168404Spjd * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15168404Spjd * If applicable, add the following below this CDDL HEADER, with the 16168404Spjd * fields enclosed by brackets "[]" replaced with your own identifying 17168404Spjd * information: Portions Copyright [yyyy] [name of copyright owner] 18168404Spjd * 19168404Spjd * CDDL HEADER END 20168404Spjd */ 21168404Spjd/* 22219089Spjd * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 23229565Smm * Copyright (c) 2011 Pawel Jakub Dawidek <pawel@dawidek.net>. 24229565Smm * All rights reserved. 25168404Spjd */ 26168404Spjd 27219089Spjd/* Portions Copyright 2010 Robert Milkowski */ 28219089Spjd 29168404Spjd#include <sys/types.h> 30168404Spjd#include <sys/param.h> 31168404Spjd#include <sys/systm.h> 32168404Spjd#include <sys/kernel.h> 33168404Spjd#include <sys/sysmacros.h> 34168404Spjd#include <sys/kmem.h> 35168404Spjd#include <sys/acl.h> 36168404Spjd#include <sys/vnode.h> 37168404Spjd#include <sys/vfs.h> 38168404Spjd#include <sys/mntent.h> 39168404Spjd#include <sys/mount.h> 40168404Spjd#include <sys/cmn_err.h> 41168404Spjd#include <sys/zfs_znode.h> 42168404Spjd#include <sys/zfs_dir.h> 43168404Spjd#include <sys/zil.h> 44168404Spjd#include <sys/fs/zfs.h> 45168404Spjd#include <sys/dmu.h> 46168404Spjd#include <sys/dsl_prop.h> 47168404Spjd#include <sys/dsl_dataset.h> 48185029Spjd#include <sys/dsl_deleg.h> 49168404Spjd#include <sys/spa.h> 50168404Spjd#include <sys/zap.h> 51219089Spjd#include <sys/sa.h> 52168404Spjd#include <sys/varargs.h> 53168962Spjd#include <sys/policy.h> 54168404Spjd#include <sys/atomic.h> 55168404Spjd#include <sys/zfs_ioctl.h> 56168404Spjd#include <sys/zfs_ctldir.h> 57185029Spjd#include <sys/zfs_fuid.h> 58168962Spjd#include <sys/sunddi.h> 59168404Spjd#include <sys/dnlc.h> 60185029Spjd#include <sys/dmu_objset.h> 61185029Spjd#include <sys/spa_boot.h> 62219089Spjd#include <sys/sa.h> 63232728Smm#include <sys/jail.h> 64219089Spjd#include "zfs_comutil.h" 65168404Spjd 66168404Spjdstruct mtx zfs_debug_mtx; 67168404SpjdMTX_SYSINIT(zfs_debug_mtx, &zfs_debug_mtx, "zfs_debug", MTX_DEF); 68185029Spjd 69168404SpjdSYSCTL_NODE(_vfs, OID_AUTO, zfs, CTLFLAG_RW, 0, "ZFS file system"); 70185029Spjd 71219089Spjdint zfs_super_owner; 72185029SpjdSYSCTL_INT(_vfs_zfs, OID_AUTO, super_owner, CTLFLAG_RW, &zfs_super_owner, 0, 73185029Spjd "File system owner can perform privileged operation on his file systems"); 74185029Spjd 75219089Spjdint zfs_debug_level; 76168713SpjdTUNABLE_INT("vfs.zfs.debug", &zfs_debug_level); 77168404SpjdSYSCTL_INT(_vfs_zfs, OID_AUTO, debug, CTLFLAG_RW, &zfs_debug_level, 0, 78168404Spjd "Debug level"); 79168404Spjd 80185029SpjdSYSCTL_NODE(_vfs_zfs, OID_AUTO, version, CTLFLAG_RD, 0, "ZFS versions"); 81185029Spjdstatic int zfs_version_acl = ZFS_ACL_VERSION; 82185029SpjdSYSCTL_INT(_vfs_zfs_version, OID_AUTO, acl, CTLFLAG_RD, &zfs_version_acl, 0, 83185029Spjd "ZFS_ACL_VERSION"); 84185029Spjdstatic int zfs_version_spa = SPA_VERSION; 85185029SpjdSYSCTL_INT(_vfs_zfs_version, OID_AUTO, spa, CTLFLAG_RD, &zfs_version_spa, 0, 86185029Spjd "SPA_VERSION"); 87185029Spjdstatic int zfs_version_zpl = ZPL_VERSION; 88185029SpjdSYSCTL_INT(_vfs_zfs_version, OID_AUTO, zpl, CTLFLAG_RD, &zfs_version_zpl, 0, 89185029Spjd "ZPL_VERSION"); 90185029Spjd 91191990Sattiliostatic int zfs_mount(vfs_t *vfsp); 92191990Sattiliostatic int zfs_umount(vfs_t *vfsp, int fflag); 93191990Sattiliostatic int zfs_root(vfs_t *vfsp, int flags, vnode_t **vpp); 94191990Sattiliostatic int zfs_statfs(vfs_t *vfsp, struct statfs *statp); 95168404Spjdstatic int zfs_vget(vfs_t *vfsp, ino_t ino, int flags, vnode_t **vpp); 96191990Sattiliostatic int zfs_sync(vfs_t *vfsp, int waitfor); 97196982Spjdstatic int zfs_checkexp(vfs_t *vfsp, struct sockaddr *nam, int *extflagsp, 98196982Spjd struct ucred **credanonp, int *numsecflavors, int **secflavors); 99222167Srmacklemstatic int zfs_fhtovp(vfs_t *vfsp, fid_t *fidp, int flags, vnode_t **vpp); 100168404Spjdstatic void zfs_objset_close(zfsvfs_t *zfsvfs); 101168404Spjdstatic void zfs_freevfs(vfs_t *vfsp); 102168404Spjd 103168404Spjdstatic struct vfsops zfs_vfsops = { 104168404Spjd .vfs_mount = zfs_mount, 105168404Spjd .vfs_unmount = zfs_umount, 106168404Spjd .vfs_root = zfs_root, 107168404Spjd .vfs_statfs = zfs_statfs, 108168404Spjd .vfs_vget = zfs_vget, 109168404Spjd .vfs_sync = zfs_sync, 110196982Spjd .vfs_checkexp = zfs_checkexp, 111168404Spjd .vfs_fhtovp = zfs_fhtovp, 112168404Spjd}; 113168404Spjd 114185029SpjdVFS_SET(zfs_vfsops, zfs, VFCF_JAIL | VFCF_DELEGADMIN); 115168404Spjd 116168404Spjd/* 117168404Spjd * We need to keep a count of active fs's. 118168404Spjd * This is necessary to prevent our module 119168404Spjd * from being unloaded after a umount -f 120168404Spjd */ 121168404Spjdstatic uint32_t zfs_active_fs_count = 0; 122168404Spjd 123168404Spjd/*ARGSUSED*/ 124168404Spjdstatic int 125191990Sattiliozfs_sync(vfs_t *vfsp, int waitfor) 126168404Spjd{ 127168404Spjd 128168404Spjd /* 129168404Spjd * Data integrity is job one. We don't want a compromised kernel 130168404Spjd * writing to the storage pool, so we never sync during panic. 131168404Spjd */ 132168404Spjd if (panicstr) 133168404Spjd return (0); 134168404Spjd 135168404Spjd if (vfsp != NULL) { 136168404Spjd /* 137168404Spjd * Sync a specific filesystem. 138168404Spjd */ 139168404Spjd zfsvfs_t *zfsvfs = vfsp->vfs_data; 140209962Smm dsl_pool_t *dp; 141168404Spjd int error; 142168404Spjd 143191990Sattilio error = vfs_stdsync(vfsp, waitfor); 144168404Spjd if (error != 0) 145168404Spjd return (error); 146168404Spjd 147168404Spjd ZFS_ENTER(zfsvfs); 148209962Smm dp = dmu_objset_pool(zfsvfs->z_os); 149209962Smm 150209962Smm /* 151209962Smm * If the system is shutting down, then skip any 152209962Smm * filesystems which may exist on a suspended pool. 153209962Smm */ 154209962Smm if (sys_shutdown && spa_suspended(dp->dp_spa)) { 155209962Smm ZFS_EXIT(zfsvfs); 156209962Smm return (0); 157209962Smm } 158209962Smm 159168404Spjd if (zfsvfs->z_log != NULL) 160219089Spjd zil_commit(zfsvfs->z_log, 0); 161219089Spjd 162168404Spjd ZFS_EXIT(zfsvfs); 163168404Spjd } else { 164168404Spjd /* 165168404Spjd * Sync all ZFS filesystems. This is what happens when you 166168404Spjd * run sync(1M). Unlike other filesystems, ZFS honors the 167168404Spjd * request by waiting for all pools to commit all dirty data. 168168404Spjd */ 169168404Spjd spa_sync_allpools(); 170168404Spjd } 171168404Spjd 172168404Spjd return (0); 173168404Spjd} 174168404Spjd 175219089Spjd#ifndef __FreeBSD__ 176219089Spjdstatic int 177219089Spjdzfs_create_unique_device(dev_t *dev) 178219089Spjd{ 179219089Spjd major_t new_major; 180219089Spjd 181219089Spjd do { 182219089Spjd ASSERT3U(zfs_minor, <=, MAXMIN32); 183219089Spjd minor_t start = zfs_minor; 184219089Spjd do { 185219089Spjd mutex_enter(&zfs_dev_mtx); 186219089Spjd if (zfs_minor >= MAXMIN32) { 187219089Spjd /* 188219089Spjd * If we're still using the real major 189219089Spjd * keep out of /dev/zfs and /dev/zvol minor 190219089Spjd * number space. If we're using a getudev()'ed 191219089Spjd * major number, we can use all of its minors. 192219089Spjd */ 193219089Spjd if (zfs_major == ddi_name_to_major(ZFS_DRIVER)) 194219089Spjd zfs_minor = ZFS_MIN_MINOR; 195219089Spjd else 196219089Spjd zfs_minor = 0; 197219089Spjd } else { 198219089Spjd zfs_minor++; 199219089Spjd } 200219089Spjd *dev = makedevice(zfs_major, zfs_minor); 201219089Spjd mutex_exit(&zfs_dev_mtx); 202219089Spjd } while (vfs_devismounted(*dev) && zfs_minor != start); 203219089Spjd if (zfs_minor == start) { 204219089Spjd /* 205219089Spjd * We are using all ~262,000 minor numbers for the 206219089Spjd * current major number. Create a new major number. 207219089Spjd */ 208219089Spjd if ((new_major = getudev()) == (major_t)-1) { 209219089Spjd cmn_err(CE_WARN, 210219089Spjd "zfs_mount: Can't get unique major " 211219089Spjd "device number."); 212219089Spjd return (-1); 213219089Spjd } 214219089Spjd mutex_enter(&zfs_dev_mtx); 215219089Spjd zfs_major = new_major; 216219089Spjd zfs_minor = 0; 217219089Spjd 218219089Spjd mutex_exit(&zfs_dev_mtx); 219219089Spjd } else { 220219089Spjd break; 221219089Spjd } 222219089Spjd /* CONSTANTCONDITION */ 223219089Spjd } while (1); 224219089Spjd 225219089Spjd return (0); 226219089Spjd} 227219089Spjd#endif /* !__FreeBSD__ */ 228219089Spjd 229168404Spjdstatic void 230168404Spjdatime_changed_cb(void *arg, uint64_t newval) 231168404Spjd{ 232168404Spjd zfsvfs_t *zfsvfs = arg; 233168404Spjd 234168404Spjd if (newval == TRUE) { 235168404Spjd zfsvfs->z_atime = TRUE; 236168404Spjd zfsvfs->z_vfs->vfs_flag &= ~MNT_NOATIME; 237168404Spjd vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NOATIME); 238168404Spjd vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_ATIME, NULL, 0); 239168404Spjd } else { 240168404Spjd zfsvfs->z_atime = FALSE; 241168404Spjd zfsvfs->z_vfs->vfs_flag |= MNT_NOATIME; 242168404Spjd vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_ATIME); 243168404Spjd vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NOATIME, NULL, 0); 244168404Spjd } 245168404Spjd} 246168404Spjd 247168404Spjdstatic void 248168404Spjdxattr_changed_cb(void *arg, uint64_t newval) 249168404Spjd{ 250168404Spjd zfsvfs_t *zfsvfs = arg; 251168404Spjd 252168404Spjd if (newval == TRUE) { 253168404Spjd /* XXX locking on vfs_flag? */ 254168404Spjd#ifdef TODO 255168404Spjd zfsvfs->z_vfs->vfs_flag |= VFS_XATTR; 256168404Spjd#endif 257168404Spjd vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NOXATTR); 258168404Spjd vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_XATTR, NULL, 0); 259168404Spjd } else { 260168404Spjd /* XXX locking on vfs_flag? */ 261168404Spjd#ifdef TODO 262168404Spjd zfsvfs->z_vfs->vfs_flag &= ~VFS_XATTR; 263168404Spjd#endif 264168404Spjd vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_XATTR); 265168404Spjd vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NOXATTR, NULL, 0); 266168404Spjd } 267168404Spjd} 268168404Spjd 269168404Spjdstatic void 270168404Spjdblksz_changed_cb(void *arg, uint64_t newval) 271168404Spjd{ 272168404Spjd zfsvfs_t *zfsvfs = arg; 273168404Spjd 274168404Spjd if (newval < SPA_MINBLOCKSIZE || 275168404Spjd newval > SPA_MAXBLOCKSIZE || !ISP2(newval)) 276168404Spjd newval = SPA_MAXBLOCKSIZE; 277168404Spjd 278168404Spjd zfsvfs->z_max_blksz = newval; 279204101Spjd zfsvfs->z_vfs->mnt_stat.f_iosize = newval; 280168404Spjd} 281168404Spjd 282168404Spjdstatic void 283168404Spjdreadonly_changed_cb(void *arg, uint64_t newval) 284168404Spjd{ 285168404Spjd zfsvfs_t *zfsvfs = arg; 286168404Spjd 287168404Spjd if (newval) { 288168404Spjd /* XXX locking on vfs_flag? */ 289168404Spjd zfsvfs->z_vfs->vfs_flag |= VFS_RDONLY; 290168404Spjd vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_RW); 291168404Spjd vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_RO, NULL, 0); 292168404Spjd } else { 293168404Spjd /* XXX locking on vfs_flag? */ 294168404Spjd zfsvfs->z_vfs->vfs_flag &= ~VFS_RDONLY; 295168404Spjd vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_RO); 296168404Spjd vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_RW, NULL, 0); 297168404Spjd } 298168404Spjd} 299168404Spjd 300168404Spjdstatic void 301168404Spjdsetuid_changed_cb(void *arg, uint64_t newval) 302168404Spjd{ 303168404Spjd zfsvfs_t *zfsvfs = arg; 304168404Spjd 305168404Spjd if (newval == FALSE) { 306168404Spjd zfsvfs->z_vfs->vfs_flag |= VFS_NOSETUID; 307168404Spjd vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_SETUID); 308168404Spjd vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NOSETUID, NULL, 0); 309168404Spjd } else { 310168404Spjd zfsvfs->z_vfs->vfs_flag &= ~VFS_NOSETUID; 311168404Spjd vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NOSETUID); 312168404Spjd vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_SETUID, NULL, 0); 313168404Spjd } 314168404Spjd} 315168404Spjd 316168404Spjdstatic void 317168404Spjdexec_changed_cb(void *arg, uint64_t newval) 318168404Spjd{ 319168404Spjd zfsvfs_t *zfsvfs = arg; 320168404Spjd 321168404Spjd if (newval == FALSE) { 322168404Spjd zfsvfs->z_vfs->vfs_flag |= VFS_NOEXEC; 323168404Spjd vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_EXEC); 324168404Spjd vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NOEXEC, NULL, 0); 325168404Spjd } else { 326168404Spjd zfsvfs->z_vfs->vfs_flag &= ~VFS_NOEXEC; 327168404Spjd vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NOEXEC); 328168404Spjd vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_EXEC, NULL, 0); 329168404Spjd } 330168404Spjd} 331168404Spjd 332185029Spjd/* 333185029Spjd * The nbmand mount option can be changed at mount time. 334185029Spjd * We can't allow it to be toggled on live file systems or incorrect 335185029Spjd * behavior may be seen from cifs clients 336185029Spjd * 337185029Spjd * This property isn't registered via dsl_prop_register(), but this callback 338185029Spjd * will be called when a file system is first mounted 339185029Spjd */ 340168404Spjdstatic void 341185029Spjdnbmand_changed_cb(void *arg, uint64_t newval) 342185029Spjd{ 343185029Spjd zfsvfs_t *zfsvfs = arg; 344185029Spjd if (newval == FALSE) { 345185029Spjd vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NBMAND); 346185029Spjd vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NONBMAND, NULL, 0); 347185029Spjd } else { 348185029Spjd vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NONBMAND); 349185029Spjd vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NBMAND, NULL, 0); 350185029Spjd } 351185029Spjd} 352185029Spjd 353185029Spjdstatic void 354168404Spjdsnapdir_changed_cb(void *arg, uint64_t newval) 355168404Spjd{ 356168404Spjd zfsvfs_t *zfsvfs = arg; 357168404Spjd 358168404Spjd zfsvfs->z_show_ctldir = newval; 359168404Spjd} 360168404Spjd 361168404Spjdstatic void 362185029Spjdvscan_changed_cb(void *arg, uint64_t newval) 363185029Spjd{ 364185029Spjd zfsvfs_t *zfsvfs = arg; 365185029Spjd 366185029Spjd zfsvfs->z_vscan = newval; 367185029Spjd} 368185029Spjd 369185029Spjdstatic void 370224174Smmacl_mode_changed_cb(void *arg, uint64_t newval) 371224174Smm{ 372224174Smm zfsvfs_t *zfsvfs = arg; 373224174Smm 374224174Smm zfsvfs->z_acl_mode = newval; 375224174Smm} 376224174Smm 377224174Smmstatic void 378168404Spjdacl_inherit_changed_cb(void *arg, uint64_t newval) 379168404Spjd{ 380168404Spjd zfsvfs_t *zfsvfs = arg; 381168404Spjd 382168404Spjd zfsvfs->z_acl_inherit = newval; 383168404Spjd} 384168404Spjd 385168404Spjdstatic int 386168404Spjdzfs_register_callbacks(vfs_t *vfsp) 387168404Spjd{ 388168404Spjd struct dsl_dataset *ds = NULL; 389168404Spjd objset_t *os = NULL; 390168404Spjd zfsvfs_t *zfsvfs = NULL; 391185029Spjd uint64_t nbmand; 392219089Spjd int readonly, do_readonly = B_FALSE; 393219089Spjd int setuid, do_setuid = B_FALSE; 394219089Spjd int exec, do_exec = B_FALSE; 395219089Spjd int xattr, do_xattr = B_FALSE; 396219089Spjd int atime, do_atime = B_FALSE; 397168404Spjd int error = 0; 398168404Spjd 399168404Spjd ASSERT(vfsp); 400168404Spjd zfsvfs = vfsp->vfs_data; 401168404Spjd ASSERT(zfsvfs); 402168404Spjd os = zfsvfs->z_os; 403168404Spjd 404168404Spjd /* 405196965Spjd * This function can be called for a snapshot when we update snapshot's 406196965Spjd * mount point, which isn't really supported. 407196965Spjd */ 408196965Spjd if (dmu_objset_is_snapshot(os)) 409196965Spjd return (EOPNOTSUPP); 410196965Spjd 411196965Spjd /* 412168404Spjd * The act of registering our callbacks will destroy any mount 413168404Spjd * options we may have. In order to enable temporary overrides 414168404Spjd * of mount options, we stash away the current values and 415168404Spjd * restore them after we register the callbacks. 416168404Spjd */ 417219089Spjd if (vfs_optionisset(vfsp, MNTOPT_RO, NULL) || 418219089Spjd !spa_writeable(dmu_objset_spa(os))) { 419168404Spjd readonly = B_TRUE; 420168404Spjd do_readonly = B_TRUE; 421168404Spjd } else if (vfs_optionisset(vfsp, MNTOPT_RW, NULL)) { 422168404Spjd readonly = B_FALSE; 423168404Spjd do_readonly = B_TRUE; 424168404Spjd } 425168404Spjd if (vfs_optionisset(vfsp, MNTOPT_NOSUID, NULL)) { 426168404Spjd setuid = B_FALSE; 427168404Spjd do_setuid = B_TRUE; 428168404Spjd } else { 429168404Spjd if (vfs_optionisset(vfsp, MNTOPT_NOSETUID, NULL)) { 430168404Spjd setuid = B_FALSE; 431168404Spjd do_setuid = B_TRUE; 432168404Spjd } else if (vfs_optionisset(vfsp, MNTOPT_SETUID, NULL)) { 433168404Spjd setuid = B_TRUE; 434168404Spjd do_setuid = B_TRUE; 435168404Spjd } 436168404Spjd } 437168404Spjd if (vfs_optionisset(vfsp, MNTOPT_NOEXEC, NULL)) { 438168404Spjd exec = B_FALSE; 439168404Spjd do_exec = B_TRUE; 440168404Spjd } else if (vfs_optionisset(vfsp, MNTOPT_EXEC, NULL)) { 441168404Spjd exec = B_TRUE; 442168404Spjd do_exec = B_TRUE; 443168404Spjd } 444168404Spjd if (vfs_optionisset(vfsp, MNTOPT_NOXATTR, NULL)) { 445168404Spjd xattr = B_FALSE; 446168404Spjd do_xattr = B_TRUE; 447168404Spjd } else if (vfs_optionisset(vfsp, MNTOPT_XATTR, NULL)) { 448168404Spjd xattr = B_TRUE; 449168404Spjd do_xattr = B_TRUE; 450168404Spjd } 451185029Spjd if (vfs_optionisset(vfsp, MNTOPT_NOATIME, NULL)) { 452185029Spjd atime = B_FALSE; 453185029Spjd do_atime = B_TRUE; 454185029Spjd } else if (vfs_optionisset(vfsp, MNTOPT_ATIME, NULL)) { 455185029Spjd atime = B_TRUE; 456185029Spjd do_atime = B_TRUE; 457185029Spjd } 458168404Spjd 459168404Spjd /* 460185029Spjd * nbmand is a special property. It can only be changed at 461185029Spjd * mount time. 462185029Spjd * 463185029Spjd * This is weird, but it is documented to only be changeable 464185029Spjd * at mount time. 465185029Spjd */ 466185029Spjd if (vfs_optionisset(vfsp, MNTOPT_NONBMAND, NULL)) { 467185029Spjd nbmand = B_FALSE; 468185029Spjd } else if (vfs_optionisset(vfsp, MNTOPT_NBMAND, NULL)) { 469185029Spjd nbmand = B_TRUE; 470185029Spjd } else { 471185029Spjd char osname[MAXNAMELEN]; 472185029Spjd 473185029Spjd dmu_objset_name(os, osname); 474185029Spjd if (error = dsl_prop_get_integer(osname, "nbmand", &nbmand, 475185029Spjd NULL)) { 476185029Spjd return (error); 477185029Spjd } 478185029Spjd } 479185029Spjd 480185029Spjd /* 481168404Spjd * Register property callbacks. 482168404Spjd * 483168404Spjd * It would probably be fine to just check for i/o error from 484168404Spjd * the first prop_register(), but I guess I like to go 485168404Spjd * overboard... 486168404Spjd */ 487168404Spjd ds = dmu_objset_ds(os); 488168404Spjd error = dsl_prop_register(ds, "atime", atime_changed_cb, zfsvfs); 489168404Spjd error = error ? error : dsl_prop_register(ds, 490168404Spjd "xattr", xattr_changed_cb, zfsvfs); 491168404Spjd error = error ? error : dsl_prop_register(ds, 492168404Spjd "recordsize", blksz_changed_cb, zfsvfs); 493168404Spjd error = error ? error : dsl_prop_register(ds, 494168404Spjd "readonly", readonly_changed_cb, zfsvfs); 495168404Spjd error = error ? error : dsl_prop_register(ds, 496168404Spjd "setuid", setuid_changed_cb, zfsvfs); 497168404Spjd error = error ? error : dsl_prop_register(ds, 498168404Spjd "exec", exec_changed_cb, zfsvfs); 499168404Spjd error = error ? error : dsl_prop_register(ds, 500168404Spjd "snapdir", snapdir_changed_cb, zfsvfs); 501168404Spjd error = error ? error : dsl_prop_register(ds, 502224174Smm "aclmode", acl_mode_changed_cb, zfsvfs); 503224174Smm error = error ? error : dsl_prop_register(ds, 504168404Spjd "aclinherit", acl_inherit_changed_cb, zfsvfs); 505185029Spjd error = error ? error : dsl_prop_register(ds, 506185029Spjd "vscan", vscan_changed_cb, zfsvfs); 507168404Spjd if (error) 508168404Spjd goto unregister; 509168404Spjd 510168404Spjd /* 511168404Spjd * Invoke our callbacks to restore temporary mount options. 512168404Spjd */ 513168404Spjd if (do_readonly) 514168404Spjd readonly_changed_cb(zfsvfs, readonly); 515168404Spjd if (do_setuid) 516168404Spjd setuid_changed_cb(zfsvfs, setuid); 517168404Spjd if (do_exec) 518168404Spjd exec_changed_cb(zfsvfs, exec); 519168404Spjd if (do_xattr) 520168404Spjd xattr_changed_cb(zfsvfs, xattr); 521185029Spjd if (do_atime) 522185029Spjd atime_changed_cb(zfsvfs, atime); 523168404Spjd 524185029Spjd nbmand_changed_cb(zfsvfs, nbmand); 525185029Spjd 526168404Spjd return (0); 527168404Spjd 528168404Spjdunregister: 529168404Spjd /* 530168404Spjd * We may attempt to unregister some callbacks that are not 531168404Spjd * registered, but this is OK; it will simply return ENOMSG, 532168404Spjd * which we will ignore. 533168404Spjd */ 534168404Spjd (void) dsl_prop_unregister(ds, "atime", atime_changed_cb, zfsvfs); 535168404Spjd (void) dsl_prop_unregister(ds, "xattr", xattr_changed_cb, zfsvfs); 536168404Spjd (void) dsl_prop_unregister(ds, "recordsize", blksz_changed_cb, zfsvfs); 537168404Spjd (void) dsl_prop_unregister(ds, "readonly", readonly_changed_cb, zfsvfs); 538168404Spjd (void) dsl_prop_unregister(ds, "setuid", setuid_changed_cb, zfsvfs); 539168404Spjd (void) dsl_prop_unregister(ds, "exec", exec_changed_cb, zfsvfs); 540168404Spjd (void) dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb, zfsvfs); 541224174Smm (void) dsl_prop_unregister(ds, "aclmode", acl_mode_changed_cb, zfsvfs); 542168404Spjd (void) dsl_prop_unregister(ds, "aclinherit", acl_inherit_changed_cb, 543168404Spjd zfsvfs); 544185029Spjd (void) dsl_prop_unregister(ds, "vscan", vscan_changed_cb, zfsvfs); 545168404Spjd return (error); 546168404Spjd 547168404Spjd} 548168404Spjd 549219089Spjdstatic int 550219089Spjdzfs_space_delta_cb(dmu_object_type_t bonustype, void *data, 551219089Spjd uint64_t *userp, uint64_t *groupp) 552209962Smm{ 553219089Spjd znode_phys_t *znp = data; 554219089Spjd int error = 0; 555209962Smm 556219089Spjd /* 557219089Spjd * Is it a valid type of object to track? 558219089Spjd */ 559219089Spjd if (bonustype != DMU_OT_ZNODE && bonustype != DMU_OT_SA) 560219089Spjd return (ENOENT); 561209962Smm 562219089Spjd /* 563219089Spjd * If we have a NULL data pointer 564219089Spjd * then assume the id's aren't changing and 565219089Spjd * return EEXIST to the dmu to let it know to 566219089Spjd * use the same ids 567219089Spjd */ 568219089Spjd if (data == NULL) 569219089Spjd return (EEXIST); 570209962Smm 571219089Spjd if (bonustype == DMU_OT_ZNODE) { 572219089Spjd *userp = znp->zp_uid; 573219089Spjd *groupp = znp->zp_gid; 574219089Spjd } else { 575219089Spjd int hdrsize; 576209962Smm 577219089Spjd ASSERT(bonustype == DMU_OT_SA); 578219089Spjd hdrsize = sa_hdrsize(data); 579209962Smm 580219089Spjd if (hdrsize != 0) { 581219089Spjd *userp = *((uint64_t *)((uintptr_t)data + hdrsize + 582219089Spjd SA_UID_OFFSET)); 583219089Spjd *groupp = *((uint64_t *)((uintptr_t)data + hdrsize + 584219089Spjd SA_GID_OFFSET)); 585219089Spjd } else { 586219089Spjd /* 587219089Spjd * This should only happen for newly created 588219089Spjd * files that haven't had the znode data filled 589219089Spjd * in yet. 590219089Spjd */ 591219089Spjd *userp = 0; 592219089Spjd *groupp = 0; 593219089Spjd } 594209962Smm } 595219089Spjd return (error); 596209962Smm} 597209962Smm 598209962Smmstatic void 599209962Smmfuidstr_to_sid(zfsvfs_t *zfsvfs, const char *fuidstr, 600209962Smm char *domainbuf, int buflen, uid_t *ridp) 601209962Smm{ 602209962Smm uint64_t fuid; 603209962Smm const char *domain; 604209962Smm 605209962Smm fuid = strtonum(fuidstr, NULL); 606209962Smm 607209962Smm domain = zfs_fuid_find_by_idx(zfsvfs, FUID_INDEX(fuid)); 608209962Smm if (domain) 609209962Smm (void) strlcpy(domainbuf, domain, buflen); 610209962Smm else 611209962Smm domainbuf[0] = '\0'; 612209962Smm *ridp = FUID_RID(fuid); 613209962Smm} 614209962Smm 615209962Smmstatic uint64_t 616209962Smmzfs_userquota_prop_to_obj(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type) 617209962Smm{ 618209962Smm switch (type) { 619209962Smm case ZFS_PROP_USERUSED: 620209962Smm return (DMU_USERUSED_OBJECT); 621209962Smm case ZFS_PROP_GROUPUSED: 622209962Smm return (DMU_GROUPUSED_OBJECT); 623209962Smm case ZFS_PROP_USERQUOTA: 624209962Smm return (zfsvfs->z_userquota_obj); 625209962Smm case ZFS_PROP_GROUPQUOTA: 626209962Smm return (zfsvfs->z_groupquota_obj); 627209962Smm } 628209962Smm return (0); 629209962Smm} 630209962Smm 631209962Smmint 632209962Smmzfs_userspace_many(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type, 633209962Smm uint64_t *cookiep, void *vbuf, uint64_t *bufsizep) 634209962Smm{ 635209962Smm int error; 636209962Smm zap_cursor_t zc; 637209962Smm zap_attribute_t za; 638209962Smm zfs_useracct_t *buf = vbuf; 639209962Smm uint64_t obj; 640209962Smm 641209962Smm if (!dmu_objset_userspace_present(zfsvfs->z_os)) 642209962Smm return (ENOTSUP); 643209962Smm 644209962Smm obj = zfs_userquota_prop_to_obj(zfsvfs, type); 645209962Smm if (obj == 0) { 646209962Smm *bufsizep = 0; 647209962Smm return (0); 648209962Smm } 649209962Smm 650209962Smm for (zap_cursor_init_serialized(&zc, zfsvfs->z_os, obj, *cookiep); 651209962Smm (error = zap_cursor_retrieve(&zc, &za)) == 0; 652209962Smm zap_cursor_advance(&zc)) { 653209962Smm if ((uintptr_t)buf - (uintptr_t)vbuf + sizeof (zfs_useracct_t) > 654209962Smm *bufsizep) 655209962Smm break; 656209962Smm 657209962Smm fuidstr_to_sid(zfsvfs, za.za_name, 658209962Smm buf->zu_domain, sizeof (buf->zu_domain), &buf->zu_rid); 659209962Smm 660209962Smm buf->zu_space = za.za_first_integer; 661209962Smm buf++; 662209962Smm } 663209962Smm if (error == ENOENT) 664209962Smm error = 0; 665209962Smm 666209962Smm ASSERT3U((uintptr_t)buf - (uintptr_t)vbuf, <=, *bufsizep); 667209962Smm *bufsizep = (uintptr_t)buf - (uintptr_t)vbuf; 668209962Smm *cookiep = zap_cursor_serialize(&zc); 669209962Smm zap_cursor_fini(&zc); 670209962Smm return (error); 671209962Smm} 672209962Smm 673209962Smm/* 674209962Smm * buf must be big enough (eg, 32 bytes) 675209962Smm */ 676168404Spjdstatic int 677209962Smmid_to_fuidstr(zfsvfs_t *zfsvfs, const char *domain, uid_t rid, 678209962Smm char *buf, boolean_t addok) 679209962Smm{ 680209962Smm uint64_t fuid; 681209962Smm int domainid = 0; 682209962Smm 683209962Smm if (domain && domain[0]) { 684209962Smm domainid = zfs_fuid_find_by_domain(zfsvfs, domain, NULL, addok); 685209962Smm if (domainid == -1) 686209962Smm return (ENOENT); 687209962Smm } 688209962Smm fuid = FUID_ENCODE(domainid, rid); 689209962Smm (void) sprintf(buf, "%llx", (longlong_t)fuid); 690209962Smm return (0); 691209962Smm} 692209962Smm 693209962Smmint 694209962Smmzfs_userspace_one(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type, 695209962Smm const char *domain, uint64_t rid, uint64_t *valp) 696209962Smm{ 697209962Smm char buf[32]; 698209962Smm int err; 699209962Smm uint64_t obj; 700209962Smm 701209962Smm *valp = 0; 702209962Smm 703209962Smm if (!dmu_objset_userspace_present(zfsvfs->z_os)) 704209962Smm return (ENOTSUP); 705209962Smm 706209962Smm obj = zfs_userquota_prop_to_obj(zfsvfs, type); 707209962Smm if (obj == 0) 708209962Smm return (0); 709209962Smm 710209962Smm err = id_to_fuidstr(zfsvfs, domain, rid, buf, B_FALSE); 711209962Smm if (err) 712209962Smm return (err); 713209962Smm 714209962Smm err = zap_lookup(zfsvfs->z_os, obj, buf, 8, 1, valp); 715209962Smm if (err == ENOENT) 716209962Smm err = 0; 717209962Smm return (err); 718209962Smm} 719209962Smm 720209962Smmint 721209962Smmzfs_set_userquota(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type, 722209962Smm const char *domain, uint64_t rid, uint64_t quota) 723209962Smm{ 724209962Smm char buf[32]; 725209962Smm int err; 726209962Smm dmu_tx_t *tx; 727209962Smm uint64_t *objp; 728209962Smm boolean_t fuid_dirtied; 729209962Smm 730209962Smm if (type != ZFS_PROP_USERQUOTA && type != ZFS_PROP_GROUPQUOTA) 731209962Smm return (EINVAL); 732209962Smm 733209962Smm if (zfsvfs->z_version < ZPL_VERSION_USERSPACE) 734209962Smm return (ENOTSUP); 735209962Smm 736209962Smm objp = (type == ZFS_PROP_USERQUOTA) ? &zfsvfs->z_userquota_obj : 737209962Smm &zfsvfs->z_groupquota_obj; 738209962Smm 739209962Smm err = id_to_fuidstr(zfsvfs, domain, rid, buf, B_TRUE); 740209962Smm if (err) 741209962Smm return (err); 742209962Smm fuid_dirtied = zfsvfs->z_fuid_dirty; 743209962Smm 744209962Smm tx = dmu_tx_create(zfsvfs->z_os); 745209962Smm dmu_tx_hold_zap(tx, *objp ? *objp : DMU_NEW_OBJECT, B_TRUE, NULL); 746209962Smm if (*objp == 0) { 747209962Smm dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE, 748209962Smm zfs_userquota_prop_prefixes[type]); 749209962Smm } 750209962Smm if (fuid_dirtied) 751209962Smm zfs_fuid_txhold(zfsvfs, tx); 752209962Smm err = dmu_tx_assign(tx, TXG_WAIT); 753209962Smm if (err) { 754209962Smm dmu_tx_abort(tx); 755209962Smm return (err); 756209962Smm } 757209962Smm 758209962Smm mutex_enter(&zfsvfs->z_lock); 759209962Smm if (*objp == 0) { 760209962Smm *objp = zap_create(zfsvfs->z_os, DMU_OT_USERGROUP_QUOTA, 761209962Smm DMU_OT_NONE, 0, tx); 762209962Smm VERIFY(0 == zap_add(zfsvfs->z_os, MASTER_NODE_OBJ, 763209962Smm zfs_userquota_prop_prefixes[type], 8, 1, objp, tx)); 764209962Smm } 765209962Smm mutex_exit(&zfsvfs->z_lock); 766209962Smm 767209962Smm if (quota == 0) { 768209962Smm err = zap_remove(zfsvfs->z_os, *objp, buf, tx); 769209962Smm if (err == ENOENT) 770209962Smm err = 0; 771209962Smm } else { 772209962Smm err = zap_update(zfsvfs->z_os, *objp, buf, 8, 1, "a, tx); 773209962Smm } 774209962Smm ASSERT(err == 0); 775209962Smm if (fuid_dirtied) 776209962Smm zfs_fuid_sync(zfsvfs, tx); 777209962Smm dmu_tx_commit(tx); 778209962Smm return (err); 779209962Smm} 780209962Smm 781209962Smmboolean_t 782219089Spjdzfs_fuid_overquota(zfsvfs_t *zfsvfs, boolean_t isgroup, uint64_t fuid) 783209962Smm{ 784209962Smm char buf[32]; 785209962Smm uint64_t used, quota, usedobj, quotaobj; 786209962Smm int err; 787209962Smm 788209962Smm usedobj = isgroup ? DMU_GROUPUSED_OBJECT : DMU_USERUSED_OBJECT; 789209962Smm quotaobj = isgroup ? zfsvfs->z_groupquota_obj : zfsvfs->z_userquota_obj; 790209962Smm 791209962Smm if (quotaobj == 0 || zfsvfs->z_replay) 792209962Smm return (B_FALSE); 793209962Smm 794209962Smm (void) sprintf(buf, "%llx", (longlong_t)fuid); 795209962Smm err = zap_lookup(zfsvfs->z_os, quotaobj, buf, 8, 1, "a); 796209962Smm if (err != 0) 797209962Smm return (B_FALSE); 798209962Smm 799209962Smm err = zap_lookup(zfsvfs->z_os, usedobj, buf, 8, 1, &used); 800209962Smm if (err != 0) 801209962Smm return (B_FALSE); 802209962Smm return (used >= quota); 803209962Smm} 804209962Smm 805219089Spjdboolean_t 806219089Spjdzfs_owner_overquota(zfsvfs_t *zfsvfs, znode_t *zp, boolean_t isgroup) 807219089Spjd{ 808219089Spjd uint64_t fuid; 809219089Spjd uint64_t quotaobj; 810219089Spjd 811219089Spjd quotaobj = isgroup ? zfsvfs->z_groupquota_obj : zfsvfs->z_userquota_obj; 812219089Spjd 813219089Spjd fuid = isgroup ? zp->z_gid : zp->z_uid; 814219089Spjd 815219089Spjd if (quotaobj == 0 || zfsvfs->z_replay) 816219089Spjd return (B_FALSE); 817219089Spjd 818219089Spjd return (zfs_fuid_overquota(zfsvfs, isgroup, fuid)); 819219089Spjd} 820219089Spjd 821209962Smmint 822219089Spjdzfsvfs_create(const char *osname, zfsvfs_t **zfvp) 823209962Smm{ 824209962Smm objset_t *os; 825209962Smm zfsvfs_t *zfsvfs; 826209962Smm uint64_t zval; 827209962Smm int i, error; 828219089Spjd uint64_t sa_obj; 829209962Smm 830219089Spjd zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP); 831219089Spjd 832219089Spjd /* 833219089Spjd * We claim to always be readonly so we can open snapshots; 834219089Spjd * other ZPL code will prevent us from writing to snapshots. 835219089Spjd */ 836219089Spjd error = dmu_objset_own(osname, DMU_OST_ZFS, B_TRUE, zfsvfs, &os); 837219089Spjd if (error) { 838219089Spjd kmem_free(zfsvfs, sizeof (zfsvfs_t)); 839209962Smm return (error); 840209962Smm } 841209962Smm 842209962Smm /* 843209962Smm * Initialize the zfs-specific filesystem structure. 844209962Smm * Should probably make this a kmem cache, shuffle fields, 845209962Smm * and just bzero up to z_hold_mtx[]. 846209962Smm */ 847209962Smm zfsvfs->z_vfs = NULL; 848209962Smm zfsvfs->z_parent = zfsvfs; 849209962Smm zfsvfs->z_max_blksz = SPA_MAXBLOCKSIZE; 850209962Smm zfsvfs->z_show_ctldir = ZFS_SNAPDIR_VISIBLE; 851209962Smm zfsvfs->z_os = os; 852209962Smm 853209962Smm error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zfsvfs->z_version); 854209962Smm if (error) { 855209962Smm goto out; 856219089Spjd } else if (zfsvfs->z_version > 857219089Spjd zfs_zpl_version_map(spa_version(dmu_objset_spa(os)))) { 858219089Spjd (void) printf("Can't mount a version %lld file system " 859219089Spjd "on a version %lld pool\n. Pool must be upgraded to mount " 860219089Spjd "this file system.", (u_longlong_t)zfsvfs->z_version, 861219089Spjd (u_longlong_t)spa_version(dmu_objset_spa(os))); 862209962Smm error = ENOTSUP; 863209962Smm goto out; 864209962Smm } 865209962Smm if ((error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &zval)) != 0) 866209962Smm goto out; 867209962Smm zfsvfs->z_norm = (int)zval; 868209962Smm 869209962Smm if ((error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &zval)) != 0) 870209962Smm goto out; 871209962Smm zfsvfs->z_utf8 = (zval != 0); 872209962Smm 873209962Smm if ((error = zfs_get_zplprop(os, ZFS_PROP_CASE, &zval)) != 0) 874209962Smm goto out; 875209962Smm zfsvfs->z_case = (uint_t)zval; 876209962Smm 877209962Smm /* 878209962Smm * Fold case on file systems that are always or sometimes case 879209962Smm * insensitive. 880209962Smm */ 881209962Smm if (zfsvfs->z_case == ZFS_CASE_INSENSITIVE || 882209962Smm zfsvfs->z_case == ZFS_CASE_MIXED) 883209962Smm zfsvfs->z_norm |= U8_TEXTPREP_TOUPPER; 884209962Smm 885209962Smm zfsvfs->z_use_fuids = USE_FUIDS(zfsvfs->z_version, zfsvfs->z_os); 886219089Spjd zfsvfs->z_use_sa = USE_SA(zfsvfs->z_version, zfsvfs->z_os); 887209962Smm 888219089Spjd if (zfsvfs->z_use_sa) { 889219089Spjd /* should either have both of these objects or none */ 890219089Spjd error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, 891219089Spjd &sa_obj); 892219089Spjd if (error) 893219089Spjd return (error); 894219089Spjd } else { 895219089Spjd /* 896219089Spjd * Pre SA versions file systems should never touch 897219089Spjd * either the attribute registration or layout objects. 898219089Spjd */ 899219089Spjd sa_obj = 0; 900219089Spjd } 901219089Spjd 902219089Spjd error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END, 903219089Spjd &zfsvfs->z_attr_table); 904219089Spjd if (error) 905219089Spjd goto out; 906219089Spjd 907219089Spjd if (zfsvfs->z_version >= ZPL_VERSION_SA) 908219089Spjd sa_register_update_callback(os, zfs_sa_upgrade); 909219089Spjd 910209962Smm error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1, 911209962Smm &zfsvfs->z_root); 912209962Smm if (error) 913209962Smm goto out; 914209962Smm ASSERT(zfsvfs->z_root != 0); 915209962Smm 916209962Smm error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1, 917209962Smm &zfsvfs->z_unlinkedobj); 918209962Smm if (error) 919209962Smm goto out; 920209962Smm 921209962Smm error = zap_lookup(os, MASTER_NODE_OBJ, 922209962Smm zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA], 923209962Smm 8, 1, &zfsvfs->z_userquota_obj); 924209962Smm if (error && error != ENOENT) 925209962Smm goto out; 926209962Smm 927209962Smm error = zap_lookup(os, MASTER_NODE_OBJ, 928209962Smm zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA], 929209962Smm 8, 1, &zfsvfs->z_groupquota_obj); 930209962Smm if (error && error != ENOENT) 931209962Smm goto out; 932209962Smm 933209962Smm error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1, 934209962Smm &zfsvfs->z_fuid_obj); 935209962Smm if (error && error != ENOENT) 936209962Smm goto out; 937209962Smm 938209962Smm error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SHARES_DIR, 8, 1, 939209962Smm &zfsvfs->z_shares_dir); 940209962Smm if (error && error != ENOENT) 941209962Smm goto out; 942209962Smm 943209962Smm mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); 944209962Smm mutex_init(&zfsvfs->z_lock, NULL, MUTEX_DEFAULT, NULL); 945209962Smm list_create(&zfsvfs->z_all_znodes, sizeof (znode_t), 946209962Smm offsetof(znode_t, z_link_node)); 947209962Smm rrw_init(&zfsvfs->z_teardown_lock); 948209962Smm rw_init(&zfsvfs->z_teardown_inactive_lock, NULL, RW_DEFAULT, NULL); 949209962Smm rw_init(&zfsvfs->z_fuid_lock, NULL, RW_DEFAULT, NULL); 950209962Smm for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) 951209962Smm mutex_init(&zfsvfs->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL); 952209962Smm 953219089Spjd *zfvp = zfsvfs; 954209962Smm return (0); 955209962Smm 956209962Smmout: 957219089Spjd dmu_objset_disown(os, zfsvfs); 958219089Spjd *zfvp = NULL; 959209962Smm kmem_free(zfsvfs, sizeof (zfsvfs_t)); 960209962Smm return (error); 961209962Smm} 962209962Smm 963209962Smmstatic int 964185029Spjdzfsvfs_setup(zfsvfs_t *zfsvfs, boolean_t mounting) 965168404Spjd{ 966185029Spjd int error; 967185029Spjd 968185029Spjd error = zfs_register_callbacks(zfsvfs->z_vfs); 969185029Spjd if (error) 970185029Spjd return (error); 971185029Spjd 972185029Spjd /* 973185029Spjd * Set the objset user_ptr to track its zfsvfs. 974185029Spjd */ 975219089Spjd mutex_enter(&zfsvfs->z_os->os_user_ptr_lock); 976185029Spjd dmu_objset_set_user(zfsvfs->z_os, zfsvfs); 977219089Spjd mutex_exit(&zfsvfs->z_os->os_user_ptr_lock); 978185029Spjd 979208689Smm zfsvfs->z_log = zil_open(zfsvfs->z_os, zfs_get_data); 980208689Smm 981185029Spjd /* 982185029Spjd * If we are not mounting (ie: online recv), then we don't 983185029Spjd * have to worry about replaying the log as we blocked all 984185029Spjd * operations out since we closed the ZIL. 985185029Spjd */ 986185029Spjd if (mounting) { 987185029Spjd boolean_t readonly; 988185029Spjd 989185029Spjd /* 990185029Spjd * During replay we remove the read only flag to 991185029Spjd * allow replays to succeed. 992185029Spjd */ 993185029Spjd readonly = zfsvfs->z_vfs->vfs_flag & VFS_RDONLY; 994208689Smm if (readonly != 0) 995208689Smm zfsvfs->z_vfs->vfs_flag &= ~VFS_RDONLY; 996208689Smm else 997208689Smm zfs_unlinked_drain(zfsvfs); 998185029Spjd 999219089Spjd /* 1000219089Spjd * Parse and replay the intent log. 1001219089Spjd * 1002219089Spjd * Because of ziltest, this must be done after 1003219089Spjd * zfs_unlinked_drain(). (Further note: ziltest 1004219089Spjd * doesn't use readonly mounts, where 1005219089Spjd * zfs_unlinked_drain() isn't called.) This is because 1006219089Spjd * ziltest causes spa_sync() to think it's committed, 1007219089Spjd * but actually it is not, so the intent log contains 1008219089Spjd * many txg's worth of changes. 1009219089Spjd * 1010219089Spjd * In particular, if object N is in the unlinked set in 1011219089Spjd * the last txg to actually sync, then it could be 1012219089Spjd * actually freed in a later txg and then reallocated 1013219089Spjd * in a yet later txg. This would write a "create 1014219089Spjd * object N" record to the intent log. Normally, this 1015219089Spjd * would be fine because the spa_sync() would have 1016219089Spjd * written out the fact that object N is free, before 1017219089Spjd * we could write the "create object N" intent log 1018219089Spjd * record. 1019219089Spjd * 1020219089Spjd * But when we are in ziltest mode, we advance the "open 1021219089Spjd * txg" without actually spa_sync()-ing the changes to 1022219089Spjd * disk. So we would see that object N is still 1023219089Spjd * allocated and in the unlinked set, and there is an 1024219089Spjd * intent log record saying to allocate it. 1025219089Spjd */ 1026219089Spjd if (spa_writeable(dmu_objset_spa(zfsvfs->z_os))) { 1027219089Spjd if (zil_replay_disable) { 1028219089Spjd zil_destroy(zfsvfs->z_log, B_FALSE); 1029219089Spjd } else { 1030219089Spjd zfsvfs->z_replay = B_TRUE; 1031219089Spjd zil_replay(zfsvfs->z_os, zfsvfs, 1032219089Spjd zfs_replay_vector); 1033219089Spjd zfsvfs->z_replay = B_FALSE; 1034219089Spjd } 1035208689Smm } 1036185029Spjd zfsvfs->z_vfs->vfs_flag |= readonly; /* restore readonly bit */ 1037185029Spjd } 1038185029Spjd 1039185029Spjd return (0); 1040185029Spjd} 1041185029Spjd 1042210470Smmextern krwlock_t zfsvfs_lock; /* in zfs_znode.c */ 1043210470Smm 1044209962Smmvoid 1045209962Smmzfsvfs_free(zfsvfs_t *zfsvfs) 1046185029Spjd{ 1047209962Smm int i; 1048209962Smm 1049210470Smm /* 1050210470Smm * This is a barrier to prevent the filesystem from going away in 1051210470Smm * zfs_znode_move() until we can safely ensure that the filesystem is 1052210470Smm * not unmounted. We consider the filesystem valid before the barrier 1053210470Smm * and invalid after the barrier. 1054210470Smm */ 1055210470Smm rw_enter(&zfsvfs_lock, RW_READER); 1056210470Smm rw_exit(&zfsvfs_lock); 1057210470Smm 1058209962Smm zfs_fuid_destroy(zfsvfs); 1059209962Smm 1060185029Spjd mutex_destroy(&zfsvfs->z_znodes_lock); 1061209962Smm mutex_destroy(&zfsvfs->z_lock); 1062185029Spjd list_destroy(&zfsvfs->z_all_znodes); 1063185029Spjd rrw_destroy(&zfsvfs->z_teardown_lock); 1064185029Spjd rw_destroy(&zfsvfs->z_teardown_inactive_lock); 1065185029Spjd rw_destroy(&zfsvfs->z_fuid_lock); 1066209962Smm for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) 1067209962Smm mutex_destroy(&zfsvfs->z_hold_mtx[i]); 1068185029Spjd kmem_free(zfsvfs, sizeof (zfsvfs_t)); 1069185029Spjd} 1070185029Spjd 1071209962Smmstatic void 1072209962Smmzfs_set_fuid_feature(zfsvfs_t *zfsvfs) 1073209962Smm{ 1074209962Smm zfsvfs->z_use_fuids = USE_FUIDS(zfsvfs->z_version, zfsvfs->z_os); 1075219089Spjd if (zfsvfs->z_vfs) { 1076219089Spjd if (zfsvfs->z_use_fuids) { 1077219089Spjd vfs_set_feature(zfsvfs->z_vfs, VFSFT_XVATTR); 1078219089Spjd vfs_set_feature(zfsvfs->z_vfs, VFSFT_SYSATTR_VIEWS); 1079219089Spjd vfs_set_feature(zfsvfs->z_vfs, VFSFT_ACEMASKONACCESS); 1080219089Spjd vfs_set_feature(zfsvfs->z_vfs, VFSFT_ACLONCREATE); 1081219089Spjd vfs_set_feature(zfsvfs->z_vfs, VFSFT_ACCESS_FILTER); 1082219089Spjd vfs_set_feature(zfsvfs->z_vfs, VFSFT_REPARSE); 1083219089Spjd } else { 1084219089Spjd vfs_clear_feature(zfsvfs->z_vfs, VFSFT_XVATTR); 1085219089Spjd vfs_clear_feature(zfsvfs->z_vfs, VFSFT_SYSATTR_VIEWS); 1086219089Spjd vfs_clear_feature(zfsvfs->z_vfs, VFSFT_ACEMASKONACCESS); 1087219089Spjd vfs_clear_feature(zfsvfs->z_vfs, VFSFT_ACLONCREATE); 1088219089Spjd vfs_clear_feature(zfsvfs->z_vfs, VFSFT_ACCESS_FILTER); 1089219089Spjd vfs_clear_feature(zfsvfs->z_vfs, VFSFT_REPARSE); 1090219089Spjd } 1091209962Smm } 1092219089Spjd zfsvfs->z_use_sa = USE_SA(zfsvfs->z_version, zfsvfs->z_os); 1093209962Smm} 1094209962Smm 1095185029Spjdstatic int 1096185029Spjdzfs_domount(vfs_t *vfsp, char *osname) 1097185029Spjd{ 1098209962Smm uint64_t recordsize, fsid_guid; 1099168404Spjd int error = 0; 1100168404Spjd zfsvfs_t *zfsvfs; 1101209962Smm vnode_t *vp; 1102168404Spjd 1103168404Spjd ASSERT(vfsp); 1104168404Spjd ASSERT(osname); 1105168404Spjd 1106219089Spjd error = zfsvfs_create(osname, &zfsvfs); 1107209962Smm if (error) 1108209962Smm return (error); 1109168404Spjd zfsvfs->z_vfs = vfsp; 1110168404Spjd 1111168404Spjd if (error = dsl_prop_get_integer(osname, "recordsize", &recordsize, 1112168404Spjd NULL)) 1113168404Spjd goto out; 1114204101Spjd zfsvfs->z_vfs->vfs_bsize = SPA_MINBLOCKSIZE; 1115204101Spjd zfsvfs->z_vfs->mnt_stat.f_iosize = recordsize; 1116168404Spjd 1117168404Spjd vfsp->vfs_data = zfsvfs; 1118218386Strasz vfsp->mnt_flag |= MNT_LOCAL; 1119168404Spjd vfsp->mnt_kern_flag |= MNTK_MPSAFE; 1120168404Spjd vfsp->mnt_kern_flag |= MNTK_LOOKUP_SHARED; 1121193440Sps vfsp->mnt_kern_flag |= MNTK_SHARED_WRITES; 1122243493Savg vfsp->mnt_kern_flag |= MNTK_EXTENDED_SHARED; 1123168404Spjd 1124209962Smm /* 1125209962Smm * The fsid is 64 bits, composed of an 8-bit fs type, which 1126209962Smm * separates our fsid from any other filesystem types, and a 1127209962Smm * 56-bit objset unique ID. The objset unique ID is unique to 1128209962Smm * all objsets open on this system, provided by unique_create(). 1129209962Smm * The 8-bit fs type must be put in the low bits of fsid[1] 1130209962Smm * because that's where other Solaris filesystems put it. 1131209962Smm */ 1132209962Smm fsid_guid = dmu_objset_fsid_guid(zfsvfs->z_os); 1133209962Smm ASSERT((fsid_guid & ~((1ULL<<56)-1)) == 0); 1134209962Smm vfsp->vfs_fsid.val[0] = fsid_guid; 1135209962Smm vfsp->vfs_fsid.val[1] = ((fsid_guid>>32) << 8) | 1136209962Smm vfsp->mnt_vfc->vfc_typenum & 0xFF; 1137168404Spjd 1138185029Spjd /* 1139185029Spjd * Set features for file system. 1140185029Spjd */ 1141209962Smm zfs_set_fuid_feature(zfsvfs); 1142185029Spjd if (zfsvfs->z_case == ZFS_CASE_INSENSITIVE) { 1143185029Spjd vfs_set_feature(vfsp, VFSFT_DIRENTFLAGS); 1144185029Spjd vfs_set_feature(vfsp, VFSFT_CASEINSENSITIVE); 1145185029Spjd vfs_set_feature(vfsp, VFSFT_NOCASESENSITIVE); 1146185029Spjd } else if (zfsvfs->z_case == ZFS_CASE_MIXED) { 1147185029Spjd vfs_set_feature(vfsp, VFSFT_DIRENTFLAGS); 1148185029Spjd vfs_set_feature(vfsp, VFSFT_CASEINSENSITIVE); 1149185029Spjd } 1150219089Spjd vfs_set_feature(vfsp, VFSFT_ZEROCOPY_SUPPORTED); 1151185029Spjd 1152168404Spjd if (dmu_objset_is_snapshot(zfsvfs->z_os)) { 1153185029Spjd uint64_t pval; 1154168404Spjd 1155168404Spjd atime_changed_cb(zfsvfs, B_FALSE); 1156168404Spjd readonly_changed_cb(zfsvfs, B_TRUE); 1157185029Spjd if (error = dsl_prop_get_integer(osname, "xattr", &pval, NULL)) 1158168404Spjd goto out; 1159185029Spjd xattr_changed_cb(zfsvfs, pval); 1160168404Spjd zfsvfs->z_issnap = B_TRUE; 1161219089Spjd zfsvfs->z_os->os_sync = ZFS_SYNC_DISABLED; 1162209962Smm 1163219089Spjd mutex_enter(&zfsvfs->z_os->os_user_ptr_lock); 1164209962Smm dmu_objset_set_user(zfsvfs->z_os, zfsvfs); 1165219089Spjd mutex_exit(&zfsvfs->z_os->os_user_ptr_lock); 1166168404Spjd } else { 1167185029Spjd error = zfsvfs_setup(zfsvfs, B_TRUE); 1168168404Spjd } 1169168404Spjd 1170168404Spjd vfs_mountedfrom(vfsp, osname); 1171209962Smm /* Grab extra reference. */ 1172209962Smm VERIFY(VFS_ROOT(vfsp, LK_EXCLUSIVE, &vp) == 0); 1173209962Smm VOP_UNLOCK(vp, 0); 1174168404Spjd 1175168404Spjd if (!zfsvfs->z_issnap) 1176168404Spjd zfsctl_create(zfsvfs); 1177168404Spjdout: 1178168404Spjd if (error) { 1179219089Spjd dmu_objset_disown(zfsvfs->z_os, zfsvfs); 1180209962Smm zfsvfs_free(zfsvfs); 1181168404Spjd } else { 1182168404Spjd atomic_add_32(&zfs_active_fs_count, 1); 1183168404Spjd } 1184168404Spjd 1185168404Spjd return (error); 1186168404Spjd} 1187168404Spjd 1188168404Spjdvoid 1189168404Spjdzfs_unregister_callbacks(zfsvfs_t *zfsvfs) 1190168404Spjd{ 1191168404Spjd objset_t *os = zfsvfs->z_os; 1192168404Spjd struct dsl_dataset *ds; 1193168404Spjd 1194168404Spjd /* 1195168404Spjd * Unregister properties. 1196168404Spjd */ 1197168404Spjd if (!dmu_objset_is_snapshot(os)) { 1198168404Spjd ds = dmu_objset_ds(os); 1199168404Spjd VERIFY(dsl_prop_unregister(ds, "atime", atime_changed_cb, 1200168404Spjd zfsvfs) == 0); 1201168404Spjd 1202168404Spjd VERIFY(dsl_prop_unregister(ds, "xattr", xattr_changed_cb, 1203168404Spjd zfsvfs) == 0); 1204168404Spjd 1205168404Spjd VERIFY(dsl_prop_unregister(ds, "recordsize", blksz_changed_cb, 1206168404Spjd zfsvfs) == 0); 1207168404Spjd 1208168404Spjd VERIFY(dsl_prop_unregister(ds, "readonly", readonly_changed_cb, 1209168404Spjd zfsvfs) == 0); 1210168404Spjd 1211168404Spjd VERIFY(dsl_prop_unregister(ds, "setuid", setuid_changed_cb, 1212168404Spjd zfsvfs) == 0); 1213168404Spjd 1214168404Spjd VERIFY(dsl_prop_unregister(ds, "exec", exec_changed_cb, 1215168404Spjd zfsvfs) == 0); 1216168404Spjd 1217168404Spjd VERIFY(dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb, 1218168404Spjd zfsvfs) == 0); 1219168404Spjd 1220224174Smm VERIFY(dsl_prop_unregister(ds, "aclmode", acl_mode_changed_cb, 1221224174Smm zfsvfs) == 0); 1222224174Smm 1223168404Spjd VERIFY(dsl_prop_unregister(ds, "aclinherit", 1224168404Spjd acl_inherit_changed_cb, zfsvfs) == 0); 1225185029Spjd 1226185029Spjd VERIFY(dsl_prop_unregister(ds, "vscan", 1227185029Spjd vscan_changed_cb, zfsvfs) == 0); 1228168404Spjd } 1229168404Spjd} 1230168404Spjd 1231219089Spjd#ifdef SECLABEL 1232219089Spjd/* 1233219089Spjd * Convert a decimal digit string to a uint64_t integer. 1234219089Spjd */ 1235219089Spjdstatic int 1236219089Spjdstr_to_uint64(char *str, uint64_t *objnum) 1237219089Spjd{ 1238219089Spjd uint64_t num = 0; 1239219089Spjd 1240219089Spjd while (*str) { 1241219089Spjd if (*str < '0' || *str > '9') 1242219089Spjd return (EINVAL); 1243219089Spjd 1244219089Spjd num = num*10 + *str++ - '0'; 1245219089Spjd } 1246219089Spjd 1247219089Spjd *objnum = num; 1248219089Spjd return (0); 1249219089Spjd} 1250219089Spjd 1251219089Spjd/* 1252219089Spjd * The boot path passed from the boot loader is in the form of 1253219089Spjd * "rootpool-name/root-filesystem-object-number'. Convert this 1254219089Spjd * string to a dataset name: "rootpool-name/root-filesystem-name". 1255219089Spjd */ 1256219089Spjdstatic int 1257219089Spjdzfs_parse_bootfs(char *bpath, char *outpath) 1258219089Spjd{ 1259219089Spjd char *slashp; 1260219089Spjd uint64_t objnum; 1261219089Spjd int error; 1262219089Spjd 1263219089Spjd if (*bpath == 0 || *bpath == '/') 1264219089Spjd return (EINVAL); 1265219089Spjd 1266219089Spjd (void) strcpy(outpath, bpath); 1267219089Spjd 1268219089Spjd slashp = strchr(bpath, '/'); 1269219089Spjd 1270219089Spjd /* if no '/', just return the pool name */ 1271219089Spjd if (slashp == NULL) { 1272219089Spjd return (0); 1273219089Spjd } 1274219089Spjd 1275219089Spjd /* if not a number, just return the root dataset name */ 1276219089Spjd if (str_to_uint64(slashp+1, &objnum)) { 1277219089Spjd return (0); 1278219089Spjd } 1279219089Spjd 1280219089Spjd *slashp = '\0'; 1281219089Spjd error = dsl_dsobj_to_dsname(bpath, objnum, outpath); 1282219089Spjd *slashp = '/'; 1283219089Spjd 1284219089Spjd return (error); 1285219089Spjd} 1286219089Spjd 1287219089Spjd/* 1288219089Spjd * zfs_check_global_label: 1289219089Spjd * Check that the hex label string is appropriate for the dataset 1290219089Spjd * being mounted into the global_zone proper. 1291219089Spjd * 1292219089Spjd * Return an error if the hex label string is not default or 1293219089Spjd * admin_low/admin_high. For admin_low labels, the corresponding 1294219089Spjd * dataset must be readonly. 1295219089Spjd */ 1296219089Spjdint 1297219089Spjdzfs_check_global_label(const char *dsname, const char *hexsl) 1298219089Spjd{ 1299219089Spjd if (strcasecmp(hexsl, ZFS_MLSLABEL_DEFAULT) == 0) 1300219089Spjd return (0); 1301219089Spjd if (strcasecmp(hexsl, ADMIN_HIGH) == 0) 1302219089Spjd return (0); 1303219089Spjd if (strcasecmp(hexsl, ADMIN_LOW) == 0) { 1304219089Spjd /* must be readonly */ 1305219089Spjd uint64_t rdonly; 1306219089Spjd 1307219089Spjd if (dsl_prop_get_integer(dsname, 1308219089Spjd zfs_prop_to_name(ZFS_PROP_READONLY), &rdonly, NULL)) 1309219089Spjd return (EACCES); 1310219089Spjd return (rdonly ? 0 : EACCES); 1311219089Spjd } 1312219089Spjd return (EACCES); 1313219089Spjd} 1314219089Spjd 1315219089Spjd/* 1316219089Spjd * zfs_mount_label_policy: 1317219089Spjd * Determine whether the mount is allowed according to MAC check. 1318219089Spjd * by comparing (where appropriate) label of the dataset against 1319219089Spjd * the label of the zone being mounted into. If the dataset has 1320219089Spjd * no label, create one. 1321219089Spjd * 1322219089Spjd * Returns: 1323219089Spjd * 0 : access allowed 1324219089Spjd * >0 : error code, such as EACCES 1325219089Spjd */ 1326219089Spjdstatic int 1327219089Spjdzfs_mount_label_policy(vfs_t *vfsp, char *osname) 1328219089Spjd{ 1329219089Spjd int error, retv; 1330219089Spjd zone_t *mntzone = NULL; 1331219089Spjd ts_label_t *mnt_tsl; 1332219089Spjd bslabel_t *mnt_sl; 1333219089Spjd bslabel_t ds_sl; 1334219089Spjd char ds_hexsl[MAXNAMELEN]; 1335219089Spjd 1336219089Spjd retv = EACCES; /* assume the worst */ 1337219089Spjd 1338219089Spjd /* 1339219089Spjd * Start by getting the dataset label if it exists. 1340219089Spjd */ 1341219089Spjd error = dsl_prop_get(osname, zfs_prop_to_name(ZFS_PROP_MLSLABEL), 1342219089Spjd 1, sizeof (ds_hexsl), &ds_hexsl, NULL); 1343219089Spjd if (error) 1344219089Spjd return (EACCES); 1345219089Spjd 1346219089Spjd /* 1347219089Spjd * If labeling is NOT enabled, then disallow the mount of datasets 1348219089Spjd * which have a non-default label already. No other label checks 1349219089Spjd * are needed. 1350219089Spjd */ 1351219089Spjd if (!is_system_labeled()) { 1352219089Spjd if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) == 0) 1353219089Spjd return (0); 1354219089Spjd return (EACCES); 1355219089Spjd } 1356219089Spjd 1357219089Spjd /* 1358219089Spjd * Get the label of the mountpoint. If mounting into the global 1359219089Spjd * zone (i.e. mountpoint is not within an active zone and the 1360219089Spjd * zoned property is off), the label must be default or 1361219089Spjd * admin_low/admin_high only; no other checks are needed. 1362219089Spjd */ 1363219089Spjd mntzone = zone_find_by_any_path(refstr_value(vfsp->vfs_mntpt), B_FALSE); 1364219089Spjd if (mntzone->zone_id == GLOBAL_ZONEID) { 1365219089Spjd uint64_t zoned; 1366219089Spjd 1367219089Spjd zone_rele(mntzone); 1368219089Spjd 1369219089Spjd if (dsl_prop_get_integer(osname, 1370219089Spjd zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL)) 1371219089Spjd return (EACCES); 1372219089Spjd if (!zoned) 1373219089Spjd return (zfs_check_global_label(osname, ds_hexsl)); 1374219089Spjd else 1375219089Spjd /* 1376219089Spjd * This is the case of a zone dataset being mounted 1377219089Spjd * initially, before the zone has been fully created; 1378219089Spjd * allow this mount into global zone. 1379219089Spjd */ 1380219089Spjd return (0); 1381219089Spjd } 1382219089Spjd 1383219089Spjd mnt_tsl = mntzone->zone_slabel; 1384219089Spjd ASSERT(mnt_tsl != NULL); 1385219089Spjd label_hold(mnt_tsl); 1386219089Spjd mnt_sl = label2bslabel(mnt_tsl); 1387219089Spjd 1388219089Spjd if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) == 0) { 1389219089Spjd /* 1390219089Spjd * The dataset doesn't have a real label, so fabricate one. 1391219089Spjd */ 1392219089Spjd char *str = NULL; 1393219089Spjd 1394219089Spjd if (l_to_str_internal(mnt_sl, &str) == 0 && 1395219089Spjd dsl_prop_set(osname, zfs_prop_to_name(ZFS_PROP_MLSLABEL), 1396219089Spjd ZPROP_SRC_LOCAL, 1, strlen(str) + 1, str) == 0) 1397219089Spjd retv = 0; 1398219089Spjd if (str != NULL) 1399219089Spjd kmem_free(str, strlen(str) + 1); 1400219089Spjd } else if (hexstr_to_label(ds_hexsl, &ds_sl) == 0) { 1401219089Spjd /* 1402219089Spjd * Now compare labels to complete the MAC check. If the 1403219089Spjd * labels are equal then allow access. If the mountpoint 1404219089Spjd * label dominates the dataset label, allow readonly access. 1405219089Spjd * Otherwise, access is denied. 1406219089Spjd */ 1407219089Spjd if (blequal(mnt_sl, &ds_sl)) 1408219089Spjd retv = 0; 1409219089Spjd else if (bldominates(mnt_sl, &ds_sl)) { 1410219089Spjd vfs_setmntopt(vfsp, MNTOPT_RO, NULL, 0); 1411219089Spjd retv = 0; 1412219089Spjd } 1413219089Spjd } 1414219089Spjd 1415219089Spjd label_rele(mnt_tsl); 1416219089Spjd zone_rele(mntzone); 1417219089Spjd return (retv); 1418219089Spjd} 1419219089Spjd#endif /* SECLABEL */ 1420219089Spjd 1421219089Spjd#ifdef OPENSOLARIS_MOUNTROOT 1422219089Spjdstatic int 1423219089Spjdzfs_mountroot(vfs_t *vfsp, enum whymountroot why) 1424219089Spjd{ 1425219089Spjd int error = 0; 1426219089Spjd static int zfsrootdone = 0; 1427219089Spjd zfsvfs_t *zfsvfs = NULL; 1428219089Spjd znode_t *zp = NULL; 1429219089Spjd vnode_t *vp = NULL; 1430219089Spjd char *zfs_bootfs; 1431219089Spjd char *zfs_devid; 1432219089Spjd 1433219089Spjd ASSERT(vfsp); 1434219089Spjd 1435219089Spjd /* 1436219089Spjd * The filesystem that we mount as root is defined in the 1437219089Spjd * boot property "zfs-bootfs" with a format of 1438219089Spjd * "poolname/root-dataset-objnum". 1439219089Spjd */ 1440219089Spjd if (why == ROOT_INIT) { 1441219089Spjd if (zfsrootdone++) 1442219089Spjd return (EBUSY); 1443219089Spjd /* 1444219089Spjd * the process of doing a spa_load will require the 1445219089Spjd * clock to be set before we could (for example) do 1446219089Spjd * something better by looking at the timestamp on 1447219089Spjd * an uberblock, so just set it to -1. 1448219089Spjd */ 1449219089Spjd clkset(-1); 1450219089Spjd 1451219089Spjd if ((zfs_bootfs = spa_get_bootprop("zfs-bootfs")) == NULL) { 1452219089Spjd cmn_err(CE_NOTE, "spa_get_bootfs: can not get " 1453219089Spjd "bootfs name"); 1454219089Spjd return (EINVAL); 1455219089Spjd } 1456219089Spjd zfs_devid = spa_get_bootprop("diskdevid"); 1457219089Spjd error = spa_import_rootpool(rootfs.bo_name, zfs_devid); 1458219089Spjd if (zfs_devid) 1459219089Spjd spa_free_bootprop(zfs_devid); 1460219089Spjd if (error) { 1461219089Spjd spa_free_bootprop(zfs_bootfs); 1462219089Spjd cmn_err(CE_NOTE, "spa_import_rootpool: error %d", 1463219089Spjd error); 1464219089Spjd return (error); 1465219089Spjd } 1466219089Spjd if (error = zfs_parse_bootfs(zfs_bootfs, rootfs.bo_name)) { 1467219089Spjd spa_free_bootprop(zfs_bootfs); 1468219089Spjd cmn_err(CE_NOTE, "zfs_parse_bootfs: error %d", 1469219089Spjd error); 1470219089Spjd return (error); 1471219089Spjd } 1472219089Spjd 1473219089Spjd spa_free_bootprop(zfs_bootfs); 1474219089Spjd 1475219089Spjd if (error = vfs_lock(vfsp)) 1476219089Spjd return (error); 1477219089Spjd 1478219089Spjd if (error = zfs_domount(vfsp, rootfs.bo_name)) { 1479219089Spjd cmn_err(CE_NOTE, "zfs_domount: error %d", error); 1480219089Spjd goto out; 1481219089Spjd } 1482219089Spjd 1483219089Spjd zfsvfs = (zfsvfs_t *)vfsp->vfs_data; 1484219089Spjd ASSERT(zfsvfs); 1485219089Spjd if (error = zfs_zget(zfsvfs, zfsvfs->z_root, &zp)) { 1486219089Spjd cmn_err(CE_NOTE, "zfs_zget: error %d", error); 1487219089Spjd goto out; 1488219089Spjd } 1489219089Spjd 1490219089Spjd vp = ZTOV(zp); 1491219089Spjd mutex_enter(&vp->v_lock); 1492219089Spjd vp->v_flag |= VROOT; 1493219089Spjd mutex_exit(&vp->v_lock); 1494219089Spjd rootvp = vp; 1495219089Spjd 1496219089Spjd /* 1497219089Spjd * Leave rootvp held. The root file system is never unmounted. 1498219089Spjd */ 1499219089Spjd 1500219089Spjd vfs_add((struct vnode *)0, vfsp, 1501219089Spjd (vfsp->vfs_flag & VFS_RDONLY) ? MS_RDONLY : 0); 1502219089Spjdout: 1503219089Spjd vfs_unlock(vfsp); 1504219089Spjd return (error); 1505219089Spjd } else if (why == ROOT_REMOUNT) { 1506219089Spjd readonly_changed_cb(vfsp->vfs_data, B_FALSE); 1507219089Spjd vfsp->vfs_flag |= VFS_REMOUNT; 1508219089Spjd 1509219089Spjd /* refresh mount options */ 1510219089Spjd zfs_unregister_callbacks(vfsp->vfs_data); 1511219089Spjd return (zfs_register_callbacks(vfsp)); 1512219089Spjd 1513219089Spjd } else if (why == ROOT_UNMOUNT) { 1514219089Spjd zfs_unregister_callbacks((zfsvfs_t *)vfsp->vfs_data); 1515219089Spjd (void) zfs_sync(vfsp, 0, 0); 1516219089Spjd return (0); 1517219089Spjd } 1518219089Spjd 1519219089Spjd /* 1520219089Spjd * if "why" is equal to anything else other than ROOT_INIT, 1521219089Spjd * ROOT_REMOUNT, or ROOT_UNMOUNT, we do not support it. 1522219089Spjd */ 1523219089Spjd return (ENOTSUP); 1524219089Spjd} 1525219089Spjd#endif /* OPENSOLARIS_MOUNTROOT */ 1526219089Spjd 1527242554Savgstatic int 1528242554Savggetpoolname(const char *osname, char *poolname) 1529242554Savg{ 1530242554Savg char *p; 1531242554Savg 1532242554Savg p = strchr(osname, '/'); 1533242554Savg if (p == NULL) { 1534242554Savg if (strlen(osname) >= MAXNAMELEN) 1535242554Savg return (ENAMETOOLONG); 1536242554Savg (void) strcpy(poolname, osname); 1537242554Savg } else { 1538242554Savg if (p - osname >= MAXNAMELEN) 1539242554Savg return (ENAMETOOLONG); 1540242554Savg (void) strncpy(poolname, osname, p - osname); 1541242554Savg poolname[p - osname] = '\0'; 1542242554Savg } 1543242554Savg return (0); 1544242554Savg} 1545242554Savg 1546168404Spjd/*ARGSUSED*/ 1547168404Spjdstatic int 1548191990Sattiliozfs_mount(vfs_t *vfsp) 1549168404Spjd{ 1550191990Sattilio kthread_t *td = curthread; 1551185029Spjd vnode_t *mvp = vfsp->mnt_vnodecovered; 1552185029Spjd cred_t *cr = td->td_ucred; 1553185029Spjd char *osname; 1554185029Spjd int error = 0; 1555185029Spjd int canwrite; 1556168404Spjd 1557232728Smm if (!prison_allow(td->td_ucred, PR_ALLOW_MOUNT_ZFS)) 1558232728Smm return (EPERM); 1559232728Smm 1560185029Spjd if (vfs_getopt(vfsp->mnt_optnew, "from", (void **)&osname, NULL)) 1561185029Spjd return (EINVAL); 1562185029Spjd 1563168404Spjd /* 1564185029Spjd * If full-owner-access is enabled and delegated administration is 1565185029Spjd * turned on, we must set nosuid. 1566185029Spjd */ 1567185029Spjd if (zfs_super_owner && 1568185029Spjd dsl_deleg_access(osname, ZFS_DELEG_PERM_MOUNT, cr) != ECANCELED) { 1569185029Spjd secpolicy_fs_mount_clearopts(cr, vfsp); 1570185029Spjd } 1571185029Spjd 1572185029Spjd /* 1573185029Spjd * Check for mount privilege? 1574185029Spjd * 1575185029Spjd * If we don't have privilege then see if 1576185029Spjd * we have local permission to allow it 1577185029Spjd */ 1578185029Spjd error = secpolicy_fs_mount(cr, mvp, vfsp); 1579185029Spjd if (error) { 1580212694Smm if (dsl_deleg_access(osname, ZFS_DELEG_PERM_MOUNT, cr) != 0) 1581196944Spjd goto out; 1582196944Spjd 1583196944Spjd if (!(vfsp->vfs_flag & MS_REMOUNT)) { 1584185029Spjd vattr_t vattr; 1585185029Spjd 1586185029Spjd /* 1587185029Spjd * Make sure user is the owner of the mount point 1588185029Spjd * or has sufficient privileges. 1589185029Spjd */ 1590185029Spjd 1591185029Spjd vattr.va_mask = AT_UID; 1592185029Spjd 1593196662Spjd vn_lock(mvp, LK_SHARED | LK_RETRY); 1594212694Smm if (VOP_GETATTR(mvp, &vattr, cr)) { 1595196662Spjd VOP_UNLOCK(mvp, 0); 1596185029Spjd goto out; 1597185029Spjd } 1598185029Spjd 1599185029Spjd if (secpolicy_vnode_owner(mvp, cr, vattr.va_uid) != 0 && 1600185029Spjd VOP_ACCESS(mvp, VWRITE, cr, td) != 0) { 1601196662Spjd VOP_UNLOCK(mvp, 0); 1602185029Spjd goto out; 1603185029Spjd } 1604196662Spjd VOP_UNLOCK(mvp, 0); 1605196944Spjd } 1606185029Spjd 1607196944Spjd secpolicy_fs_mount_clearopts(cr, vfsp); 1608185029Spjd } 1609185029Spjd 1610185029Spjd /* 1611185029Spjd * Refuse to mount a filesystem if we are in a local zone and the 1612185029Spjd * dataset is not visible. 1613185029Spjd */ 1614185029Spjd if (!INGLOBALZONE(curthread) && 1615185029Spjd (!zone_dataset_visible(osname, &canwrite) || !canwrite)) { 1616185029Spjd error = EPERM; 1617185029Spjd goto out; 1618185029Spjd } 1619185029Spjd 1620219089Spjd#ifdef SECLABEL 1621219089Spjd error = zfs_mount_label_policy(vfsp, osname); 1622219089Spjd if (error) 1623219089Spjd goto out; 1624219089Spjd#endif 1625219089Spjd 1626218386Strasz vfsp->vfs_flag |= MNT_NFS4ACLS; 1627218386Strasz 1628185029Spjd /* 1629168404Spjd * When doing a remount, we simply refresh our temporary properties 1630168404Spjd * according to those options set in the current VFS options. 1631168404Spjd */ 1632185029Spjd if (vfsp->vfs_flag & MS_REMOUNT) { 1633185029Spjd /* refresh mount options */ 1634185029Spjd zfs_unregister_callbacks(vfsp->vfs_data); 1635185029Spjd error = zfs_register_callbacks(vfsp); 1636185029Spjd goto out; 1637185029Spjd } 1638168404Spjd 1639242554Savg /* Initial root mount: try hard to import the requested root pool. */ 1640242554Savg if ((vfsp->vfs_flag & MNT_ROOTFS) != 0 && 1641242554Savg (vfsp->vfs_flag & MNT_UPDATE) == 0) { 1642242554Savg char pname[MAXNAMELEN]; 1643242554Savg spa_t *spa; 1644242554Savg int prefer_cache; 1645242554Savg 1646242554Savg error = getpoolname(osname, pname); 1647242554Savg if (error) 1648242554Savg goto out; 1649242554Savg 1650242554Savg prefer_cache = 1; 1651242554Savg TUNABLE_INT_FETCH("vfs.zfs.rootpool.prefer_cached_config", 1652242554Savg &prefer_cache); 1653242554Savg mutex_enter(&spa_namespace_lock); 1654242554Savg spa = spa_lookup(pname); 1655242554Savg mutex_exit(&spa_namespace_lock); 1656242554Savg if (!prefer_cache || spa == NULL) { 1657242554Savg error = spa_import_rootpool(pname); 1658242554Savg if (error) 1659242554Savg goto out; 1660242554Savg } 1661242554Savg } 1662168510Spjd DROP_GIANT(); 1663185029Spjd error = zfs_domount(vfsp, osname); 1664168510Spjd PICKUP_GIANT(); 1665209962Smm 1666215260Smm#ifdef sun 1667209962Smm /* 1668209962Smm * Add an extra VFS_HOLD on our parent vfs so that it can't 1669209962Smm * disappear due to a forced unmount. 1670209962Smm */ 1671209962Smm if (error == 0 && ((zfsvfs_t *)vfsp->vfs_data)->z_issnap) 1672209962Smm VFS_HOLD(mvp->v_vfsp); 1673215260Smm#endif /* sun */ 1674209962Smm 1675185029Spjdout: 1676168510Spjd return (error); 1677168404Spjd} 1678168404Spjd 1679168404Spjdstatic int 1680191990Sattiliozfs_statfs(vfs_t *vfsp, struct statfs *statp) 1681169170Spjd{ 1682168404Spjd zfsvfs_t *zfsvfs = vfsp->vfs_data; 1683168404Spjd uint64_t refdbytes, availbytes, usedobjs, availobjs; 1684168404Spjd 1685168404Spjd statp->f_version = STATFS_VERSION; 1686168404Spjd 1687168404Spjd ZFS_ENTER(zfsvfs); 1688168404Spjd 1689168404Spjd dmu_objset_space(zfsvfs->z_os, 1690168404Spjd &refdbytes, &availbytes, &usedobjs, &availobjs); 1691168404Spjd 1692168404Spjd /* 1693168404Spjd * The underlying storage pool actually uses multiple block sizes. 1694168404Spjd * We report the fragsize as the smallest block size we support, 1695168404Spjd * and we report our blocksize as the filesystem's maximum blocksize. 1696168404Spjd */ 1697204101Spjd statp->f_bsize = SPA_MINBLOCKSIZE; 1698204101Spjd statp->f_iosize = zfsvfs->z_vfs->mnt_stat.f_iosize; 1699168404Spjd 1700168404Spjd /* 1701168404Spjd * The following report "total" blocks of various kinds in the 1702168404Spjd * file system, but reported in terms of f_frsize - the 1703168404Spjd * "fragment" size. 1704168404Spjd */ 1705168404Spjd 1706204101Spjd statp->f_blocks = (refdbytes + availbytes) >> SPA_MINBLOCKSHIFT; 1707168404Spjd statp->f_bfree = availbytes / statp->f_bsize; 1708168404Spjd statp->f_bavail = statp->f_bfree; /* no root reservation */ 1709168404Spjd 1710168404Spjd /* 1711168404Spjd * statvfs() should really be called statufs(), because it assumes 1712168404Spjd * static metadata. ZFS doesn't preallocate files, so the best 1713168404Spjd * we can do is report the max that could possibly fit in f_files, 1714168404Spjd * and that minus the number actually used in f_ffree. 1715168404Spjd * For f_ffree, report the smaller of the number of object available 1716168404Spjd * and the number of blocks (each object will take at least a block). 1717168404Spjd */ 1718168404Spjd statp->f_ffree = MIN(availobjs, statp->f_bfree); 1719168404Spjd statp->f_files = statp->f_ffree + usedobjs; 1720168404Spjd 1721168404Spjd /* 1722168404Spjd * We're a zfs filesystem. 1723168404Spjd */ 1724168404Spjd (void) strlcpy(statp->f_fstypename, "zfs", sizeof(statp->f_fstypename)); 1725168404Spjd 1726168404Spjd strlcpy(statp->f_mntfromname, vfsp->mnt_stat.f_mntfromname, 1727168404Spjd sizeof(statp->f_mntfromname)); 1728168404Spjd strlcpy(statp->f_mntonname, vfsp->mnt_stat.f_mntonname, 1729168404Spjd sizeof(statp->f_mntonname)); 1730168404Spjd 1731168404Spjd statp->f_namemax = ZFS_MAXNAMELEN; 1732168404Spjd 1733168404Spjd ZFS_EXIT(zfsvfs); 1734168404Spjd return (0); 1735168404Spjd} 1736168404Spjd 1737219089Spjdint 1738219089Spjdzfs_vnode_lock(vnode_t *vp, int flags) 1739219089Spjd{ 1740219089Spjd int error; 1741219089Spjd 1742219089Spjd ASSERT(vp != NULL); 1743219089Spjd 1744219089Spjd error = vn_lock(vp, flags); 1745219089Spjd return (error); 1746219089Spjd} 1747219089Spjd 1748168404Spjdstatic int 1749191990Sattiliozfs_root(vfs_t *vfsp, int flags, vnode_t **vpp) 1750168404Spjd{ 1751168404Spjd zfsvfs_t *zfsvfs = vfsp->vfs_data; 1752168404Spjd znode_t *rootzp; 1753168404Spjd int error; 1754168404Spjd 1755197459Spjd ZFS_ENTER_NOERROR(zfsvfs); 1756168404Spjd 1757168404Spjd error = zfs_zget(zfsvfs, zfsvfs->z_root, &rootzp); 1758219089Spjd if (error == 0) 1759219089Spjd *vpp = ZTOV(rootzp); 1760206667Spjd 1761206667Spjd ZFS_EXIT(zfsvfs); 1762206667Spjd 1763168404Spjd if (error == 0) { 1764219089Spjd error = zfs_vnode_lock(*vpp, flags); 1765219089Spjd if (error == 0) 1766219089Spjd (*vpp)->v_vflag |= VV_ROOT; 1767168404Spjd } 1768219089Spjd if (error != 0) 1769219089Spjd *vpp = NULL; 1770168404Spjd 1771168404Spjd return (error); 1772168404Spjd} 1773168404Spjd 1774185029Spjd/* 1775185029Spjd * Teardown the zfsvfs::z_os. 1776185029Spjd * 1777185029Spjd * Note, if 'unmounting' if FALSE, we return with the 'z_teardown_lock' 1778185029Spjd * and 'z_teardown_inactive_lock' held. 1779185029Spjd */ 1780185029Spjdstatic int 1781185029Spjdzfsvfs_teardown(zfsvfs_t *zfsvfs, boolean_t unmounting) 1782185029Spjd{ 1783185029Spjd znode_t *zp; 1784185029Spjd 1785185029Spjd rrw_enter(&zfsvfs->z_teardown_lock, RW_WRITER, FTAG); 1786185029Spjd 1787185029Spjd if (!unmounting) { 1788185029Spjd /* 1789185029Spjd * We purge the parent filesystem's vfsp as the parent 1790185029Spjd * filesystem and all of its snapshots have their vnode's 1791185029Spjd * v_vfsp set to the parent's filesystem's vfsp. Note, 1792185029Spjd * 'z_parent' is self referential for non-snapshots. 1793185029Spjd */ 1794185029Spjd (void) dnlc_purge_vfsp(zfsvfs->z_parent->z_vfs, 0); 1795197351Spjd#ifdef FREEBSD_NAMECACHE 1796197351Spjd cache_purgevfs(zfsvfs->z_parent->z_vfs); 1797197351Spjd#endif 1798185029Spjd } 1799185029Spjd 1800185029Spjd /* 1801185029Spjd * Close the zil. NB: Can't close the zil while zfs_inactive 1802185029Spjd * threads are blocked as zil_close can call zfs_inactive. 1803185029Spjd */ 1804185029Spjd if (zfsvfs->z_log) { 1805185029Spjd zil_close(zfsvfs->z_log); 1806185029Spjd zfsvfs->z_log = NULL; 1807185029Spjd } 1808185029Spjd 1809185029Spjd rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_WRITER); 1810185029Spjd 1811185029Spjd /* 1812185029Spjd * If we are not unmounting (ie: online recv) and someone already 1813185029Spjd * unmounted this file system while we were doing the switcheroo, 1814185029Spjd * or a reopen of z_os failed then just bail out now. 1815185029Spjd */ 1816185029Spjd if (!unmounting && (zfsvfs->z_unmounted || zfsvfs->z_os == NULL)) { 1817185029Spjd rw_exit(&zfsvfs->z_teardown_inactive_lock); 1818185029Spjd rrw_exit(&zfsvfs->z_teardown_lock, FTAG); 1819185029Spjd return (EIO); 1820185029Spjd } 1821185029Spjd 1822185029Spjd /* 1823185029Spjd * At this point there are no vops active, and any new vops will 1824185029Spjd * fail with EIO since we have z_teardown_lock for writer (only 1825185029Spjd * relavent for forced unmount). 1826185029Spjd * 1827185029Spjd * Release all holds on dbufs. 1828185029Spjd */ 1829185029Spjd mutex_enter(&zfsvfs->z_znodes_lock); 1830185029Spjd for (zp = list_head(&zfsvfs->z_all_znodes); zp != NULL; 1831185029Spjd zp = list_next(&zfsvfs->z_all_znodes, zp)) 1832219089Spjd if (zp->z_sa_hdl) { 1833196297Spjd ASSERT(ZTOV(zp)->v_count >= 0); 1834185029Spjd zfs_znode_dmu_fini(zp); 1835185029Spjd } 1836185029Spjd mutex_exit(&zfsvfs->z_znodes_lock); 1837185029Spjd 1838185029Spjd /* 1839185029Spjd * If we are unmounting, set the unmounted flag and let new vops 1840185029Spjd * unblock. zfs_inactive will have the unmounted behavior, and all 1841185029Spjd * other vops will fail with EIO. 1842185029Spjd */ 1843185029Spjd if (unmounting) { 1844185029Spjd zfsvfs->z_unmounted = B_TRUE; 1845185029Spjd rrw_exit(&zfsvfs->z_teardown_lock, FTAG); 1846185029Spjd rw_exit(&zfsvfs->z_teardown_inactive_lock); 1847197133Spjd 1848197133Spjd#ifdef __FreeBSD__ 1849197133Spjd /* 1850197133Spjd * Some znodes might not be fully reclaimed, wait for them. 1851197133Spjd */ 1852197133Spjd mutex_enter(&zfsvfs->z_znodes_lock); 1853197133Spjd while (list_head(&zfsvfs->z_all_znodes) != NULL) { 1854197133Spjd msleep(zfsvfs, &zfsvfs->z_znodes_lock, 0, 1855197133Spjd "zteardown", 0); 1856197133Spjd } 1857197133Spjd mutex_exit(&zfsvfs->z_znodes_lock); 1858197133Spjd#endif 1859185029Spjd } 1860185029Spjd 1861185029Spjd /* 1862185029Spjd * z_os will be NULL if there was an error in attempting to reopen 1863185029Spjd * zfsvfs, so just return as the properties had already been 1864185029Spjd * unregistered and cached data had been evicted before. 1865185029Spjd */ 1866185029Spjd if (zfsvfs->z_os == NULL) 1867185029Spjd return (0); 1868185029Spjd 1869185029Spjd /* 1870185029Spjd * Unregister properties. 1871185029Spjd */ 1872185029Spjd zfs_unregister_callbacks(zfsvfs); 1873185029Spjd 1874185029Spjd /* 1875185029Spjd * Evict cached data 1876185029Spjd */ 1877219089Spjd if (dmu_objset_is_dirty_anywhere(zfsvfs->z_os)) 1878219089Spjd if (!(zfsvfs->z_vfs->vfs_flag & VFS_RDONLY)) 1879219089Spjd txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), 0); 1880219089Spjd (void) dmu_objset_evict_dbufs(zfsvfs->z_os); 1881185029Spjd 1882185029Spjd return (0); 1883185029Spjd} 1884185029Spjd 1885168404Spjd/*ARGSUSED*/ 1886168404Spjdstatic int 1887191990Sattiliozfs_umount(vfs_t *vfsp, int fflag) 1888168404Spjd{ 1889209962Smm kthread_t *td = curthread; 1890168404Spjd zfsvfs_t *zfsvfs = vfsp->vfs_data; 1891185029Spjd objset_t *os; 1892209962Smm cred_t *cr = td->td_ucred; 1893168404Spjd int ret; 1894168404Spjd 1895185029Spjd ret = secpolicy_fs_unmount(cr, vfsp); 1896185029Spjd if (ret) { 1897212694Smm if (dsl_deleg_access((char *)refstr_value(vfsp->vfs_resource), 1898212694Smm ZFS_DELEG_PERM_MOUNT, cr)) 1899185029Spjd return (ret); 1900185029Spjd } 1901219089Spjd 1902185029Spjd /* 1903185029Spjd * We purge the parent filesystem's vfsp as the parent filesystem 1904185029Spjd * and all of its snapshots have their vnode's v_vfsp set to the 1905185029Spjd * parent's filesystem's vfsp. Note, 'z_parent' is self 1906185029Spjd * referential for non-snapshots. 1907185029Spjd */ 1908185029Spjd (void) dnlc_purge_vfsp(zfsvfs->z_parent->z_vfs, 0); 1909168404Spjd 1910168404Spjd /* 1911168404Spjd * Unmount any snapshots mounted under .zfs before unmounting the 1912168404Spjd * dataset itself. 1913168404Spjd */ 1914169170Spjd if (zfsvfs->z_ctldir != NULL) { 1915168404Spjd if ((ret = zfsctl_umount_snapshots(vfsp, fflag, cr)) != 0) 1916168404Spjd return (ret); 1917209962Smm ret = vflush(vfsp, 0, 0, td); 1918168404Spjd ASSERT(ret == EBUSY); 1919168404Spjd if (!(fflag & MS_FORCE)) { 1920168404Spjd if (zfsvfs->z_ctldir->v_count > 1) 1921168404Spjd return (EBUSY); 1922168404Spjd ASSERT(zfsvfs->z_ctldir->v_count == 1); 1923168404Spjd } 1924168404Spjd zfsctl_destroy(zfsvfs); 1925168404Spjd ASSERT(zfsvfs->z_ctldir == NULL); 1926168404Spjd } 1927168404Spjd 1928197459Spjd if (fflag & MS_FORCE) { 1929197459Spjd /* 1930197459Spjd * Mark file system as unmounted before calling 1931197459Spjd * vflush(FORCECLOSE). This way we ensure no future vnops 1932197459Spjd * will be called and risk operating on DOOMED vnodes. 1933197459Spjd */ 1934197459Spjd rrw_enter(&zfsvfs->z_teardown_lock, RW_WRITER, FTAG); 1935197459Spjd zfsvfs->z_unmounted = B_TRUE; 1936197459Spjd rrw_exit(&zfsvfs->z_teardown_lock, FTAG); 1937197459Spjd } 1938197459Spjd 1939168404Spjd /* 1940168404Spjd * Flush all the files. 1941168404Spjd */ 1942209962Smm ret = vflush(vfsp, 1, (fflag & MS_FORCE) ? FORCECLOSE : 0, td); 1943168404Spjd if (ret != 0) { 1944168404Spjd if (!zfsvfs->z_issnap) { 1945168404Spjd zfsctl_create(zfsvfs); 1946168404Spjd ASSERT(zfsvfs->z_ctldir != NULL); 1947168404Spjd } 1948168404Spjd return (ret); 1949168404Spjd } 1950168404Spjd 1951185029Spjd if (!(fflag & MS_FORCE)) { 1952185029Spjd /* 1953185029Spjd * Check the number of active vnodes in the file system. 1954185029Spjd * Our count is maintained in the vfs structure, but the 1955185029Spjd * number is off by 1 to indicate a hold on the vfs 1956185029Spjd * structure itself. 1957185029Spjd * 1958185029Spjd * The '.zfs' directory maintains a reference of its 1959185029Spjd * own, and any active references underneath are 1960185029Spjd * reflected in the vnode count. 1961185029Spjd */ 1962185029Spjd if (zfsvfs->z_ctldir == NULL) { 1963185029Spjd if (vfsp->vfs_count > 1) 1964185029Spjd return (EBUSY); 1965185029Spjd } else { 1966185029Spjd if (vfsp->vfs_count > 2 || 1967185029Spjd zfsvfs->z_ctldir->v_count > 1) 1968185029Spjd return (EBUSY); 1969185029Spjd } 1970185029Spjd } 1971168404Spjd 1972185029Spjd VERIFY(zfsvfs_teardown(zfsvfs, B_TRUE) == 0); 1973185029Spjd os = zfsvfs->z_os; 1974185029Spjd 1975185029Spjd /* 1976185029Spjd * z_os will be NULL if there was an error in 1977185029Spjd * attempting to reopen zfsvfs. 1978185029Spjd */ 1979185029Spjd if (os != NULL) { 1980168404Spjd /* 1981185029Spjd * Unset the objset user_ptr. 1982168404Spjd */ 1983219089Spjd mutex_enter(&os->os_user_ptr_lock); 1984185029Spjd dmu_objset_set_user(os, NULL); 1985219089Spjd mutex_exit(&os->os_user_ptr_lock); 1986185029Spjd 1987185029Spjd /* 1988185029Spjd * Finally release the objset 1989185029Spjd */ 1990219089Spjd dmu_objset_disown(os, zfsvfs); 1991168404Spjd } 1992168404Spjd 1993185029Spjd /* 1994185029Spjd * We can now safely destroy the '.zfs' directory node. 1995185029Spjd */ 1996185029Spjd if (zfsvfs->z_ctldir != NULL) 1997185029Spjd zfsctl_destroy(zfsvfs); 1998185029Spjd if (zfsvfs->z_issnap) { 1999185029Spjd vnode_t *svp = vfsp->mnt_vnodecovered; 2000185029Spjd 2001197515Spjd if (svp->v_count >= 2) 2002192211Skmacy VN_RELE(svp); 2003185029Spjd } 2004168404Spjd zfs_freevfs(vfsp); 2005168404Spjd 2006168404Spjd return (0); 2007168404Spjd} 2008168404Spjd 2009168404Spjdstatic int 2010168404Spjdzfs_vget(vfs_t *vfsp, ino_t ino, int flags, vnode_t **vpp) 2011168404Spjd{ 2012168404Spjd zfsvfs_t *zfsvfs = vfsp->vfs_data; 2013168404Spjd znode_t *zp; 2014168404Spjd int err; 2015168404Spjd 2016197167Spjd /* 2017215397Savg * zfs_zget() can't operate on virtual entries like .zfs/ or 2018211855Spjd * .zfs/snapshot/ directories, that's why we return EOPNOTSUPP. 2019211855Spjd * This will make NFS to switch to LOOKUP instead of using VGET. 2020197167Spjd */ 2021197167Spjd if (ino == ZFSCTL_INO_ROOT || ino == ZFSCTL_INO_SNAPDIR) 2022197167Spjd return (EOPNOTSUPP); 2023197167Spjd 2024168404Spjd ZFS_ENTER(zfsvfs); 2025168404Spjd err = zfs_zget(zfsvfs, ino, &zp); 2026168404Spjd if (err == 0 && zp->z_unlinked) { 2027168404Spjd VN_RELE(ZTOV(zp)); 2028168404Spjd err = EINVAL; 2029168404Spjd } 2030219089Spjd if (err == 0) 2031219089Spjd *vpp = ZTOV(zp); 2032206667Spjd ZFS_EXIT(zfsvfs); 2033219089Spjd if (err == 0) 2034219089Spjd err = zfs_vnode_lock(*vpp, flags); 2035168404Spjd if (err != 0) 2036168404Spjd *vpp = NULL; 2037171063Sdfr return (err); 2038168404Spjd} 2039168404Spjd 2040168404Spjdstatic int 2041196982Spjdzfs_checkexp(vfs_t *vfsp, struct sockaddr *nam, int *extflagsp, 2042196982Spjd struct ucred **credanonp, int *numsecflavors, int **secflavors) 2043196982Spjd{ 2044196982Spjd zfsvfs_t *zfsvfs = vfsp->vfs_data; 2045196982Spjd 2046196982Spjd /* 2047196982Spjd * If this is regular file system vfsp is the same as 2048196982Spjd * zfsvfs->z_parent->z_vfs, but if it is snapshot, 2049196982Spjd * zfsvfs->z_parent->z_vfs represents parent file system 2050196982Spjd * which we have to use here, because only this file system 2051196982Spjd * has mnt_export configured. 2052196982Spjd */ 2053196982Spjd return (vfs_stdcheckexp(zfsvfs->z_parent->z_vfs, nam, extflagsp, 2054196982Spjd credanonp, numsecflavors, secflavors)); 2055196982Spjd} 2056196982Spjd 2057197151SpjdCTASSERT(SHORT_FID_LEN <= sizeof(struct fid)); 2058197151SpjdCTASSERT(LONG_FID_LEN <= sizeof(struct fid)); 2059196982Spjd 2060196982Spjdstatic int 2061222167Srmacklemzfs_fhtovp(vfs_t *vfsp, fid_t *fidp, int flags, vnode_t **vpp) 2062168404Spjd{ 2063168404Spjd zfsvfs_t *zfsvfs = vfsp->vfs_data; 2064168404Spjd znode_t *zp; 2065168404Spjd uint64_t object = 0; 2066168404Spjd uint64_t fid_gen = 0; 2067168404Spjd uint64_t gen_mask; 2068168404Spjd uint64_t zp_gen; 2069219089Spjd int i, err; 2070168404Spjd 2071168404Spjd *vpp = NULL; 2072168404Spjd 2073168404Spjd ZFS_ENTER(zfsvfs); 2074168404Spjd 2075196979Spjd /* 2076197177Spjd * On FreeBSD we can get snapshot's mount point or its parent file 2077197177Spjd * system mount point depending if snapshot is already mounted or not. 2078196979Spjd */ 2079197177Spjd if (zfsvfs->z_parent == zfsvfs && fidp->fid_len == LONG_FID_LEN) { 2080168404Spjd zfid_long_t *zlfid = (zfid_long_t *)fidp; 2081168404Spjd uint64_t objsetid = 0; 2082168404Spjd uint64_t setgen = 0; 2083168404Spjd 2084168404Spjd for (i = 0; i < sizeof (zlfid->zf_setid); i++) 2085168404Spjd objsetid |= ((uint64_t)zlfid->zf_setid[i]) << (8 * i); 2086168404Spjd 2087168404Spjd for (i = 0; i < sizeof (zlfid->zf_setgen); i++) 2088168404Spjd setgen |= ((uint64_t)zlfid->zf_setgen[i]) << (8 * i); 2089168404Spjd 2090168404Spjd ZFS_EXIT(zfsvfs); 2091168404Spjd 2092168404Spjd err = zfsctl_lookup_objset(vfsp, objsetid, &zfsvfs); 2093168404Spjd if (err) 2094168404Spjd return (EINVAL); 2095168404Spjd ZFS_ENTER(zfsvfs); 2096168404Spjd } 2097168404Spjd 2098168404Spjd if (fidp->fid_len == SHORT_FID_LEN || fidp->fid_len == LONG_FID_LEN) { 2099168404Spjd zfid_short_t *zfid = (zfid_short_t *)fidp; 2100168404Spjd 2101168404Spjd for (i = 0; i < sizeof (zfid->zf_object); i++) 2102168404Spjd object |= ((uint64_t)zfid->zf_object[i]) << (8 * i); 2103168404Spjd 2104168404Spjd for (i = 0; i < sizeof (zfid->zf_gen); i++) 2105168404Spjd fid_gen |= ((uint64_t)zfid->zf_gen[i]) << (8 * i); 2106168404Spjd } else { 2107168404Spjd ZFS_EXIT(zfsvfs); 2108168404Spjd return (EINVAL); 2109168404Spjd } 2110168404Spjd 2111168404Spjd /* A zero fid_gen means we are in the .zfs control directories */ 2112168404Spjd if (fid_gen == 0 && 2113168404Spjd (object == ZFSCTL_INO_ROOT || object == ZFSCTL_INO_SNAPDIR)) { 2114168404Spjd *vpp = zfsvfs->z_ctldir; 2115168404Spjd ASSERT(*vpp != NULL); 2116168404Spjd if (object == ZFSCTL_INO_SNAPDIR) { 2117168404Spjd VERIFY(zfsctl_root_lookup(*vpp, "snapshot", vpp, NULL, 2118185029Spjd 0, NULL, NULL, NULL, NULL, NULL) == 0); 2119168404Spjd } else { 2120168404Spjd VN_HOLD(*vpp); 2121168404Spjd } 2122206667Spjd ZFS_EXIT(zfsvfs); 2123222199Srmacklem err = zfs_vnode_lock(*vpp, flags | LK_RETRY); 2124219089Spjd if (err != 0) 2125219089Spjd *vpp = NULL; 2126219089Spjd return (err); 2127168404Spjd } 2128168404Spjd 2129168404Spjd gen_mask = -1ULL >> (64 - 8 * i); 2130168404Spjd 2131168404Spjd dprintf("getting %llu [%u mask %llx]\n", object, fid_gen, gen_mask); 2132168404Spjd if (err = zfs_zget(zfsvfs, object, &zp)) { 2133168404Spjd ZFS_EXIT(zfsvfs); 2134168404Spjd return (err); 2135168404Spjd } 2136219089Spjd (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zfsvfs), &zp_gen, 2137219089Spjd sizeof (uint64_t)); 2138219089Spjd zp_gen = zp_gen & gen_mask; 2139168404Spjd if (zp_gen == 0) 2140168404Spjd zp_gen = 1; 2141168404Spjd if (zp->z_unlinked || zp_gen != fid_gen) { 2142168404Spjd dprintf("znode gen (%u) != fid gen (%u)\n", zp_gen, fid_gen); 2143168404Spjd VN_RELE(ZTOV(zp)); 2144168404Spjd ZFS_EXIT(zfsvfs); 2145168404Spjd return (EINVAL); 2146168404Spjd } 2147168404Spjd 2148219089Spjd *vpp = ZTOV(zp); 2149206667Spjd ZFS_EXIT(zfsvfs); 2150222199Srmacklem err = zfs_vnode_lock(*vpp, flags | LK_RETRY); 2151219089Spjd if (err == 0) 2152219089Spjd vnode_create_vobject(*vpp, zp->z_size, curthread); 2153219089Spjd else 2154219089Spjd *vpp = NULL; 2155219089Spjd return (err); 2156168404Spjd} 2157168404Spjd 2158185029Spjd/* 2159185029Spjd * Block out VOPs and close zfsvfs_t::z_os 2160185029Spjd * 2161185029Spjd * Note, if successful, then we return with the 'z_teardown_lock' and 2162185029Spjd * 'z_teardown_inactive_lock' write held. 2163185029Spjd */ 2164185029Spjdint 2165219089Spjdzfs_suspend_fs(zfsvfs_t *zfsvfs) 2166168404Spjd{ 2167185029Spjd int error; 2168168404Spjd 2169185029Spjd if ((error = zfsvfs_teardown(zfsvfs, B_FALSE)) != 0) 2170185029Spjd return (error); 2171219089Spjd dmu_objset_disown(zfsvfs->z_os, zfsvfs); 2172168404Spjd 2173185029Spjd return (0); 2174185029Spjd} 2175168404Spjd 2176185029Spjd/* 2177185029Spjd * Reopen zfsvfs_t::z_os and release VOPs. 2178185029Spjd */ 2179185029Spjdint 2180219089Spjdzfs_resume_fs(zfsvfs_t *zfsvfs, const char *osname) 2181185029Spjd{ 2182185029Spjd int err; 2183168404Spjd 2184185029Spjd ASSERT(RRW_WRITE_HELD(&zfsvfs->z_teardown_lock)); 2185185029Spjd ASSERT(RW_WRITE_HELD(&zfsvfs->z_teardown_inactive_lock)); 2186185029Spjd 2187219089Spjd err = dmu_objset_own(osname, DMU_OST_ZFS, B_FALSE, zfsvfs, 2188219089Spjd &zfsvfs->z_os); 2189185029Spjd if (err) { 2190185029Spjd zfsvfs->z_os = NULL; 2191185029Spjd } else { 2192185029Spjd znode_t *zp; 2193219089Spjd uint64_t sa_obj = 0; 2194185029Spjd 2195219089Spjd /* 2196219089Spjd * Make sure version hasn't changed 2197219089Spjd */ 2198219089Spjd 2199219089Spjd err = zfs_get_zplprop(zfsvfs->z_os, ZFS_PROP_VERSION, 2200219089Spjd &zfsvfs->z_version); 2201219089Spjd 2202219089Spjd if (err) 2203219089Spjd goto bail; 2204219089Spjd 2205219089Spjd err = zap_lookup(zfsvfs->z_os, MASTER_NODE_OBJ, 2206219089Spjd ZFS_SA_ATTRS, 8, 1, &sa_obj); 2207219089Spjd 2208219089Spjd if (err && zfsvfs->z_version >= ZPL_VERSION_SA) 2209219089Spjd goto bail; 2210219089Spjd 2211219089Spjd if ((err = sa_setup(zfsvfs->z_os, sa_obj, 2212219089Spjd zfs_attr_table, ZPL_END, &zfsvfs->z_attr_table)) != 0) 2213219089Spjd goto bail; 2214219089Spjd 2215219089Spjd if (zfsvfs->z_version >= ZPL_VERSION_SA) 2216219089Spjd sa_register_update_callback(zfsvfs->z_os, 2217219089Spjd zfs_sa_upgrade); 2218219089Spjd 2219185029Spjd VERIFY(zfsvfs_setup(zfsvfs, B_FALSE) == 0); 2220185029Spjd 2221219089Spjd zfs_set_fuid_feature(zfsvfs); 2222219089Spjd 2223185029Spjd /* 2224185029Spjd * Attempt to re-establish all the active znodes with 2225185029Spjd * their dbufs. If a zfs_rezget() fails, then we'll let 2226185029Spjd * any potential callers discover that via ZFS_ENTER_VERIFY_VP 2227185029Spjd * when they try to use their znode. 2228185029Spjd */ 2229185029Spjd mutex_enter(&zfsvfs->z_znodes_lock); 2230185029Spjd for (zp = list_head(&zfsvfs->z_all_znodes); zp; 2231185029Spjd zp = list_next(&zfsvfs->z_all_znodes, zp)) { 2232185029Spjd (void) zfs_rezget(zp); 2233185029Spjd } 2234185029Spjd mutex_exit(&zfsvfs->z_znodes_lock); 2235168404Spjd } 2236168404Spjd 2237219089Spjdbail: 2238185029Spjd /* release the VOPs */ 2239185029Spjd rw_exit(&zfsvfs->z_teardown_inactive_lock); 2240185029Spjd rrw_exit(&zfsvfs->z_teardown_lock, FTAG); 2241185029Spjd 2242185029Spjd if (err) { 2243185029Spjd /* 2244219089Spjd * Since we couldn't reopen zfsvfs::z_os, or 2245219089Spjd * setup the sa framework force unmount this file system. 2246185029Spjd */ 2247185029Spjd if (vn_vfswlock(zfsvfs->z_vfs->vfs_vnodecovered) == 0) 2248185029Spjd (void) dounmount(zfsvfs->z_vfs, MS_FORCE, curthread); 2249168404Spjd } 2250185029Spjd return (err); 2251168404Spjd} 2252168404Spjd 2253168404Spjdstatic void 2254168404Spjdzfs_freevfs(vfs_t *vfsp) 2255168404Spjd{ 2256168404Spjd zfsvfs_t *zfsvfs = vfsp->vfs_data; 2257168404Spjd 2258215260Smm#ifdef sun 2259209962Smm /* 2260209962Smm * If this is a snapshot, we have an extra VFS_HOLD on our parent 2261219089Spjd * from zfs_mount(). Release it here. If we came through 2262219089Spjd * zfs_mountroot() instead, we didn't grab an extra hold, so 2263219089Spjd * skip the VFS_RELE for rootvfs. 2264209962Smm */ 2265219089Spjd if (zfsvfs->z_issnap && (vfsp != rootvfs)) 2266209962Smm VFS_RELE(zfsvfs->z_parent->z_vfs); 2267215260Smm#endif /* sun */ 2268168404Spjd 2269209962Smm zfsvfs_free(zfsvfs); 2270185029Spjd 2271168404Spjd atomic_add_32(&zfs_active_fs_count, -1); 2272168404Spjd} 2273168404Spjd 2274172135Spjd#ifdef __i386__ 2275172135Spjdstatic int desiredvnodes_backup; 2276172135Spjd#endif 2277172135Spjd 2278172135Spjdstatic void 2279172135Spjdzfs_vnodes_adjust(void) 2280172135Spjd{ 2281172135Spjd#ifdef __i386__ 2282185029Spjd int newdesiredvnodes; 2283172135Spjd 2284172135Spjd desiredvnodes_backup = desiredvnodes; 2285172135Spjd 2286172135Spjd /* 2287172135Spjd * We calculate newdesiredvnodes the same way it is done in 2288172135Spjd * vntblinit(). If it is equal to desiredvnodes, it means that 2289172135Spjd * it wasn't tuned by the administrator and we can tune it down. 2290172135Spjd */ 2291185029Spjd newdesiredvnodes = min(maxproc + cnt.v_page_count / 4, 2 * 2292185029Spjd vm_kmem_size / (5 * (sizeof(struct vm_object) + 2293185029Spjd sizeof(struct vnode)))); 2294185029Spjd if (newdesiredvnodes == desiredvnodes) 2295185029Spjd desiredvnodes = (3 * newdesiredvnodes) / 4; 2296172135Spjd#endif 2297172135Spjd} 2298172135Spjd 2299172135Spjdstatic void 2300172135Spjdzfs_vnodes_adjust_back(void) 2301172135Spjd{ 2302172135Spjd 2303172135Spjd#ifdef __i386__ 2304172135Spjd desiredvnodes = desiredvnodes_backup; 2305172135Spjd#endif 2306172135Spjd} 2307172135Spjd 2308168404Spjdvoid 2309168404Spjdzfs_init(void) 2310168404Spjd{ 2311168404Spjd 2312202129Sdelphij printf("ZFS filesystem version " ZPL_VERSION_STRING "\n"); 2313168404Spjd 2314168404Spjd /* 2315219089Spjd * Initialize .zfs directory structures 2316168404Spjd */ 2317219089Spjd zfsctl_init(); 2318168404Spjd 2319168404Spjd /* 2320219089Spjd * Initialize znode cache, vnode ops, etc... 2321168404Spjd */ 2322219089Spjd zfs_znode_init(); 2323172135Spjd 2324172135Spjd /* 2325219089Spjd * Reduce number of vnodes. Originally number of vnodes is calculated 2326172135Spjd * with UFS inode in mind. We reduce it here, because it's too big for 2327172135Spjd * ZFS/i386. 2328172135Spjd */ 2329172135Spjd zfs_vnodes_adjust(); 2330209962Smm 2331209962Smm dmu_objset_register_type(DMU_OST_ZFS, zfs_space_delta_cb); 2332168404Spjd} 2333168404Spjd 2334168404Spjdvoid 2335168404Spjdzfs_fini(void) 2336168404Spjd{ 2337168404Spjd zfsctl_fini(); 2338168404Spjd zfs_znode_fini(); 2339172135Spjd zfs_vnodes_adjust_back(); 2340168404Spjd} 2341168404Spjd 2342168404Spjdint 2343168404Spjdzfs_busy(void) 2344168404Spjd{ 2345168404Spjd return (zfs_active_fs_count != 0); 2346168404Spjd} 2347185029Spjd 2348185029Spjdint 2349209962Smmzfs_set_version(zfsvfs_t *zfsvfs, uint64_t newvers) 2350185029Spjd{ 2351185029Spjd int error; 2352209962Smm objset_t *os = zfsvfs->z_os; 2353185029Spjd dmu_tx_t *tx; 2354185029Spjd 2355185029Spjd if (newvers < ZPL_VERSION_INITIAL || newvers > ZPL_VERSION) 2356185029Spjd return (EINVAL); 2357185029Spjd 2358209962Smm if (newvers < zfsvfs->z_version) 2359209962Smm return (EINVAL); 2360185029Spjd 2361219089Spjd if (zfs_spa_version_map(newvers) > 2362219089Spjd spa_version(dmu_objset_spa(zfsvfs->z_os))) 2363219089Spjd return (ENOTSUP); 2364219089Spjd 2365185029Spjd tx = dmu_tx_create(os); 2366209962Smm dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_FALSE, ZPL_VERSION_STR); 2367219089Spjd if (newvers >= ZPL_VERSION_SA && !zfsvfs->z_use_sa) { 2368219089Spjd dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE, 2369219089Spjd ZFS_SA_ATTRS); 2370219089Spjd dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); 2371219089Spjd } 2372185029Spjd error = dmu_tx_assign(tx, TXG_WAIT); 2373185029Spjd if (error) { 2374185029Spjd dmu_tx_abort(tx); 2375209962Smm return (error); 2376185029Spjd } 2377219089Spjd 2378209962Smm error = zap_update(os, MASTER_NODE_OBJ, ZPL_VERSION_STR, 2379209962Smm 8, 1, &newvers, tx); 2380185029Spjd 2381209962Smm if (error) { 2382209962Smm dmu_tx_commit(tx); 2383209962Smm return (error); 2384209962Smm } 2385209962Smm 2386219089Spjd if (newvers >= ZPL_VERSION_SA && !zfsvfs->z_use_sa) { 2387219089Spjd uint64_t sa_obj; 2388219089Spjd 2389219089Spjd ASSERT3U(spa_version(dmu_objset_spa(zfsvfs->z_os)), >=, 2390219089Spjd SPA_VERSION_SA); 2391219089Spjd sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE, 2392219089Spjd DMU_OT_NONE, 0, tx); 2393219089Spjd 2394219089Spjd error = zap_add(os, MASTER_NODE_OBJ, 2395219089Spjd ZFS_SA_ATTRS, 8, 1, &sa_obj, tx); 2396219089Spjd ASSERT3U(error, ==, 0); 2397219089Spjd 2398219089Spjd VERIFY(0 == sa_set_sa_object(os, sa_obj)); 2399219089Spjd sa_register_update_callback(os, zfs_sa_upgrade); 2400219089Spjd } 2401219089Spjd 2402219089Spjd spa_history_log_internal(LOG_DS_UPGRADE, 2403219089Spjd dmu_objset_spa(os), tx, "oldver=%llu newver=%llu dataset = %llu", 2404209962Smm zfsvfs->z_version, newvers, dmu_objset_id(os)); 2405209962Smm 2406185029Spjd dmu_tx_commit(tx); 2407185029Spjd 2408209962Smm zfsvfs->z_version = newvers; 2409209962Smm 2410219089Spjd zfs_set_fuid_feature(zfsvfs); 2411209962Smm 2412209962Smm return (0); 2413185029Spjd} 2414219089Spjd 2415185029Spjd/* 2416185029Spjd * Read a property stored within the master node. 2417185029Spjd */ 2418185029Spjdint 2419185029Spjdzfs_get_zplprop(objset_t *os, zfs_prop_t prop, uint64_t *value) 2420185029Spjd{ 2421185029Spjd const char *pname; 2422185029Spjd int error = ENOENT; 2423185029Spjd 2424185029Spjd /* 2425185029Spjd * Look up the file system's value for the property. For the 2426185029Spjd * version property, we look up a slightly different string. 2427185029Spjd */ 2428185029Spjd if (prop == ZFS_PROP_VERSION) 2429185029Spjd pname = ZPL_VERSION_STR; 2430185029Spjd else 2431185029Spjd pname = zfs_prop_to_name(prop); 2432185029Spjd 2433185029Spjd if (os != NULL) 2434185029Spjd error = zap_lookup(os, MASTER_NODE_OBJ, pname, 8, 1, value); 2435185029Spjd 2436185029Spjd if (error == ENOENT) { 2437185029Spjd /* No value set, use the default value */ 2438185029Spjd switch (prop) { 2439185029Spjd case ZFS_PROP_VERSION: 2440185029Spjd *value = ZPL_VERSION; 2441185029Spjd break; 2442185029Spjd case ZFS_PROP_NORMALIZE: 2443185029Spjd case ZFS_PROP_UTF8ONLY: 2444185029Spjd *value = 0; 2445185029Spjd break; 2446185029Spjd case ZFS_PROP_CASE: 2447185029Spjd *value = ZFS_CASE_SENSITIVE; 2448185029Spjd break; 2449185029Spjd default: 2450185029Spjd return (error); 2451185029Spjd } 2452185029Spjd error = 0; 2453185029Spjd } 2454185029Spjd return (error); 2455185029Spjd} 2456229565Smm 2457229565Smm#ifdef _KERNEL 2458229565Smmvoid 2459229565Smmzfsvfs_update_fromname(const char *oldname, const char *newname) 2460229565Smm{ 2461229565Smm char tmpbuf[MAXPATHLEN]; 2462229565Smm struct mount *mp; 2463229565Smm char *fromname; 2464229565Smm size_t oldlen; 2465229565Smm 2466229565Smm oldlen = strlen(oldname); 2467229565Smm 2468229565Smm mtx_lock(&mountlist_mtx); 2469229565Smm TAILQ_FOREACH(mp, &mountlist, mnt_list) { 2470229565Smm fromname = mp->mnt_stat.f_mntfromname; 2471229565Smm if (strcmp(fromname, oldname) == 0) { 2472229565Smm (void)strlcpy(fromname, newname, 2473229565Smm sizeof(mp->mnt_stat.f_mntfromname)); 2474229565Smm continue; 2475229565Smm } 2476229565Smm if (strncmp(fromname, oldname, oldlen) == 0 && 2477229565Smm (fromname[oldlen] == '/' || fromname[oldlen] == '@')) { 2478229565Smm (void)snprintf(tmpbuf, sizeof(tmpbuf), "%s%s", 2479229565Smm newname, fromname + oldlen); 2480229565Smm (void)strlcpy(fromname, tmpbuf, 2481229565Smm sizeof(mp->mnt_stat.f_mntfromname)); 2482229565Smm continue; 2483229565Smm } 2484229565Smm } 2485229565Smm mtx_unlock(&mountlist_mtx); 2486229565Smm} 2487229565Smm#endif 2488