zfs_vfsops.c revision 229565
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> 63219089Spjd#include "zfs_comutil.h" 64168404Spjd 65168404Spjdstruct mtx zfs_debug_mtx; 66168404SpjdMTX_SYSINIT(zfs_debug_mtx, &zfs_debug_mtx, "zfs_debug", MTX_DEF); 67185029Spjd 68168404SpjdSYSCTL_NODE(_vfs, OID_AUTO, zfs, CTLFLAG_RW, 0, "ZFS file system"); 69185029Spjd 70219089Spjdint zfs_super_owner; 71185029SpjdSYSCTL_INT(_vfs_zfs, OID_AUTO, super_owner, CTLFLAG_RW, &zfs_super_owner, 0, 72185029Spjd "File system owner can perform privileged operation on his file systems"); 73185029Spjd 74219089Spjdint zfs_debug_level; 75168713SpjdTUNABLE_INT("vfs.zfs.debug", &zfs_debug_level); 76168404SpjdSYSCTL_INT(_vfs_zfs, OID_AUTO, debug, CTLFLAG_RW, &zfs_debug_level, 0, 77168404Spjd "Debug level"); 78168404Spjd 79185029SpjdSYSCTL_NODE(_vfs_zfs, OID_AUTO, version, CTLFLAG_RD, 0, "ZFS versions"); 80185029Spjdstatic int zfs_version_acl = ZFS_ACL_VERSION; 81185029SpjdSYSCTL_INT(_vfs_zfs_version, OID_AUTO, acl, CTLFLAG_RD, &zfs_version_acl, 0, 82185029Spjd "ZFS_ACL_VERSION"); 83185029Spjdstatic int zfs_version_spa = SPA_VERSION; 84185029SpjdSYSCTL_INT(_vfs_zfs_version, OID_AUTO, spa, CTLFLAG_RD, &zfs_version_spa, 0, 85185029Spjd "SPA_VERSION"); 86185029Spjdstatic int zfs_version_zpl = ZPL_VERSION; 87185029SpjdSYSCTL_INT(_vfs_zfs_version, OID_AUTO, zpl, CTLFLAG_RD, &zfs_version_zpl, 0, 88185029Spjd "ZPL_VERSION"); 89185029Spjd 90191990Sattiliostatic int zfs_mount(vfs_t *vfsp); 91191990Sattiliostatic int zfs_umount(vfs_t *vfsp, int fflag); 92191990Sattiliostatic int zfs_root(vfs_t *vfsp, int flags, vnode_t **vpp); 93191990Sattiliostatic int zfs_statfs(vfs_t *vfsp, struct statfs *statp); 94168404Spjdstatic int zfs_vget(vfs_t *vfsp, ino_t ino, int flags, vnode_t **vpp); 95191990Sattiliostatic int zfs_sync(vfs_t *vfsp, int waitfor); 96196982Spjdstatic int zfs_checkexp(vfs_t *vfsp, struct sockaddr *nam, int *extflagsp, 97196982Spjd struct ucred **credanonp, int *numsecflavors, int **secflavors); 98222167Srmacklemstatic int zfs_fhtovp(vfs_t *vfsp, fid_t *fidp, int flags, vnode_t **vpp); 99168404Spjdstatic void zfs_objset_close(zfsvfs_t *zfsvfs); 100168404Spjdstatic void zfs_freevfs(vfs_t *vfsp); 101168404Spjd 102168404Spjdstatic struct vfsops zfs_vfsops = { 103168404Spjd .vfs_mount = zfs_mount, 104168404Spjd .vfs_unmount = zfs_umount, 105168404Spjd .vfs_root = zfs_root, 106168404Spjd .vfs_statfs = zfs_statfs, 107168404Spjd .vfs_vget = zfs_vget, 108168404Spjd .vfs_sync = zfs_sync, 109196982Spjd .vfs_checkexp = zfs_checkexp, 110168404Spjd .vfs_fhtovp = zfs_fhtovp, 111168404Spjd}; 112168404Spjd 113185029SpjdVFS_SET(zfs_vfsops, zfs, VFCF_JAIL | VFCF_DELEGADMIN); 114168404Spjd 115168404Spjd/* 116168404Spjd * We need to keep a count of active fs's. 117168404Spjd * This is necessary to prevent our module 118168404Spjd * from being unloaded after a umount -f 119168404Spjd */ 120168404Spjdstatic uint32_t zfs_active_fs_count = 0; 121168404Spjd 122168404Spjd/*ARGSUSED*/ 123168404Spjdstatic int 124191990Sattiliozfs_sync(vfs_t *vfsp, int waitfor) 125168404Spjd{ 126168404Spjd 127168404Spjd /* 128168404Spjd * Data integrity is job one. We don't want a compromised kernel 129168404Spjd * writing to the storage pool, so we never sync during panic. 130168404Spjd */ 131168404Spjd if (panicstr) 132168404Spjd return (0); 133168404Spjd 134168404Spjd if (vfsp != NULL) { 135168404Spjd /* 136168404Spjd * Sync a specific filesystem. 137168404Spjd */ 138168404Spjd zfsvfs_t *zfsvfs = vfsp->vfs_data; 139209962Smm dsl_pool_t *dp; 140168404Spjd int error; 141168404Spjd 142191990Sattilio error = vfs_stdsync(vfsp, waitfor); 143168404Spjd if (error != 0) 144168404Spjd return (error); 145168404Spjd 146168404Spjd ZFS_ENTER(zfsvfs); 147209962Smm dp = dmu_objset_pool(zfsvfs->z_os); 148209962Smm 149209962Smm /* 150209962Smm * If the system is shutting down, then skip any 151209962Smm * filesystems which may exist on a suspended pool. 152209962Smm */ 153209962Smm if (sys_shutdown && spa_suspended(dp->dp_spa)) { 154209962Smm ZFS_EXIT(zfsvfs); 155209962Smm return (0); 156209962Smm } 157209962Smm 158168404Spjd if (zfsvfs->z_log != NULL) 159219089Spjd zil_commit(zfsvfs->z_log, 0); 160219089Spjd 161168404Spjd ZFS_EXIT(zfsvfs); 162168404Spjd } else { 163168404Spjd /* 164168404Spjd * Sync all ZFS filesystems. This is what happens when you 165168404Spjd * run sync(1M). Unlike other filesystems, ZFS honors the 166168404Spjd * request by waiting for all pools to commit all dirty data. 167168404Spjd */ 168168404Spjd spa_sync_allpools(); 169168404Spjd } 170168404Spjd 171168404Spjd return (0); 172168404Spjd} 173168404Spjd 174219089Spjd#ifndef __FreeBSD__ 175219089Spjdstatic int 176219089Spjdzfs_create_unique_device(dev_t *dev) 177219089Spjd{ 178219089Spjd major_t new_major; 179219089Spjd 180219089Spjd do { 181219089Spjd ASSERT3U(zfs_minor, <=, MAXMIN32); 182219089Spjd minor_t start = zfs_minor; 183219089Spjd do { 184219089Spjd mutex_enter(&zfs_dev_mtx); 185219089Spjd if (zfs_minor >= MAXMIN32) { 186219089Spjd /* 187219089Spjd * If we're still using the real major 188219089Spjd * keep out of /dev/zfs and /dev/zvol minor 189219089Spjd * number space. If we're using a getudev()'ed 190219089Spjd * major number, we can use all of its minors. 191219089Spjd */ 192219089Spjd if (zfs_major == ddi_name_to_major(ZFS_DRIVER)) 193219089Spjd zfs_minor = ZFS_MIN_MINOR; 194219089Spjd else 195219089Spjd zfs_minor = 0; 196219089Spjd } else { 197219089Spjd zfs_minor++; 198219089Spjd } 199219089Spjd *dev = makedevice(zfs_major, zfs_minor); 200219089Spjd mutex_exit(&zfs_dev_mtx); 201219089Spjd } while (vfs_devismounted(*dev) && zfs_minor != start); 202219089Spjd if (zfs_minor == start) { 203219089Spjd /* 204219089Spjd * We are using all ~262,000 minor numbers for the 205219089Spjd * current major number. Create a new major number. 206219089Spjd */ 207219089Spjd if ((new_major = getudev()) == (major_t)-1) { 208219089Spjd cmn_err(CE_WARN, 209219089Spjd "zfs_mount: Can't get unique major " 210219089Spjd "device number."); 211219089Spjd return (-1); 212219089Spjd } 213219089Spjd mutex_enter(&zfs_dev_mtx); 214219089Spjd zfs_major = new_major; 215219089Spjd zfs_minor = 0; 216219089Spjd 217219089Spjd mutex_exit(&zfs_dev_mtx); 218219089Spjd } else { 219219089Spjd break; 220219089Spjd } 221219089Spjd /* CONSTANTCONDITION */ 222219089Spjd } while (1); 223219089Spjd 224219089Spjd return (0); 225219089Spjd} 226219089Spjd#endif /* !__FreeBSD__ */ 227219089Spjd 228168404Spjdstatic void 229168404Spjdatime_changed_cb(void *arg, uint64_t newval) 230168404Spjd{ 231168404Spjd zfsvfs_t *zfsvfs = arg; 232168404Spjd 233168404Spjd if (newval == TRUE) { 234168404Spjd zfsvfs->z_atime = TRUE; 235168404Spjd zfsvfs->z_vfs->vfs_flag &= ~MNT_NOATIME; 236168404Spjd vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NOATIME); 237168404Spjd vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_ATIME, NULL, 0); 238168404Spjd } else { 239168404Spjd zfsvfs->z_atime = FALSE; 240168404Spjd zfsvfs->z_vfs->vfs_flag |= MNT_NOATIME; 241168404Spjd vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_ATIME); 242168404Spjd vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NOATIME, NULL, 0); 243168404Spjd } 244168404Spjd} 245168404Spjd 246168404Spjdstatic void 247168404Spjdxattr_changed_cb(void *arg, uint64_t newval) 248168404Spjd{ 249168404Spjd zfsvfs_t *zfsvfs = arg; 250168404Spjd 251168404Spjd if (newval == TRUE) { 252168404Spjd /* XXX locking on vfs_flag? */ 253168404Spjd#ifdef TODO 254168404Spjd zfsvfs->z_vfs->vfs_flag |= VFS_XATTR; 255168404Spjd#endif 256168404Spjd vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NOXATTR); 257168404Spjd vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_XATTR, NULL, 0); 258168404Spjd } else { 259168404Spjd /* XXX locking on vfs_flag? */ 260168404Spjd#ifdef TODO 261168404Spjd zfsvfs->z_vfs->vfs_flag &= ~VFS_XATTR; 262168404Spjd#endif 263168404Spjd vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_XATTR); 264168404Spjd vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NOXATTR, NULL, 0); 265168404Spjd } 266168404Spjd} 267168404Spjd 268168404Spjdstatic void 269168404Spjdblksz_changed_cb(void *arg, uint64_t newval) 270168404Spjd{ 271168404Spjd zfsvfs_t *zfsvfs = arg; 272168404Spjd 273168404Spjd if (newval < SPA_MINBLOCKSIZE || 274168404Spjd newval > SPA_MAXBLOCKSIZE || !ISP2(newval)) 275168404Spjd newval = SPA_MAXBLOCKSIZE; 276168404Spjd 277168404Spjd zfsvfs->z_max_blksz = newval; 278204101Spjd zfsvfs->z_vfs->mnt_stat.f_iosize = newval; 279168404Spjd} 280168404Spjd 281168404Spjdstatic void 282168404Spjdreadonly_changed_cb(void *arg, uint64_t newval) 283168404Spjd{ 284168404Spjd zfsvfs_t *zfsvfs = arg; 285168404Spjd 286168404Spjd if (newval) { 287168404Spjd /* XXX locking on vfs_flag? */ 288168404Spjd zfsvfs->z_vfs->vfs_flag |= VFS_RDONLY; 289168404Spjd vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_RW); 290168404Spjd vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_RO, NULL, 0); 291168404Spjd } else { 292168404Spjd /* XXX locking on vfs_flag? */ 293168404Spjd zfsvfs->z_vfs->vfs_flag &= ~VFS_RDONLY; 294168404Spjd vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_RO); 295168404Spjd vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_RW, NULL, 0); 296168404Spjd } 297168404Spjd} 298168404Spjd 299168404Spjdstatic void 300168404Spjdsetuid_changed_cb(void *arg, uint64_t newval) 301168404Spjd{ 302168404Spjd zfsvfs_t *zfsvfs = arg; 303168404Spjd 304168404Spjd if (newval == FALSE) { 305168404Spjd zfsvfs->z_vfs->vfs_flag |= VFS_NOSETUID; 306168404Spjd vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_SETUID); 307168404Spjd vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NOSETUID, NULL, 0); 308168404Spjd } else { 309168404Spjd zfsvfs->z_vfs->vfs_flag &= ~VFS_NOSETUID; 310168404Spjd vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NOSETUID); 311168404Spjd vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_SETUID, NULL, 0); 312168404Spjd } 313168404Spjd} 314168404Spjd 315168404Spjdstatic void 316168404Spjdexec_changed_cb(void *arg, uint64_t newval) 317168404Spjd{ 318168404Spjd zfsvfs_t *zfsvfs = arg; 319168404Spjd 320168404Spjd if (newval == FALSE) { 321168404Spjd zfsvfs->z_vfs->vfs_flag |= VFS_NOEXEC; 322168404Spjd vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_EXEC); 323168404Spjd vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NOEXEC, NULL, 0); 324168404Spjd } else { 325168404Spjd zfsvfs->z_vfs->vfs_flag &= ~VFS_NOEXEC; 326168404Spjd vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NOEXEC); 327168404Spjd vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_EXEC, NULL, 0); 328168404Spjd } 329168404Spjd} 330168404Spjd 331185029Spjd/* 332185029Spjd * The nbmand mount option can be changed at mount time. 333185029Spjd * We can't allow it to be toggled on live file systems or incorrect 334185029Spjd * behavior may be seen from cifs clients 335185029Spjd * 336185029Spjd * This property isn't registered via dsl_prop_register(), but this callback 337185029Spjd * will be called when a file system is first mounted 338185029Spjd */ 339168404Spjdstatic void 340185029Spjdnbmand_changed_cb(void *arg, uint64_t newval) 341185029Spjd{ 342185029Spjd zfsvfs_t *zfsvfs = arg; 343185029Spjd if (newval == FALSE) { 344185029Spjd vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NBMAND); 345185029Spjd vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NONBMAND, NULL, 0); 346185029Spjd } else { 347185029Spjd vfs_clearmntopt(zfsvfs->z_vfs, MNTOPT_NONBMAND); 348185029Spjd vfs_setmntopt(zfsvfs->z_vfs, MNTOPT_NBMAND, NULL, 0); 349185029Spjd } 350185029Spjd} 351185029Spjd 352185029Spjdstatic void 353168404Spjdsnapdir_changed_cb(void *arg, uint64_t newval) 354168404Spjd{ 355168404Spjd zfsvfs_t *zfsvfs = arg; 356168404Spjd 357168404Spjd zfsvfs->z_show_ctldir = newval; 358168404Spjd} 359168404Spjd 360168404Spjdstatic void 361185029Spjdvscan_changed_cb(void *arg, uint64_t newval) 362185029Spjd{ 363185029Spjd zfsvfs_t *zfsvfs = arg; 364185029Spjd 365185029Spjd zfsvfs->z_vscan = newval; 366185029Spjd} 367185029Spjd 368185029Spjdstatic void 369224174Smmacl_mode_changed_cb(void *arg, uint64_t newval) 370224174Smm{ 371224174Smm zfsvfs_t *zfsvfs = arg; 372224174Smm 373224174Smm zfsvfs->z_acl_mode = newval; 374224174Smm} 375224174Smm 376224174Smmstatic void 377168404Spjdacl_inherit_changed_cb(void *arg, uint64_t newval) 378168404Spjd{ 379168404Spjd zfsvfs_t *zfsvfs = arg; 380168404Spjd 381168404Spjd zfsvfs->z_acl_inherit = newval; 382168404Spjd} 383168404Spjd 384168404Spjdstatic int 385168404Spjdzfs_register_callbacks(vfs_t *vfsp) 386168404Spjd{ 387168404Spjd struct dsl_dataset *ds = NULL; 388168404Spjd objset_t *os = NULL; 389168404Spjd zfsvfs_t *zfsvfs = NULL; 390185029Spjd uint64_t nbmand; 391219089Spjd int readonly, do_readonly = B_FALSE; 392219089Spjd int setuid, do_setuid = B_FALSE; 393219089Spjd int exec, do_exec = B_FALSE; 394219089Spjd int xattr, do_xattr = B_FALSE; 395219089Spjd int atime, do_atime = B_FALSE; 396168404Spjd int error = 0; 397168404Spjd 398168404Spjd ASSERT(vfsp); 399168404Spjd zfsvfs = vfsp->vfs_data; 400168404Spjd ASSERT(zfsvfs); 401168404Spjd os = zfsvfs->z_os; 402168404Spjd 403168404Spjd /* 404196965Spjd * This function can be called for a snapshot when we update snapshot's 405196965Spjd * mount point, which isn't really supported. 406196965Spjd */ 407196965Spjd if (dmu_objset_is_snapshot(os)) 408196965Spjd return (EOPNOTSUPP); 409196965Spjd 410196965Spjd /* 411168404Spjd * The act of registering our callbacks will destroy any mount 412168404Spjd * options we may have. In order to enable temporary overrides 413168404Spjd * of mount options, we stash away the current values and 414168404Spjd * restore them after we register the callbacks. 415168404Spjd */ 416219089Spjd if (vfs_optionisset(vfsp, MNTOPT_RO, NULL) || 417219089Spjd !spa_writeable(dmu_objset_spa(os))) { 418168404Spjd readonly = B_TRUE; 419168404Spjd do_readonly = B_TRUE; 420168404Spjd } else if (vfs_optionisset(vfsp, MNTOPT_RW, NULL)) { 421168404Spjd readonly = B_FALSE; 422168404Spjd do_readonly = B_TRUE; 423168404Spjd } 424168404Spjd if (vfs_optionisset(vfsp, MNTOPT_NOSUID, NULL)) { 425168404Spjd setuid = B_FALSE; 426168404Spjd do_setuid = B_TRUE; 427168404Spjd } else { 428168404Spjd if (vfs_optionisset(vfsp, MNTOPT_NOSETUID, NULL)) { 429168404Spjd setuid = B_FALSE; 430168404Spjd do_setuid = B_TRUE; 431168404Spjd } else if (vfs_optionisset(vfsp, MNTOPT_SETUID, NULL)) { 432168404Spjd setuid = B_TRUE; 433168404Spjd do_setuid = B_TRUE; 434168404Spjd } 435168404Spjd } 436168404Spjd if (vfs_optionisset(vfsp, MNTOPT_NOEXEC, NULL)) { 437168404Spjd exec = B_FALSE; 438168404Spjd do_exec = B_TRUE; 439168404Spjd } else if (vfs_optionisset(vfsp, MNTOPT_EXEC, NULL)) { 440168404Spjd exec = B_TRUE; 441168404Spjd do_exec = B_TRUE; 442168404Spjd } 443168404Spjd if (vfs_optionisset(vfsp, MNTOPT_NOXATTR, NULL)) { 444168404Spjd xattr = B_FALSE; 445168404Spjd do_xattr = B_TRUE; 446168404Spjd } else if (vfs_optionisset(vfsp, MNTOPT_XATTR, NULL)) { 447168404Spjd xattr = B_TRUE; 448168404Spjd do_xattr = B_TRUE; 449168404Spjd } 450185029Spjd if (vfs_optionisset(vfsp, MNTOPT_NOATIME, NULL)) { 451185029Spjd atime = B_FALSE; 452185029Spjd do_atime = B_TRUE; 453185029Spjd } else if (vfs_optionisset(vfsp, MNTOPT_ATIME, NULL)) { 454185029Spjd atime = B_TRUE; 455185029Spjd do_atime = B_TRUE; 456185029Spjd } 457168404Spjd 458168404Spjd /* 459185029Spjd * nbmand is a special property. It can only be changed at 460185029Spjd * mount time. 461185029Spjd * 462185029Spjd * This is weird, but it is documented to only be changeable 463185029Spjd * at mount time. 464185029Spjd */ 465185029Spjd if (vfs_optionisset(vfsp, MNTOPT_NONBMAND, NULL)) { 466185029Spjd nbmand = B_FALSE; 467185029Spjd } else if (vfs_optionisset(vfsp, MNTOPT_NBMAND, NULL)) { 468185029Spjd nbmand = B_TRUE; 469185029Spjd } else { 470185029Spjd char osname[MAXNAMELEN]; 471185029Spjd 472185029Spjd dmu_objset_name(os, osname); 473185029Spjd if (error = dsl_prop_get_integer(osname, "nbmand", &nbmand, 474185029Spjd NULL)) { 475185029Spjd return (error); 476185029Spjd } 477185029Spjd } 478185029Spjd 479185029Spjd /* 480168404Spjd * Register property callbacks. 481168404Spjd * 482168404Spjd * It would probably be fine to just check for i/o error from 483168404Spjd * the first prop_register(), but I guess I like to go 484168404Spjd * overboard... 485168404Spjd */ 486168404Spjd ds = dmu_objset_ds(os); 487168404Spjd error = dsl_prop_register(ds, "atime", atime_changed_cb, zfsvfs); 488168404Spjd error = error ? error : dsl_prop_register(ds, 489168404Spjd "xattr", xattr_changed_cb, zfsvfs); 490168404Spjd error = error ? error : dsl_prop_register(ds, 491168404Spjd "recordsize", blksz_changed_cb, zfsvfs); 492168404Spjd error = error ? error : dsl_prop_register(ds, 493168404Spjd "readonly", readonly_changed_cb, zfsvfs); 494168404Spjd error = error ? error : dsl_prop_register(ds, 495168404Spjd "setuid", setuid_changed_cb, zfsvfs); 496168404Spjd error = error ? error : dsl_prop_register(ds, 497168404Spjd "exec", exec_changed_cb, zfsvfs); 498168404Spjd error = error ? error : dsl_prop_register(ds, 499168404Spjd "snapdir", snapdir_changed_cb, zfsvfs); 500168404Spjd error = error ? error : dsl_prop_register(ds, 501224174Smm "aclmode", acl_mode_changed_cb, zfsvfs); 502224174Smm error = error ? error : dsl_prop_register(ds, 503168404Spjd "aclinherit", acl_inherit_changed_cb, zfsvfs); 504185029Spjd error = error ? error : dsl_prop_register(ds, 505185029Spjd "vscan", vscan_changed_cb, zfsvfs); 506168404Spjd if (error) 507168404Spjd goto unregister; 508168404Spjd 509168404Spjd /* 510168404Spjd * Invoke our callbacks to restore temporary mount options. 511168404Spjd */ 512168404Spjd if (do_readonly) 513168404Spjd readonly_changed_cb(zfsvfs, readonly); 514168404Spjd if (do_setuid) 515168404Spjd setuid_changed_cb(zfsvfs, setuid); 516168404Spjd if (do_exec) 517168404Spjd exec_changed_cb(zfsvfs, exec); 518168404Spjd if (do_xattr) 519168404Spjd xattr_changed_cb(zfsvfs, xattr); 520185029Spjd if (do_atime) 521185029Spjd atime_changed_cb(zfsvfs, atime); 522168404Spjd 523185029Spjd nbmand_changed_cb(zfsvfs, nbmand); 524185029Spjd 525168404Spjd return (0); 526168404Spjd 527168404Spjdunregister: 528168404Spjd /* 529168404Spjd * We may attempt to unregister some callbacks that are not 530168404Spjd * registered, but this is OK; it will simply return ENOMSG, 531168404Spjd * which we will ignore. 532168404Spjd */ 533168404Spjd (void) dsl_prop_unregister(ds, "atime", atime_changed_cb, zfsvfs); 534168404Spjd (void) dsl_prop_unregister(ds, "xattr", xattr_changed_cb, zfsvfs); 535168404Spjd (void) dsl_prop_unregister(ds, "recordsize", blksz_changed_cb, zfsvfs); 536168404Spjd (void) dsl_prop_unregister(ds, "readonly", readonly_changed_cb, zfsvfs); 537168404Spjd (void) dsl_prop_unregister(ds, "setuid", setuid_changed_cb, zfsvfs); 538168404Spjd (void) dsl_prop_unregister(ds, "exec", exec_changed_cb, zfsvfs); 539168404Spjd (void) dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb, zfsvfs); 540224174Smm (void) dsl_prop_unregister(ds, "aclmode", acl_mode_changed_cb, zfsvfs); 541168404Spjd (void) dsl_prop_unregister(ds, "aclinherit", acl_inherit_changed_cb, 542168404Spjd zfsvfs); 543185029Spjd (void) dsl_prop_unregister(ds, "vscan", vscan_changed_cb, zfsvfs); 544168404Spjd return (error); 545168404Spjd 546168404Spjd} 547168404Spjd 548219089Spjdstatic int 549219089Spjdzfs_space_delta_cb(dmu_object_type_t bonustype, void *data, 550219089Spjd uint64_t *userp, uint64_t *groupp) 551209962Smm{ 552219089Spjd znode_phys_t *znp = data; 553219089Spjd int error = 0; 554209962Smm 555219089Spjd /* 556219089Spjd * Is it a valid type of object to track? 557219089Spjd */ 558219089Spjd if (bonustype != DMU_OT_ZNODE && bonustype != DMU_OT_SA) 559219089Spjd return (ENOENT); 560209962Smm 561219089Spjd /* 562219089Spjd * If we have a NULL data pointer 563219089Spjd * then assume the id's aren't changing and 564219089Spjd * return EEXIST to the dmu to let it know to 565219089Spjd * use the same ids 566219089Spjd */ 567219089Spjd if (data == NULL) 568219089Spjd return (EEXIST); 569209962Smm 570219089Spjd if (bonustype == DMU_OT_ZNODE) { 571219089Spjd *userp = znp->zp_uid; 572219089Spjd *groupp = znp->zp_gid; 573219089Spjd } else { 574219089Spjd int hdrsize; 575209962Smm 576219089Spjd ASSERT(bonustype == DMU_OT_SA); 577219089Spjd hdrsize = sa_hdrsize(data); 578209962Smm 579219089Spjd if (hdrsize != 0) { 580219089Spjd *userp = *((uint64_t *)((uintptr_t)data + hdrsize + 581219089Spjd SA_UID_OFFSET)); 582219089Spjd *groupp = *((uint64_t *)((uintptr_t)data + hdrsize + 583219089Spjd SA_GID_OFFSET)); 584219089Spjd } else { 585219089Spjd /* 586219089Spjd * This should only happen for newly created 587219089Spjd * files that haven't had the znode data filled 588219089Spjd * in yet. 589219089Spjd */ 590219089Spjd *userp = 0; 591219089Spjd *groupp = 0; 592219089Spjd } 593209962Smm } 594219089Spjd return (error); 595209962Smm} 596209962Smm 597209962Smmstatic void 598209962Smmfuidstr_to_sid(zfsvfs_t *zfsvfs, const char *fuidstr, 599209962Smm char *domainbuf, int buflen, uid_t *ridp) 600209962Smm{ 601209962Smm uint64_t fuid; 602209962Smm const char *domain; 603209962Smm 604209962Smm fuid = strtonum(fuidstr, NULL); 605209962Smm 606209962Smm domain = zfs_fuid_find_by_idx(zfsvfs, FUID_INDEX(fuid)); 607209962Smm if (domain) 608209962Smm (void) strlcpy(domainbuf, domain, buflen); 609209962Smm else 610209962Smm domainbuf[0] = '\0'; 611209962Smm *ridp = FUID_RID(fuid); 612209962Smm} 613209962Smm 614209962Smmstatic uint64_t 615209962Smmzfs_userquota_prop_to_obj(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type) 616209962Smm{ 617209962Smm switch (type) { 618209962Smm case ZFS_PROP_USERUSED: 619209962Smm return (DMU_USERUSED_OBJECT); 620209962Smm case ZFS_PROP_GROUPUSED: 621209962Smm return (DMU_GROUPUSED_OBJECT); 622209962Smm case ZFS_PROP_USERQUOTA: 623209962Smm return (zfsvfs->z_userquota_obj); 624209962Smm case ZFS_PROP_GROUPQUOTA: 625209962Smm return (zfsvfs->z_groupquota_obj); 626209962Smm } 627209962Smm return (0); 628209962Smm} 629209962Smm 630209962Smmint 631209962Smmzfs_userspace_many(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type, 632209962Smm uint64_t *cookiep, void *vbuf, uint64_t *bufsizep) 633209962Smm{ 634209962Smm int error; 635209962Smm zap_cursor_t zc; 636209962Smm zap_attribute_t za; 637209962Smm zfs_useracct_t *buf = vbuf; 638209962Smm uint64_t obj; 639209962Smm 640209962Smm if (!dmu_objset_userspace_present(zfsvfs->z_os)) 641209962Smm return (ENOTSUP); 642209962Smm 643209962Smm obj = zfs_userquota_prop_to_obj(zfsvfs, type); 644209962Smm if (obj == 0) { 645209962Smm *bufsizep = 0; 646209962Smm return (0); 647209962Smm } 648209962Smm 649209962Smm for (zap_cursor_init_serialized(&zc, zfsvfs->z_os, obj, *cookiep); 650209962Smm (error = zap_cursor_retrieve(&zc, &za)) == 0; 651209962Smm zap_cursor_advance(&zc)) { 652209962Smm if ((uintptr_t)buf - (uintptr_t)vbuf + sizeof (zfs_useracct_t) > 653209962Smm *bufsizep) 654209962Smm break; 655209962Smm 656209962Smm fuidstr_to_sid(zfsvfs, za.za_name, 657209962Smm buf->zu_domain, sizeof (buf->zu_domain), &buf->zu_rid); 658209962Smm 659209962Smm buf->zu_space = za.za_first_integer; 660209962Smm buf++; 661209962Smm } 662209962Smm if (error == ENOENT) 663209962Smm error = 0; 664209962Smm 665209962Smm ASSERT3U((uintptr_t)buf - (uintptr_t)vbuf, <=, *bufsizep); 666209962Smm *bufsizep = (uintptr_t)buf - (uintptr_t)vbuf; 667209962Smm *cookiep = zap_cursor_serialize(&zc); 668209962Smm zap_cursor_fini(&zc); 669209962Smm return (error); 670209962Smm} 671209962Smm 672209962Smm/* 673209962Smm * buf must be big enough (eg, 32 bytes) 674209962Smm */ 675168404Spjdstatic int 676209962Smmid_to_fuidstr(zfsvfs_t *zfsvfs, const char *domain, uid_t rid, 677209962Smm char *buf, boolean_t addok) 678209962Smm{ 679209962Smm uint64_t fuid; 680209962Smm int domainid = 0; 681209962Smm 682209962Smm if (domain && domain[0]) { 683209962Smm domainid = zfs_fuid_find_by_domain(zfsvfs, domain, NULL, addok); 684209962Smm if (domainid == -1) 685209962Smm return (ENOENT); 686209962Smm } 687209962Smm fuid = FUID_ENCODE(domainid, rid); 688209962Smm (void) sprintf(buf, "%llx", (longlong_t)fuid); 689209962Smm return (0); 690209962Smm} 691209962Smm 692209962Smmint 693209962Smmzfs_userspace_one(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type, 694209962Smm const char *domain, uint64_t rid, uint64_t *valp) 695209962Smm{ 696209962Smm char buf[32]; 697209962Smm int err; 698209962Smm uint64_t obj; 699209962Smm 700209962Smm *valp = 0; 701209962Smm 702209962Smm if (!dmu_objset_userspace_present(zfsvfs->z_os)) 703209962Smm return (ENOTSUP); 704209962Smm 705209962Smm obj = zfs_userquota_prop_to_obj(zfsvfs, type); 706209962Smm if (obj == 0) 707209962Smm return (0); 708209962Smm 709209962Smm err = id_to_fuidstr(zfsvfs, domain, rid, buf, B_FALSE); 710209962Smm if (err) 711209962Smm return (err); 712209962Smm 713209962Smm err = zap_lookup(zfsvfs->z_os, obj, buf, 8, 1, valp); 714209962Smm if (err == ENOENT) 715209962Smm err = 0; 716209962Smm return (err); 717209962Smm} 718209962Smm 719209962Smmint 720209962Smmzfs_set_userquota(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type, 721209962Smm const char *domain, uint64_t rid, uint64_t quota) 722209962Smm{ 723209962Smm char buf[32]; 724209962Smm int err; 725209962Smm dmu_tx_t *tx; 726209962Smm uint64_t *objp; 727209962Smm boolean_t fuid_dirtied; 728209962Smm 729209962Smm if (type != ZFS_PROP_USERQUOTA && type != ZFS_PROP_GROUPQUOTA) 730209962Smm return (EINVAL); 731209962Smm 732209962Smm if (zfsvfs->z_version < ZPL_VERSION_USERSPACE) 733209962Smm return (ENOTSUP); 734209962Smm 735209962Smm objp = (type == ZFS_PROP_USERQUOTA) ? &zfsvfs->z_userquota_obj : 736209962Smm &zfsvfs->z_groupquota_obj; 737209962Smm 738209962Smm err = id_to_fuidstr(zfsvfs, domain, rid, buf, B_TRUE); 739209962Smm if (err) 740209962Smm return (err); 741209962Smm fuid_dirtied = zfsvfs->z_fuid_dirty; 742209962Smm 743209962Smm tx = dmu_tx_create(zfsvfs->z_os); 744209962Smm dmu_tx_hold_zap(tx, *objp ? *objp : DMU_NEW_OBJECT, B_TRUE, NULL); 745209962Smm if (*objp == 0) { 746209962Smm dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE, 747209962Smm zfs_userquota_prop_prefixes[type]); 748209962Smm } 749209962Smm if (fuid_dirtied) 750209962Smm zfs_fuid_txhold(zfsvfs, tx); 751209962Smm err = dmu_tx_assign(tx, TXG_WAIT); 752209962Smm if (err) { 753209962Smm dmu_tx_abort(tx); 754209962Smm return (err); 755209962Smm } 756209962Smm 757209962Smm mutex_enter(&zfsvfs->z_lock); 758209962Smm if (*objp == 0) { 759209962Smm *objp = zap_create(zfsvfs->z_os, DMU_OT_USERGROUP_QUOTA, 760209962Smm DMU_OT_NONE, 0, tx); 761209962Smm VERIFY(0 == zap_add(zfsvfs->z_os, MASTER_NODE_OBJ, 762209962Smm zfs_userquota_prop_prefixes[type], 8, 1, objp, tx)); 763209962Smm } 764209962Smm mutex_exit(&zfsvfs->z_lock); 765209962Smm 766209962Smm if (quota == 0) { 767209962Smm err = zap_remove(zfsvfs->z_os, *objp, buf, tx); 768209962Smm if (err == ENOENT) 769209962Smm err = 0; 770209962Smm } else { 771209962Smm err = zap_update(zfsvfs->z_os, *objp, buf, 8, 1, "a, tx); 772209962Smm } 773209962Smm ASSERT(err == 0); 774209962Smm if (fuid_dirtied) 775209962Smm zfs_fuid_sync(zfsvfs, tx); 776209962Smm dmu_tx_commit(tx); 777209962Smm return (err); 778209962Smm} 779209962Smm 780209962Smmboolean_t 781219089Spjdzfs_fuid_overquota(zfsvfs_t *zfsvfs, boolean_t isgroup, uint64_t fuid) 782209962Smm{ 783209962Smm char buf[32]; 784209962Smm uint64_t used, quota, usedobj, quotaobj; 785209962Smm int err; 786209962Smm 787209962Smm usedobj = isgroup ? DMU_GROUPUSED_OBJECT : DMU_USERUSED_OBJECT; 788209962Smm quotaobj = isgroup ? zfsvfs->z_groupquota_obj : zfsvfs->z_userquota_obj; 789209962Smm 790209962Smm if (quotaobj == 0 || zfsvfs->z_replay) 791209962Smm return (B_FALSE); 792209962Smm 793209962Smm (void) sprintf(buf, "%llx", (longlong_t)fuid); 794209962Smm err = zap_lookup(zfsvfs->z_os, quotaobj, buf, 8, 1, "a); 795209962Smm if (err != 0) 796209962Smm return (B_FALSE); 797209962Smm 798209962Smm err = zap_lookup(zfsvfs->z_os, usedobj, buf, 8, 1, &used); 799209962Smm if (err != 0) 800209962Smm return (B_FALSE); 801209962Smm return (used >= quota); 802209962Smm} 803209962Smm 804219089Spjdboolean_t 805219089Spjdzfs_owner_overquota(zfsvfs_t *zfsvfs, znode_t *zp, boolean_t isgroup) 806219089Spjd{ 807219089Spjd uint64_t fuid; 808219089Spjd uint64_t quotaobj; 809219089Spjd 810219089Spjd quotaobj = isgroup ? zfsvfs->z_groupquota_obj : zfsvfs->z_userquota_obj; 811219089Spjd 812219089Spjd fuid = isgroup ? zp->z_gid : zp->z_uid; 813219089Spjd 814219089Spjd if (quotaobj == 0 || zfsvfs->z_replay) 815219089Spjd return (B_FALSE); 816219089Spjd 817219089Spjd return (zfs_fuid_overquota(zfsvfs, isgroup, fuid)); 818219089Spjd} 819219089Spjd 820209962Smmint 821219089Spjdzfsvfs_create(const char *osname, zfsvfs_t **zfvp) 822209962Smm{ 823209962Smm objset_t *os; 824209962Smm zfsvfs_t *zfsvfs; 825209962Smm uint64_t zval; 826209962Smm int i, error; 827219089Spjd uint64_t sa_obj; 828209962Smm 829219089Spjd zfsvfs = kmem_zalloc(sizeof (zfsvfs_t), KM_SLEEP); 830219089Spjd 831219089Spjd /* 832219089Spjd * We claim to always be readonly so we can open snapshots; 833219089Spjd * other ZPL code will prevent us from writing to snapshots. 834219089Spjd */ 835219089Spjd error = dmu_objset_own(osname, DMU_OST_ZFS, B_TRUE, zfsvfs, &os); 836219089Spjd if (error) { 837219089Spjd kmem_free(zfsvfs, sizeof (zfsvfs_t)); 838209962Smm return (error); 839209962Smm } 840209962Smm 841209962Smm /* 842209962Smm * Initialize the zfs-specific filesystem structure. 843209962Smm * Should probably make this a kmem cache, shuffle fields, 844209962Smm * and just bzero up to z_hold_mtx[]. 845209962Smm */ 846209962Smm zfsvfs->z_vfs = NULL; 847209962Smm zfsvfs->z_parent = zfsvfs; 848209962Smm zfsvfs->z_max_blksz = SPA_MAXBLOCKSIZE; 849209962Smm zfsvfs->z_show_ctldir = ZFS_SNAPDIR_VISIBLE; 850209962Smm zfsvfs->z_os = os; 851209962Smm 852209962Smm error = zfs_get_zplprop(os, ZFS_PROP_VERSION, &zfsvfs->z_version); 853209962Smm if (error) { 854209962Smm goto out; 855219089Spjd } else if (zfsvfs->z_version > 856219089Spjd zfs_zpl_version_map(spa_version(dmu_objset_spa(os)))) { 857219089Spjd (void) printf("Can't mount a version %lld file system " 858219089Spjd "on a version %lld pool\n. Pool must be upgraded to mount " 859219089Spjd "this file system.", (u_longlong_t)zfsvfs->z_version, 860219089Spjd (u_longlong_t)spa_version(dmu_objset_spa(os))); 861209962Smm error = ENOTSUP; 862209962Smm goto out; 863209962Smm } 864209962Smm if ((error = zfs_get_zplprop(os, ZFS_PROP_NORMALIZE, &zval)) != 0) 865209962Smm goto out; 866209962Smm zfsvfs->z_norm = (int)zval; 867209962Smm 868209962Smm if ((error = zfs_get_zplprop(os, ZFS_PROP_UTF8ONLY, &zval)) != 0) 869209962Smm goto out; 870209962Smm zfsvfs->z_utf8 = (zval != 0); 871209962Smm 872209962Smm if ((error = zfs_get_zplprop(os, ZFS_PROP_CASE, &zval)) != 0) 873209962Smm goto out; 874209962Smm zfsvfs->z_case = (uint_t)zval; 875209962Smm 876209962Smm /* 877209962Smm * Fold case on file systems that are always or sometimes case 878209962Smm * insensitive. 879209962Smm */ 880209962Smm if (zfsvfs->z_case == ZFS_CASE_INSENSITIVE || 881209962Smm zfsvfs->z_case == ZFS_CASE_MIXED) 882209962Smm zfsvfs->z_norm |= U8_TEXTPREP_TOUPPER; 883209962Smm 884209962Smm zfsvfs->z_use_fuids = USE_FUIDS(zfsvfs->z_version, zfsvfs->z_os); 885219089Spjd zfsvfs->z_use_sa = USE_SA(zfsvfs->z_version, zfsvfs->z_os); 886209962Smm 887219089Spjd if (zfsvfs->z_use_sa) { 888219089Spjd /* should either have both of these objects or none */ 889219089Spjd error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, 890219089Spjd &sa_obj); 891219089Spjd if (error) 892219089Spjd return (error); 893219089Spjd } else { 894219089Spjd /* 895219089Spjd * Pre SA versions file systems should never touch 896219089Spjd * either the attribute registration or layout objects. 897219089Spjd */ 898219089Spjd sa_obj = 0; 899219089Spjd } 900219089Spjd 901219089Spjd error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END, 902219089Spjd &zfsvfs->z_attr_table); 903219089Spjd if (error) 904219089Spjd goto out; 905219089Spjd 906219089Spjd if (zfsvfs->z_version >= ZPL_VERSION_SA) 907219089Spjd sa_register_update_callback(os, zfs_sa_upgrade); 908219089Spjd 909209962Smm error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1, 910209962Smm &zfsvfs->z_root); 911209962Smm if (error) 912209962Smm goto out; 913209962Smm ASSERT(zfsvfs->z_root != 0); 914209962Smm 915209962Smm error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1, 916209962Smm &zfsvfs->z_unlinkedobj); 917209962Smm if (error) 918209962Smm goto out; 919209962Smm 920209962Smm error = zap_lookup(os, MASTER_NODE_OBJ, 921209962Smm zfs_userquota_prop_prefixes[ZFS_PROP_USERQUOTA], 922209962Smm 8, 1, &zfsvfs->z_userquota_obj); 923209962Smm if (error && error != ENOENT) 924209962Smm goto out; 925209962Smm 926209962Smm error = zap_lookup(os, MASTER_NODE_OBJ, 927209962Smm zfs_userquota_prop_prefixes[ZFS_PROP_GROUPQUOTA], 928209962Smm 8, 1, &zfsvfs->z_groupquota_obj); 929209962Smm if (error && error != ENOENT) 930209962Smm goto out; 931209962Smm 932209962Smm error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_FUID_TABLES, 8, 1, 933209962Smm &zfsvfs->z_fuid_obj); 934209962Smm if (error && error != ENOENT) 935209962Smm goto out; 936209962Smm 937209962Smm error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_SHARES_DIR, 8, 1, 938209962Smm &zfsvfs->z_shares_dir); 939209962Smm if (error && error != ENOENT) 940209962Smm goto out; 941209962Smm 942209962Smm mutex_init(&zfsvfs->z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); 943209962Smm mutex_init(&zfsvfs->z_lock, NULL, MUTEX_DEFAULT, NULL); 944209962Smm list_create(&zfsvfs->z_all_znodes, sizeof (znode_t), 945209962Smm offsetof(znode_t, z_link_node)); 946209962Smm rrw_init(&zfsvfs->z_teardown_lock); 947209962Smm rw_init(&zfsvfs->z_teardown_inactive_lock, NULL, RW_DEFAULT, NULL); 948209962Smm rw_init(&zfsvfs->z_fuid_lock, NULL, RW_DEFAULT, NULL); 949209962Smm for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) 950209962Smm mutex_init(&zfsvfs->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL); 951209962Smm 952219089Spjd *zfvp = zfsvfs; 953209962Smm return (0); 954209962Smm 955209962Smmout: 956219089Spjd dmu_objset_disown(os, zfsvfs); 957219089Spjd *zfvp = NULL; 958209962Smm kmem_free(zfsvfs, sizeof (zfsvfs_t)); 959209962Smm return (error); 960209962Smm} 961209962Smm 962209962Smmstatic int 963185029Spjdzfsvfs_setup(zfsvfs_t *zfsvfs, boolean_t mounting) 964168404Spjd{ 965185029Spjd int error; 966185029Spjd 967185029Spjd error = zfs_register_callbacks(zfsvfs->z_vfs); 968185029Spjd if (error) 969185029Spjd return (error); 970185029Spjd 971185029Spjd /* 972185029Spjd * Set the objset user_ptr to track its zfsvfs. 973185029Spjd */ 974219089Spjd mutex_enter(&zfsvfs->z_os->os_user_ptr_lock); 975185029Spjd dmu_objset_set_user(zfsvfs->z_os, zfsvfs); 976219089Spjd mutex_exit(&zfsvfs->z_os->os_user_ptr_lock); 977185029Spjd 978208689Smm zfsvfs->z_log = zil_open(zfsvfs->z_os, zfs_get_data); 979208689Smm 980185029Spjd /* 981185029Spjd * If we are not mounting (ie: online recv), then we don't 982185029Spjd * have to worry about replaying the log as we blocked all 983185029Spjd * operations out since we closed the ZIL. 984185029Spjd */ 985185029Spjd if (mounting) { 986185029Spjd boolean_t readonly; 987185029Spjd 988185029Spjd /* 989185029Spjd * During replay we remove the read only flag to 990185029Spjd * allow replays to succeed. 991185029Spjd */ 992185029Spjd readonly = zfsvfs->z_vfs->vfs_flag & VFS_RDONLY; 993208689Smm if (readonly != 0) 994208689Smm zfsvfs->z_vfs->vfs_flag &= ~VFS_RDONLY; 995208689Smm else 996208689Smm zfs_unlinked_drain(zfsvfs); 997185029Spjd 998219089Spjd /* 999219089Spjd * Parse and replay the intent log. 1000219089Spjd * 1001219089Spjd * Because of ziltest, this must be done after 1002219089Spjd * zfs_unlinked_drain(). (Further note: ziltest 1003219089Spjd * doesn't use readonly mounts, where 1004219089Spjd * zfs_unlinked_drain() isn't called.) This is because 1005219089Spjd * ziltest causes spa_sync() to think it's committed, 1006219089Spjd * but actually it is not, so the intent log contains 1007219089Spjd * many txg's worth of changes. 1008219089Spjd * 1009219089Spjd * In particular, if object N is in the unlinked set in 1010219089Spjd * the last txg to actually sync, then it could be 1011219089Spjd * actually freed in a later txg and then reallocated 1012219089Spjd * in a yet later txg. This would write a "create 1013219089Spjd * object N" record to the intent log. Normally, this 1014219089Spjd * would be fine because the spa_sync() would have 1015219089Spjd * written out the fact that object N is free, before 1016219089Spjd * we could write the "create object N" intent log 1017219089Spjd * record. 1018219089Spjd * 1019219089Spjd * But when we are in ziltest mode, we advance the "open 1020219089Spjd * txg" without actually spa_sync()-ing the changes to 1021219089Spjd * disk. So we would see that object N is still 1022219089Spjd * allocated and in the unlinked set, and there is an 1023219089Spjd * intent log record saying to allocate it. 1024219089Spjd */ 1025219089Spjd if (spa_writeable(dmu_objset_spa(zfsvfs->z_os))) { 1026219089Spjd if (zil_replay_disable) { 1027219089Spjd zil_destroy(zfsvfs->z_log, B_FALSE); 1028219089Spjd } else { 1029219089Spjd zfsvfs->z_replay = B_TRUE; 1030219089Spjd zil_replay(zfsvfs->z_os, zfsvfs, 1031219089Spjd zfs_replay_vector); 1032219089Spjd zfsvfs->z_replay = B_FALSE; 1033219089Spjd } 1034208689Smm } 1035185029Spjd zfsvfs->z_vfs->vfs_flag |= readonly; /* restore readonly bit */ 1036185029Spjd } 1037185029Spjd 1038185029Spjd return (0); 1039185029Spjd} 1040185029Spjd 1041210470Smmextern krwlock_t zfsvfs_lock; /* in zfs_znode.c */ 1042210470Smm 1043209962Smmvoid 1044209962Smmzfsvfs_free(zfsvfs_t *zfsvfs) 1045185029Spjd{ 1046209962Smm int i; 1047209962Smm 1048210470Smm /* 1049210470Smm * This is a barrier to prevent the filesystem from going away in 1050210470Smm * zfs_znode_move() until we can safely ensure that the filesystem is 1051210470Smm * not unmounted. We consider the filesystem valid before the barrier 1052210470Smm * and invalid after the barrier. 1053210470Smm */ 1054210470Smm rw_enter(&zfsvfs_lock, RW_READER); 1055210470Smm rw_exit(&zfsvfs_lock); 1056210470Smm 1057209962Smm zfs_fuid_destroy(zfsvfs); 1058209962Smm 1059185029Spjd mutex_destroy(&zfsvfs->z_znodes_lock); 1060209962Smm mutex_destroy(&zfsvfs->z_lock); 1061185029Spjd list_destroy(&zfsvfs->z_all_znodes); 1062185029Spjd rrw_destroy(&zfsvfs->z_teardown_lock); 1063185029Spjd rw_destroy(&zfsvfs->z_teardown_inactive_lock); 1064185029Spjd rw_destroy(&zfsvfs->z_fuid_lock); 1065209962Smm for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) 1066209962Smm mutex_destroy(&zfsvfs->z_hold_mtx[i]); 1067185029Spjd kmem_free(zfsvfs, sizeof (zfsvfs_t)); 1068185029Spjd} 1069185029Spjd 1070209962Smmstatic void 1071209962Smmzfs_set_fuid_feature(zfsvfs_t *zfsvfs) 1072209962Smm{ 1073209962Smm zfsvfs->z_use_fuids = USE_FUIDS(zfsvfs->z_version, zfsvfs->z_os); 1074219089Spjd if (zfsvfs->z_vfs) { 1075219089Spjd if (zfsvfs->z_use_fuids) { 1076219089Spjd vfs_set_feature(zfsvfs->z_vfs, VFSFT_XVATTR); 1077219089Spjd vfs_set_feature(zfsvfs->z_vfs, VFSFT_SYSATTR_VIEWS); 1078219089Spjd vfs_set_feature(zfsvfs->z_vfs, VFSFT_ACEMASKONACCESS); 1079219089Spjd vfs_set_feature(zfsvfs->z_vfs, VFSFT_ACLONCREATE); 1080219089Spjd vfs_set_feature(zfsvfs->z_vfs, VFSFT_ACCESS_FILTER); 1081219089Spjd vfs_set_feature(zfsvfs->z_vfs, VFSFT_REPARSE); 1082219089Spjd } else { 1083219089Spjd vfs_clear_feature(zfsvfs->z_vfs, VFSFT_XVATTR); 1084219089Spjd vfs_clear_feature(zfsvfs->z_vfs, VFSFT_SYSATTR_VIEWS); 1085219089Spjd vfs_clear_feature(zfsvfs->z_vfs, VFSFT_ACEMASKONACCESS); 1086219089Spjd vfs_clear_feature(zfsvfs->z_vfs, VFSFT_ACLONCREATE); 1087219089Spjd vfs_clear_feature(zfsvfs->z_vfs, VFSFT_ACCESS_FILTER); 1088219089Spjd vfs_clear_feature(zfsvfs->z_vfs, VFSFT_REPARSE); 1089219089Spjd } 1090209962Smm } 1091219089Spjd zfsvfs->z_use_sa = USE_SA(zfsvfs->z_version, zfsvfs->z_os); 1092209962Smm} 1093209962Smm 1094185029Spjdstatic int 1095185029Spjdzfs_domount(vfs_t *vfsp, char *osname) 1096185029Spjd{ 1097209962Smm uint64_t recordsize, fsid_guid; 1098168404Spjd int error = 0; 1099168404Spjd zfsvfs_t *zfsvfs; 1100209962Smm vnode_t *vp; 1101168404Spjd 1102168404Spjd ASSERT(vfsp); 1103168404Spjd ASSERT(osname); 1104168404Spjd 1105219089Spjd error = zfsvfs_create(osname, &zfsvfs); 1106209962Smm if (error) 1107209962Smm return (error); 1108168404Spjd zfsvfs->z_vfs = vfsp; 1109168404Spjd 1110168404Spjd if (error = dsl_prop_get_integer(osname, "recordsize", &recordsize, 1111168404Spjd NULL)) 1112168404Spjd goto out; 1113204101Spjd zfsvfs->z_vfs->vfs_bsize = SPA_MINBLOCKSIZE; 1114204101Spjd zfsvfs->z_vfs->mnt_stat.f_iosize = recordsize; 1115168404Spjd 1116168404Spjd vfsp->vfs_data = zfsvfs; 1117218386Strasz vfsp->mnt_flag |= MNT_LOCAL; 1118168404Spjd vfsp->mnt_kern_flag |= MNTK_MPSAFE; 1119168404Spjd vfsp->mnt_kern_flag |= MNTK_LOOKUP_SHARED; 1120193440Sps vfsp->mnt_kern_flag |= MNTK_SHARED_WRITES; 1121168404Spjd 1122209962Smm /* 1123209962Smm * The fsid is 64 bits, composed of an 8-bit fs type, which 1124209962Smm * separates our fsid from any other filesystem types, and a 1125209962Smm * 56-bit objset unique ID. The objset unique ID is unique to 1126209962Smm * all objsets open on this system, provided by unique_create(). 1127209962Smm * The 8-bit fs type must be put in the low bits of fsid[1] 1128209962Smm * because that's where other Solaris filesystems put it. 1129209962Smm */ 1130209962Smm fsid_guid = dmu_objset_fsid_guid(zfsvfs->z_os); 1131209962Smm ASSERT((fsid_guid & ~((1ULL<<56)-1)) == 0); 1132209962Smm vfsp->vfs_fsid.val[0] = fsid_guid; 1133209962Smm vfsp->vfs_fsid.val[1] = ((fsid_guid>>32) << 8) | 1134209962Smm vfsp->mnt_vfc->vfc_typenum & 0xFF; 1135168404Spjd 1136185029Spjd /* 1137185029Spjd * Set features for file system. 1138185029Spjd */ 1139209962Smm zfs_set_fuid_feature(zfsvfs); 1140185029Spjd if (zfsvfs->z_case == ZFS_CASE_INSENSITIVE) { 1141185029Spjd vfs_set_feature(vfsp, VFSFT_DIRENTFLAGS); 1142185029Spjd vfs_set_feature(vfsp, VFSFT_CASEINSENSITIVE); 1143185029Spjd vfs_set_feature(vfsp, VFSFT_NOCASESENSITIVE); 1144185029Spjd } else if (zfsvfs->z_case == ZFS_CASE_MIXED) { 1145185029Spjd vfs_set_feature(vfsp, VFSFT_DIRENTFLAGS); 1146185029Spjd vfs_set_feature(vfsp, VFSFT_CASEINSENSITIVE); 1147185029Spjd } 1148219089Spjd vfs_set_feature(vfsp, VFSFT_ZEROCOPY_SUPPORTED); 1149185029Spjd 1150168404Spjd if (dmu_objset_is_snapshot(zfsvfs->z_os)) { 1151185029Spjd uint64_t pval; 1152168404Spjd 1153168404Spjd atime_changed_cb(zfsvfs, B_FALSE); 1154168404Spjd readonly_changed_cb(zfsvfs, B_TRUE); 1155185029Spjd if (error = dsl_prop_get_integer(osname, "xattr", &pval, NULL)) 1156168404Spjd goto out; 1157185029Spjd xattr_changed_cb(zfsvfs, pval); 1158168404Spjd zfsvfs->z_issnap = B_TRUE; 1159219089Spjd zfsvfs->z_os->os_sync = ZFS_SYNC_DISABLED; 1160209962Smm 1161219089Spjd mutex_enter(&zfsvfs->z_os->os_user_ptr_lock); 1162209962Smm dmu_objset_set_user(zfsvfs->z_os, zfsvfs); 1163219089Spjd mutex_exit(&zfsvfs->z_os->os_user_ptr_lock); 1164168404Spjd } else { 1165185029Spjd error = zfsvfs_setup(zfsvfs, B_TRUE); 1166168404Spjd } 1167168404Spjd 1168168404Spjd vfs_mountedfrom(vfsp, osname); 1169209962Smm /* Grab extra reference. */ 1170209962Smm VERIFY(VFS_ROOT(vfsp, LK_EXCLUSIVE, &vp) == 0); 1171209962Smm VOP_UNLOCK(vp, 0); 1172168404Spjd 1173168404Spjd if (!zfsvfs->z_issnap) 1174168404Spjd zfsctl_create(zfsvfs); 1175168404Spjdout: 1176168404Spjd if (error) { 1177219089Spjd dmu_objset_disown(zfsvfs->z_os, zfsvfs); 1178209962Smm zfsvfs_free(zfsvfs); 1179168404Spjd } else { 1180168404Spjd atomic_add_32(&zfs_active_fs_count, 1); 1181168404Spjd } 1182168404Spjd 1183168404Spjd return (error); 1184168404Spjd} 1185168404Spjd 1186168404Spjdvoid 1187168404Spjdzfs_unregister_callbacks(zfsvfs_t *zfsvfs) 1188168404Spjd{ 1189168404Spjd objset_t *os = zfsvfs->z_os; 1190168404Spjd struct dsl_dataset *ds; 1191168404Spjd 1192168404Spjd /* 1193168404Spjd * Unregister properties. 1194168404Spjd */ 1195168404Spjd if (!dmu_objset_is_snapshot(os)) { 1196168404Spjd ds = dmu_objset_ds(os); 1197168404Spjd VERIFY(dsl_prop_unregister(ds, "atime", atime_changed_cb, 1198168404Spjd zfsvfs) == 0); 1199168404Spjd 1200168404Spjd VERIFY(dsl_prop_unregister(ds, "xattr", xattr_changed_cb, 1201168404Spjd zfsvfs) == 0); 1202168404Spjd 1203168404Spjd VERIFY(dsl_prop_unregister(ds, "recordsize", blksz_changed_cb, 1204168404Spjd zfsvfs) == 0); 1205168404Spjd 1206168404Spjd VERIFY(dsl_prop_unregister(ds, "readonly", readonly_changed_cb, 1207168404Spjd zfsvfs) == 0); 1208168404Spjd 1209168404Spjd VERIFY(dsl_prop_unregister(ds, "setuid", setuid_changed_cb, 1210168404Spjd zfsvfs) == 0); 1211168404Spjd 1212168404Spjd VERIFY(dsl_prop_unregister(ds, "exec", exec_changed_cb, 1213168404Spjd zfsvfs) == 0); 1214168404Spjd 1215168404Spjd VERIFY(dsl_prop_unregister(ds, "snapdir", snapdir_changed_cb, 1216168404Spjd zfsvfs) == 0); 1217168404Spjd 1218224174Smm VERIFY(dsl_prop_unregister(ds, "aclmode", acl_mode_changed_cb, 1219224174Smm zfsvfs) == 0); 1220224174Smm 1221168404Spjd VERIFY(dsl_prop_unregister(ds, "aclinherit", 1222168404Spjd acl_inherit_changed_cb, zfsvfs) == 0); 1223185029Spjd 1224185029Spjd VERIFY(dsl_prop_unregister(ds, "vscan", 1225185029Spjd vscan_changed_cb, zfsvfs) == 0); 1226168404Spjd } 1227168404Spjd} 1228168404Spjd 1229219089Spjd#ifdef SECLABEL 1230219089Spjd/* 1231219089Spjd * Convert a decimal digit string to a uint64_t integer. 1232219089Spjd */ 1233219089Spjdstatic int 1234219089Spjdstr_to_uint64(char *str, uint64_t *objnum) 1235219089Spjd{ 1236219089Spjd uint64_t num = 0; 1237219089Spjd 1238219089Spjd while (*str) { 1239219089Spjd if (*str < '0' || *str > '9') 1240219089Spjd return (EINVAL); 1241219089Spjd 1242219089Spjd num = num*10 + *str++ - '0'; 1243219089Spjd } 1244219089Spjd 1245219089Spjd *objnum = num; 1246219089Spjd return (0); 1247219089Spjd} 1248219089Spjd 1249219089Spjd/* 1250219089Spjd * The boot path passed from the boot loader is in the form of 1251219089Spjd * "rootpool-name/root-filesystem-object-number'. Convert this 1252219089Spjd * string to a dataset name: "rootpool-name/root-filesystem-name". 1253219089Spjd */ 1254219089Spjdstatic int 1255219089Spjdzfs_parse_bootfs(char *bpath, char *outpath) 1256219089Spjd{ 1257219089Spjd char *slashp; 1258219089Spjd uint64_t objnum; 1259219089Spjd int error; 1260219089Spjd 1261219089Spjd if (*bpath == 0 || *bpath == '/') 1262219089Spjd return (EINVAL); 1263219089Spjd 1264219089Spjd (void) strcpy(outpath, bpath); 1265219089Spjd 1266219089Spjd slashp = strchr(bpath, '/'); 1267219089Spjd 1268219089Spjd /* if no '/', just return the pool name */ 1269219089Spjd if (slashp == NULL) { 1270219089Spjd return (0); 1271219089Spjd } 1272219089Spjd 1273219089Spjd /* if not a number, just return the root dataset name */ 1274219089Spjd if (str_to_uint64(slashp+1, &objnum)) { 1275219089Spjd return (0); 1276219089Spjd } 1277219089Spjd 1278219089Spjd *slashp = '\0'; 1279219089Spjd error = dsl_dsobj_to_dsname(bpath, objnum, outpath); 1280219089Spjd *slashp = '/'; 1281219089Spjd 1282219089Spjd return (error); 1283219089Spjd} 1284219089Spjd 1285219089Spjd/* 1286219089Spjd * zfs_check_global_label: 1287219089Spjd * Check that the hex label string is appropriate for the dataset 1288219089Spjd * being mounted into the global_zone proper. 1289219089Spjd * 1290219089Spjd * Return an error if the hex label string is not default or 1291219089Spjd * admin_low/admin_high. For admin_low labels, the corresponding 1292219089Spjd * dataset must be readonly. 1293219089Spjd */ 1294219089Spjdint 1295219089Spjdzfs_check_global_label(const char *dsname, const char *hexsl) 1296219089Spjd{ 1297219089Spjd if (strcasecmp(hexsl, ZFS_MLSLABEL_DEFAULT) == 0) 1298219089Spjd return (0); 1299219089Spjd if (strcasecmp(hexsl, ADMIN_HIGH) == 0) 1300219089Spjd return (0); 1301219089Spjd if (strcasecmp(hexsl, ADMIN_LOW) == 0) { 1302219089Spjd /* must be readonly */ 1303219089Spjd uint64_t rdonly; 1304219089Spjd 1305219089Spjd if (dsl_prop_get_integer(dsname, 1306219089Spjd zfs_prop_to_name(ZFS_PROP_READONLY), &rdonly, NULL)) 1307219089Spjd return (EACCES); 1308219089Spjd return (rdonly ? 0 : EACCES); 1309219089Spjd } 1310219089Spjd return (EACCES); 1311219089Spjd} 1312219089Spjd 1313219089Spjd/* 1314219089Spjd * zfs_mount_label_policy: 1315219089Spjd * Determine whether the mount is allowed according to MAC check. 1316219089Spjd * by comparing (where appropriate) label of the dataset against 1317219089Spjd * the label of the zone being mounted into. If the dataset has 1318219089Spjd * no label, create one. 1319219089Spjd * 1320219089Spjd * Returns: 1321219089Spjd * 0 : access allowed 1322219089Spjd * >0 : error code, such as EACCES 1323219089Spjd */ 1324219089Spjdstatic int 1325219089Spjdzfs_mount_label_policy(vfs_t *vfsp, char *osname) 1326219089Spjd{ 1327219089Spjd int error, retv; 1328219089Spjd zone_t *mntzone = NULL; 1329219089Spjd ts_label_t *mnt_tsl; 1330219089Spjd bslabel_t *mnt_sl; 1331219089Spjd bslabel_t ds_sl; 1332219089Spjd char ds_hexsl[MAXNAMELEN]; 1333219089Spjd 1334219089Spjd retv = EACCES; /* assume the worst */ 1335219089Spjd 1336219089Spjd /* 1337219089Spjd * Start by getting the dataset label if it exists. 1338219089Spjd */ 1339219089Spjd error = dsl_prop_get(osname, zfs_prop_to_name(ZFS_PROP_MLSLABEL), 1340219089Spjd 1, sizeof (ds_hexsl), &ds_hexsl, NULL); 1341219089Spjd if (error) 1342219089Spjd return (EACCES); 1343219089Spjd 1344219089Spjd /* 1345219089Spjd * If labeling is NOT enabled, then disallow the mount of datasets 1346219089Spjd * which have a non-default label already. No other label checks 1347219089Spjd * are needed. 1348219089Spjd */ 1349219089Spjd if (!is_system_labeled()) { 1350219089Spjd if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) == 0) 1351219089Spjd return (0); 1352219089Spjd return (EACCES); 1353219089Spjd } 1354219089Spjd 1355219089Spjd /* 1356219089Spjd * Get the label of the mountpoint. If mounting into the global 1357219089Spjd * zone (i.e. mountpoint is not within an active zone and the 1358219089Spjd * zoned property is off), the label must be default or 1359219089Spjd * admin_low/admin_high only; no other checks are needed. 1360219089Spjd */ 1361219089Spjd mntzone = zone_find_by_any_path(refstr_value(vfsp->vfs_mntpt), B_FALSE); 1362219089Spjd if (mntzone->zone_id == GLOBAL_ZONEID) { 1363219089Spjd uint64_t zoned; 1364219089Spjd 1365219089Spjd zone_rele(mntzone); 1366219089Spjd 1367219089Spjd if (dsl_prop_get_integer(osname, 1368219089Spjd zfs_prop_to_name(ZFS_PROP_ZONED), &zoned, NULL)) 1369219089Spjd return (EACCES); 1370219089Spjd if (!zoned) 1371219089Spjd return (zfs_check_global_label(osname, ds_hexsl)); 1372219089Spjd else 1373219089Spjd /* 1374219089Spjd * This is the case of a zone dataset being mounted 1375219089Spjd * initially, before the zone has been fully created; 1376219089Spjd * allow this mount into global zone. 1377219089Spjd */ 1378219089Spjd return (0); 1379219089Spjd } 1380219089Spjd 1381219089Spjd mnt_tsl = mntzone->zone_slabel; 1382219089Spjd ASSERT(mnt_tsl != NULL); 1383219089Spjd label_hold(mnt_tsl); 1384219089Spjd mnt_sl = label2bslabel(mnt_tsl); 1385219089Spjd 1386219089Spjd if (strcasecmp(ds_hexsl, ZFS_MLSLABEL_DEFAULT) == 0) { 1387219089Spjd /* 1388219089Spjd * The dataset doesn't have a real label, so fabricate one. 1389219089Spjd */ 1390219089Spjd char *str = NULL; 1391219089Spjd 1392219089Spjd if (l_to_str_internal(mnt_sl, &str) == 0 && 1393219089Spjd dsl_prop_set(osname, zfs_prop_to_name(ZFS_PROP_MLSLABEL), 1394219089Spjd ZPROP_SRC_LOCAL, 1, strlen(str) + 1, str) == 0) 1395219089Spjd retv = 0; 1396219089Spjd if (str != NULL) 1397219089Spjd kmem_free(str, strlen(str) + 1); 1398219089Spjd } else if (hexstr_to_label(ds_hexsl, &ds_sl) == 0) { 1399219089Spjd /* 1400219089Spjd * Now compare labels to complete the MAC check. If the 1401219089Spjd * labels are equal then allow access. If the mountpoint 1402219089Spjd * label dominates the dataset label, allow readonly access. 1403219089Spjd * Otherwise, access is denied. 1404219089Spjd */ 1405219089Spjd if (blequal(mnt_sl, &ds_sl)) 1406219089Spjd retv = 0; 1407219089Spjd else if (bldominates(mnt_sl, &ds_sl)) { 1408219089Spjd vfs_setmntopt(vfsp, MNTOPT_RO, NULL, 0); 1409219089Spjd retv = 0; 1410219089Spjd } 1411219089Spjd } 1412219089Spjd 1413219089Spjd label_rele(mnt_tsl); 1414219089Spjd zone_rele(mntzone); 1415219089Spjd return (retv); 1416219089Spjd} 1417219089Spjd#endif /* SECLABEL */ 1418219089Spjd 1419219089Spjd#ifdef OPENSOLARIS_MOUNTROOT 1420219089Spjdstatic int 1421219089Spjdzfs_mountroot(vfs_t *vfsp, enum whymountroot why) 1422219089Spjd{ 1423219089Spjd int error = 0; 1424219089Spjd static int zfsrootdone = 0; 1425219089Spjd zfsvfs_t *zfsvfs = NULL; 1426219089Spjd znode_t *zp = NULL; 1427219089Spjd vnode_t *vp = NULL; 1428219089Spjd char *zfs_bootfs; 1429219089Spjd char *zfs_devid; 1430219089Spjd 1431219089Spjd ASSERT(vfsp); 1432219089Spjd 1433219089Spjd /* 1434219089Spjd * The filesystem that we mount as root is defined in the 1435219089Spjd * boot property "zfs-bootfs" with a format of 1436219089Spjd * "poolname/root-dataset-objnum". 1437219089Spjd */ 1438219089Spjd if (why == ROOT_INIT) { 1439219089Spjd if (zfsrootdone++) 1440219089Spjd return (EBUSY); 1441219089Spjd /* 1442219089Spjd * the process of doing a spa_load will require the 1443219089Spjd * clock to be set before we could (for example) do 1444219089Spjd * something better by looking at the timestamp on 1445219089Spjd * an uberblock, so just set it to -1. 1446219089Spjd */ 1447219089Spjd clkset(-1); 1448219089Spjd 1449219089Spjd if ((zfs_bootfs = spa_get_bootprop("zfs-bootfs")) == NULL) { 1450219089Spjd cmn_err(CE_NOTE, "spa_get_bootfs: can not get " 1451219089Spjd "bootfs name"); 1452219089Spjd return (EINVAL); 1453219089Spjd } 1454219089Spjd zfs_devid = spa_get_bootprop("diskdevid"); 1455219089Spjd error = spa_import_rootpool(rootfs.bo_name, zfs_devid); 1456219089Spjd if (zfs_devid) 1457219089Spjd spa_free_bootprop(zfs_devid); 1458219089Spjd if (error) { 1459219089Spjd spa_free_bootprop(zfs_bootfs); 1460219089Spjd cmn_err(CE_NOTE, "spa_import_rootpool: error %d", 1461219089Spjd error); 1462219089Spjd return (error); 1463219089Spjd } 1464219089Spjd if (error = zfs_parse_bootfs(zfs_bootfs, rootfs.bo_name)) { 1465219089Spjd spa_free_bootprop(zfs_bootfs); 1466219089Spjd cmn_err(CE_NOTE, "zfs_parse_bootfs: error %d", 1467219089Spjd error); 1468219089Spjd return (error); 1469219089Spjd } 1470219089Spjd 1471219089Spjd spa_free_bootprop(zfs_bootfs); 1472219089Spjd 1473219089Spjd if (error = vfs_lock(vfsp)) 1474219089Spjd return (error); 1475219089Spjd 1476219089Spjd if (error = zfs_domount(vfsp, rootfs.bo_name)) { 1477219089Spjd cmn_err(CE_NOTE, "zfs_domount: error %d", error); 1478219089Spjd goto out; 1479219089Spjd } 1480219089Spjd 1481219089Spjd zfsvfs = (zfsvfs_t *)vfsp->vfs_data; 1482219089Spjd ASSERT(zfsvfs); 1483219089Spjd if (error = zfs_zget(zfsvfs, zfsvfs->z_root, &zp)) { 1484219089Spjd cmn_err(CE_NOTE, "zfs_zget: error %d", error); 1485219089Spjd goto out; 1486219089Spjd } 1487219089Spjd 1488219089Spjd vp = ZTOV(zp); 1489219089Spjd mutex_enter(&vp->v_lock); 1490219089Spjd vp->v_flag |= VROOT; 1491219089Spjd mutex_exit(&vp->v_lock); 1492219089Spjd rootvp = vp; 1493219089Spjd 1494219089Spjd /* 1495219089Spjd * Leave rootvp held. The root file system is never unmounted. 1496219089Spjd */ 1497219089Spjd 1498219089Spjd vfs_add((struct vnode *)0, vfsp, 1499219089Spjd (vfsp->vfs_flag & VFS_RDONLY) ? MS_RDONLY : 0); 1500219089Spjdout: 1501219089Spjd vfs_unlock(vfsp); 1502219089Spjd return (error); 1503219089Spjd } else if (why == ROOT_REMOUNT) { 1504219089Spjd readonly_changed_cb(vfsp->vfs_data, B_FALSE); 1505219089Spjd vfsp->vfs_flag |= VFS_REMOUNT; 1506219089Spjd 1507219089Spjd /* refresh mount options */ 1508219089Spjd zfs_unregister_callbacks(vfsp->vfs_data); 1509219089Spjd return (zfs_register_callbacks(vfsp)); 1510219089Spjd 1511219089Spjd } else if (why == ROOT_UNMOUNT) { 1512219089Spjd zfs_unregister_callbacks((zfsvfs_t *)vfsp->vfs_data); 1513219089Spjd (void) zfs_sync(vfsp, 0, 0); 1514219089Spjd return (0); 1515219089Spjd } 1516219089Spjd 1517219089Spjd /* 1518219089Spjd * if "why" is equal to anything else other than ROOT_INIT, 1519219089Spjd * ROOT_REMOUNT, or ROOT_UNMOUNT, we do not support it. 1520219089Spjd */ 1521219089Spjd return (ENOTSUP); 1522219089Spjd} 1523219089Spjd#endif /* OPENSOLARIS_MOUNTROOT */ 1524219089Spjd 1525168404Spjd/*ARGSUSED*/ 1526168404Spjdstatic int 1527191990Sattiliozfs_mount(vfs_t *vfsp) 1528168404Spjd{ 1529191990Sattilio kthread_t *td = curthread; 1530185029Spjd vnode_t *mvp = vfsp->mnt_vnodecovered; 1531185029Spjd cred_t *cr = td->td_ucred; 1532185029Spjd char *osname; 1533185029Spjd int error = 0; 1534185029Spjd int canwrite; 1535168404Spjd 1536185029Spjd if (vfs_getopt(vfsp->mnt_optnew, "from", (void **)&osname, NULL)) 1537185029Spjd return (EINVAL); 1538185029Spjd 1539168404Spjd /* 1540185029Spjd * If full-owner-access is enabled and delegated administration is 1541185029Spjd * turned on, we must set nosuid. 1542185029Spjd */ 1543185029Spjd if (zfs_super_owner && 1544185029Spjd dsl_deleg_access(osname, ZFS_DELEG_PERM_MOUNT, cr) != ECANCELED) { 1545185029Spjd secpolicy_fs_mount_clearopts(cr, vfsp); 1546185029Spjd } 1547185029Spjd 1548185029Spjd /* 1549185029Spjd * Check for mount privilege? 1550185029Spjd * 1551185029Spjd * If we don't have privilege then see if 1552185029Spjd * we have local permission to allow it 1553185029Spjd */ 1554185029Spjd error = secpolicy_fs_mount(cr, mvp, vfsp); 1555185029Spjd if (error) { 1556212694Smm if (dsl_deleg_access(osname, ZFS_DELEG_PERM_MOUNT, cr) != 0) 1557196944Spjd goto out; 1558196944Spjd 1559196944Spjd if (!(vfsp->vfs_flag & MS_REMOUNT)) { 1560185029Spjd vattr_t vattr; 1561185029Spjd 1562185029Spjd /* 1563185029Spjd * Make sure user is the owner of the mount point 1564185029Spjd * or has sufficient privileges. 1565185029Spjd */ 1566185029Spjd 1567185029Spjd vattr.va_mask = AT_UID; 1568185029Spjd 1569196662Spjd vn_lock(mvp, LK_SHARED | LK_RETRY); 1570212694Smm if (VOP_GETATTR(mvp, &vattr, cr)) { 1571196662Spjd VOP_UNLOCK(mvp, 0); 1572185029Spjd goto out; 1573185029Spjd } 1574185029Spjd 1575185029Spjd if (secpolicy_vnode_owner(mvp, cr, vattr.va_uid) != 0 && 1576185029Spjd VOP_ACCESS(mvp, VWRITE, cr, td) != 0) { 1577196662Spjd VOP_UNLOCK(mvp, 0); 1578185029Spjd goto out; 1579185029Spjd } 1580196662Spjd VOP_UNLOCK(mvp, 0); 1581196944Spjd } 1582185029Spjd 1583196944Spjd secpolicy_fs_mount_clearopts(cr, vfsp); 1584185029Spjd } 1585185029Spjd 1586185029Spjd /* 1587185029Spjd * Refuse to mount a filesystem if we are in a local zone and the 1588185029Spjd * dataset is not visible. 1589185029Spjd */ 1590185029Spjd if (!INGLOBALZONE(curthread) && 1591185029Spjd (!zone_dataset_visible(osname, &canwrite) || !canwrite)) { 1592185029Spjd error = EPERM; 1593185029Spjd goto out; 1594185029Spjd } 1595185029Spjd 1596219089Spjd#ifdef SECLABEL 1597219089Spjd error = zfs_mount_label_policy(vfsp, osname); 1598219089Spjd if (error) 1599219089Spjd goto out; 1600219089Spjd#endif 1601219089Spjd 1602218386Strasz vfsp->vfs_flag |= MNT_NFS4ACLS; 1603218386Strasz 1604185029Spjd /* 1605168404Spjd * When doing a remount, we simply refresh our temporary properties 1606168404Spjd * according to those options set in the current VFS options. 1607168404Spjd */ 1608185029Spjd if (vfsp->vfs_flag & MS_REMOUNT) { 1609185029Spjd /* refresh mount options */ 1610185029Spjd zfs_unregister_callbacks(vfsp->vfs_data); 1611185029Spjd error = zfs_register_callbacks(vfsp); 1612185029Spjd goto out; 1613185029Spjd } 1614168404Spjd 1615168510Spjd DROP_GIANT(); 1616185029Spjd error = zfs_domount(vfsp, osname); 1617168510Spjd PICKUP_GIANT(); 1618209962Smm 1619215260Smm#ifdef sun 1620209962Smm /* 1621209962Smm * Add an extra VFS_HOLD on our parent vfs so that it can't 1622209962Smm * disappear due to a forced unmount. 1623209962Smm */ 1624209962Smm if (error == 0 && ((zfsvfs_t *)vfsp->vfs_data)->z_issnap) 1625209962Smm VFS_HOLD(mvp->v_vfsp); 1626215260Smm#endif /* sun */ 1627209962Smm 1628185029Spjdout: 1629168510Spjd return (error); 1630168404Spjd} 1631168404Spjd 1632168404Spjdstatic int 1633191990Sattiliozfs_statfs(vfs_t *vfsp, struct statfs *statp) 1634169170Spjd{ 1635168404Spjd zfsvfs_t *zfsvfs = vfsp->vfs_data; 1636168404Spjd uint64_t refdbytes, availbytes, usedobjs, availobjs; 1637168404Spjd 1638168404Spjd statp->f_version = STATFS_VERSION; 1639168404Spjd 1640168404Spjd ZFS_ENTER(zfsvfs); 1641168404Spjd 1642168404Spjd dmu_objset_space(zfsvfs->z_os, 1643168404Spjd &refdbytes, &availbytes, &usedobjs, &availobjs); 1644168404Spjd 1645168404Spjd /* 1646168404Spjd * The underlying storage pool actually uses multiple block sizes. 1647168404Spjd * We report the fragsize as the smallest block size we support, 1648168404Spjd * and we report our blocksize as the filesystem's maximum blocksize. 1649168404Spjd */ 1650204101Spjd statp->f_bsize = SPA_MINBLOCKSIZE; 1651204101Spjd statp->f_iosize = zfsvfs->z_vfs->mnt_stat.f_iosize; 1652168404Spjd 1653168404Spjd /* 1654168404Spjd * The following report "total" blocks of various kinds in the 1655168404Spjd * file system, but reported in terms of f_frsize - the 1656168404Spjd * "fragment" size. 1657168404Spjd */ 1658168404Spjd 1659204101Spjd statp->f_blocks = (refdbytes + availbytes) >> SPA_MINBLOCKSHIFT; 1660168404Spjd statp->f_bfree = availbytes / statp->f_bsize; 1661168404Spjd statp->f_bavail = statp->f_bfree; /* no root reservation */ 1662168404Spjd 1663168404Spjd /* 1664168404Spjd * statvfs() should really be called statufs(), because it assumes 1665168404Spjd * static metadata. ZFS doesn't preallocate files, so the best 1666168404Spjd * we can do is report the max that could possibly fit in f_files, 1667168404Spjd * and that minus the number actually used in f_ffree. 1668168404Spjd * For f_ffree, report the smaller of the number of object available 1669168404Spjd * and the number of blocks (each object will take at least a block). 1670168404Spjd */ 1671168404Spjd statp->f_ffree = MIN(availobjs, statp->f_bfree); 1672168404Spjd statp->f_files = statp->f_ffree + usedobjs; 1673168404Spjd 1674168404Spjd /* 1675168404Spjd * We're a zfs filesystem. 1676168404Spjd */ 1677168404Spjd (void) strlcpy(statp->f_fstypename, "zfs", sizeof(statp->f_fstypename)); 1678168404Spjd 1679168404Spjd strlcpy(statp->f_mntfromname, vfsp->mnt_stat.f_mntfromname, 1680168404Spjd sizeof(statp->f_mntfromname)); 1681168404Spjd strlcpy(statp->f_mntonname, vfsp->mnt_stat.f_mntonname, 1682168404Spjd sizeof(statp->f_mntonname)); 1683168404Spjd 1684168404Spjd statp->f_namemax = ZFS_MAXNAMELEN; 1685168404Spjd 1686168404Spjd ZFS_EXIT(zfsvfs); 1687168404Spjd return (0); 1688168404Spjd} 1689168404Spjd 1690219089Spjdint 1691219089Spjdzfs_vnode_lock(vnode_t *vp, int flags) 1692219089Spjd{ 1693219089Spjd int error; 1694219089Spjd 1695219089Spjd ASSERT(vp != NULL); 1696219089Spjd 1697219089Spjd /* 1698219089Spjd * Check if the file system wasn't forcibly unmounted in the meantime. 1699219089Spjd */ 1700219089Spjd error = vn_lock(vp, flags); 1701219089Spjd if (error == 0 && (vp->v_iflag & VI_DOOMED) != 0) { 1702219089Spjd VOP_UNLOCK(vp, 0); 1703219089Spjd error = ENOENT; 1704219089Spjd } 1705219089Spjd 1706219089Spjd return (error); 1707219089Spjd} 1708219089Spjd 1709168404Spjdstatic int 1710191990Sattiliozfs_root(vfs_t *vfsp, int flags, vnode_t **vpp) 1711168404Spjd{ 1712168404Spjd zfsvfs_t *zfsvfs = vfsp->vfs_data; 1713168404Spjd znode_t *rootzp; 1714168404Spjd int error; 1715168404Spjd 1716197459Spjd ZFS_ENTER_NOERROR(zfsvfs); 1717168404Spjd 1718168404Spjd error = zfs_zget(zfsvfs, zfsvfs->z_root, &rootzp); 1719219089Spjd if (error == 0) 1720219089Spjd *vpp = ZTOV(rootzp); 1721206667Spjd 1722206667Spjd ZFS_EXIT(zfsvfs); 1723206667Spjd 1724168404Spjd if (error == 0) { 1725219089Spjd error = zfs_vnode_lock(*vpp, flags); 1726219089Spjd if (error == 0) 1727219089Spjd (*vpp)->v_vflag |= VV_ROOT; 1728168404Spjd } 1729219089Spjd if (error != 0) 1730219089Spjd *vpp = NULL; 1731168404Spjd 1732168404Spjd return (error); 1733168404Spjd} 1734168404Spjd 1735185029Spjd/* 1736185029Spjd * Teardown the zfsvfs::z_os. 1737185029Spjd * 1738185029Spjd * Note, if 'unmounting' if FALSE, we return with the 'z_teardown_lock' 1739185029Spjd * and 'z_teardown_inactive_lock' held. 1740185029Spjd */ 1741185029Spjdstatic int 1742185029Spjdzfsvfs_teardown(zfsvfs_t *zfsvfs, boolean_t unmounting) 1743185029Spjd{ 1744185029Spjd znode_t *zp; 1745185029Spjd 1746185029Spjd rrw_enter(&zfsvfs->z_teardown_lock, RW_WRITER, FTAG); 1747185029Spjd 1748185029Spjd if (!unmounting) { 1749185029Spjd /* 1750185029Spjd * We purge the parent filesystem's vfsp as the parent 1751185029Spjd * filesystem and all of its snapshots have their vnode's 1752185029Spjd * v_vfsp set to the parent's filesystem's vfsp. Note, 1753185029Spjd * 'z_parent' is self referential for non-snapshots. 1754185029Spjd */ 1755185029Spjd (void) dnlc_purge_vfsp(zfsvfs->z_parent->z_vfs, 0); 1756197351Spjd#ifdef FREEBSD_NAMECACHE 1757197351Spjd cache_purgevfs(zfsvfs->z_parent->z_vfs); 1758197351Spjd#endif 1759185029Spjd } 1760185029Spjd 1761185029Spjd /* 1762185029Spjd * Close the zil. NB: Can't close the zil while zfs_inactive 1763185029Spjd * threads are blocked as zil_close can call zfs_inactive. 1764185029Spjd */ 1765185029Spjd if (zfsvfs->z_log) { 1766185029Spjd zil_close(zfsvfs->z_log); 1767185029Spjd zfsvfs->z_log = NULL; 1768185029Spjd } 1769185029Spjd 1770185029Spjd rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_WRITER); 1771185029Spjd 1772185029Spjd /* 1773185029Spjd * If we are not unmounting (ie: online recv) and someone already 1774185029Spjd * unmounted this file system while we were doing the switcheroo, 1775185029Spjd * or a reopen of z_os failed then just bail out now. 1776185029Spjd */ 1777185029Spjd if (!unmounting && (zfsvfs->z_unmounted || zfsvfs->z_os == NULL)) { 1778185029Spjd rw_exit(&zfsvfs->z_teardown_inactive_lock); 1779185029Spjd rrw_exit(&zfsvfs->z_teardown_lock, FTAG); 1780185029Spjd return (EIO); 1781185029Spjd } 1782185029Spjd 1783185029Spjd /* 1784185029Spjd * At this point there are no vops active, and any new vops will 1785185029Spjd * fail with EIO since we have z_teardown_lock for writer (only 1786185029Spjd * relavent for forced unmount). 1787185029Spjd * 1788185029Spjd * Release all holds on dbufs. 1789185029Spjd */ 1790185029Spjd mutex_enter(&zfsvfs->z_znodes_lock); 1791185029Spjd for (zp = list_head(&zfsvfs->z_all_znodes); zp != NULL; 1792185029Spjd zp = list_next(&zfsvfs->z_all_znodes, zp)) 1793219089Spjd if (zp->z_sa_hdl) { 1794196297Spjd ASSERT(ZTOV(zp)->v_count >= 0); 1795185029Spjd zfs_znode_dmu_fini(zp); 1796185029Spjd } 1797185029Spjd mutex_exit(&zfsvfs->z_znodes_lock); 1798185029Spjd 1799185029Spjd /* 1800185029Spjd * If we are unmounting, set the unmounted flag and let new vops 1801185029Spjd * unblock. zfs_inactive will have the unmounted behavior, and all 1802185029Spjd * other vops will fail with EIO. 1803185029Spjd */ 1804185029Spjd if (unmounting) { 1805185029Spjd zfsvfs->z_unmounted = B_TRUE; 1806185029Spjd rrw_exit(&zfsvfs->z_teardown_lock, FTAG); 1807185029Spjd rw_exit(&zfsvfs->z_teardown_inactive_lock); 1808197133Spjd 1809197133Spjd#ifdef __FreeBSD__ 1810197133Spjd /* 1811197133Spjd * Some znodes might not be fully reclaimed, wait for them. 1812197133Spjd */ 1813197133Spjd mutex_enter(&zfsvfs->z_znodes_lock); 1814197133Spjd while (list_head(&zfsvfs->z_all_znodes) != NULL) { 1815197133Spjd msleep(zfsvfs, &zfsvfs->z_znodes_lock, 0, 1816197133Spjd "zteardown", 0); 1817197133Spjd } 1818197133Spjd mutex_exit(&zfsvfs->z_znodes_lock); 1819197133Spjd#endif 1820185029Spjd } 1821185029Spjd 1822185029Spjd /* 1823185029Spjd * z_os will be NULL if there was an error in attempting to reopen 1824185029Spjd * zfsvfs, so just return as the properties had already been 1825185029Spjd * unregistered and cached data had been evicted before. 1826185029Spjd */ 1827185029Spjd if (zfsvfs->z_os == NULL) 1828185029Spjd return (0); 1829185029Spjd 1830185029Spjd /* 1831185029Spjd * Unregister properties. 1832185029Spjd */ 1833185029Spjd zfs_unregister_callbacks(zfsvfs); 1834185029Spjd 1835185029Spjd /* 1836185029Spjd * Evict cached data 1837185029Spjd */ 1838219089Spjd if (dmu_objset_is_dirty_anywhere(zfsvfs->z_os)) 1839219089Spjd if (!(zfsvfs->z_vfs->vfs_flag & VFS_RDONLY)) 1840219089Spjd txg_wait_synced(dmu_objset_pool(zfsvfs->z_os), 0); 1841219089Spjd (void) dmu_objset_evict_dbufs(zfsvfs->z_os); 1842185029Spjd 1843185029Spjd return (0); 1844185029Spjd} 1845185029Spjd 1846168404Spjd/*ARGSUSED*/ 1847168404Spjdstatic int 1848191990Sattiliozfs_umount(vfs_t *vfsp, int fflag) 1849168404Spjd{ 1850209962Smm kthread_t *td = curthread; 1851168404Spjd zfsvfs_t *zfsvfs = vfsp->vfs_data; 1852185029Spjd objset_t *os; 1853209962Smm cred_t *cr = td->td_ucred; 1854168404Spjd int ret; 1855168404Spjd 1856185029Spjd ret = secpolicy_fs_unmount(cr, vfsp); 1857185029Spjd if (ret) { 1858212694Smm if (dsl_deleg_access((char *)refstr_value(vfsp->vfs_resource), 1859212694Smm ZFS_DELEG_PERM_MOUNT, cr)) 1860185029Spjd return (ret); 1861185029Spjd } 1862219089Spjd 1863185029Spjd /* 1864185029Spjd * We purge the parent filesystem's vfsp as the parent filesystem 1865185029Spjd * and all of its snapshots have their vnode's v_vfsp set to the 1866185029Spjd * parent's filesystem's vfsp. Note, 'z_parent' is self 1867185029Spjd * referential for non-snapshots. 1868185029Spjd */ 1869185029Spjd (void) dnlc_purge_vfsp(zfsvfs->z_parent->z_vfs, 0); 1870168404Spjd 1871168404Spjd /* 1872168404Spjd * Unmount any snapshots mounted under .zfs before unmounting the 1873168404Spjd * dataset itself. 1874168404Spjd */ 1875169170Spjd if (zfsvfs->z_ctldir != NULL) { 1876168404Spjd if ((ret = zfsctl_umount_snapshots(vfsp, fflag, cr)) != 0) 1877168404Spjd return (ret); 1878209962Smm ret = vflush(vfsp, 0, 0, td); 1879168404Spjd ASSERT(ret == EBUSY); 1880168404Spjd if (!(fflag & MS_FORCE)) { 1881168404Spjd if (zfsvfs->z_ctldir->v_count > 1) 1882168404Spjd return (EBUSY); 1883168404Spjd ASSERT(zfsvfs->z_ctldir->v_count == 1); 1884168404Spjd } 1885168404Spjd zfsctl_destroy(zfsvfs); 1886168404Spjd ASSERT(zfsvfs->z_ctldir == NULL); 1887168404Spjd } 1888168404Spjd 1889197459Spjd if (fflag & MS_FORCE) { 1890197459Spjd /* 1891197459Spjd * Mark file system as unmounted before calling 1892197459Spjd * vflush(FORCECLOSE). This way we ensure no future vnops 1893197459Spjd * will be called and risk operating on DOOMED vnodes. 1894197459Spjd */ 1895197459Spjd rrw_enter(&zfsvfs->z_teardown_lock, RW_WRITER, FTAG); 1896197459Spjd zfsvfs->z_unmounted = B_TRUE; 1897197459Spjd rrw_exit(&zfsvfs->z_teardown_lock, FTAG); 1898197459Spjd } 1899197459Spjd 1900168404Spjd /* 1901168404Spjd * Flush all the files. 1902168404Spjd */ 1903209962Smm ret = vflush(vfsp, 1, (fflag & MS_FORCE) ? FORCECLOSE : 0, td); 1904168404Spjd if (ret != 0) { 1905168404Spjd if (!zfsvfs->z_issnap) { 1906168404Spjd zfsctl_create(zfsvfs); 1907168404Spjd ASSERT(zfsvfs->z_ctldir != NULL); 1908168404Spjd } 1909168404Spjd return (ret); 1910168404Spjd } 1911168404Spjd 1912185029Spjd if (!(fflag & MS_FORCE)) { 1913185029Spjd /* 1914185029Spjd * Check the number of active vnodes in the file system. 1915185029Spjd * Our count is maintained in the vfs structure, but the 1916185029Spjd * number is off by 1 to indicate a hold on the vfs 1917185029Spjd * structure itself. 1918185029Spjd * 1919185029Spjd * The '.zfs' directory maintains a reference of its 1920185029Spjd * own, and any active references underneath are 1921185029Spjd * reflected in the vnode count. 1922185029Spjd */ 1923185029Spjd if (zfsvfs->z_ctldir == NULL) { 1924185029Spjd if (vfsp->vfs_count > 1) 1925185029Spjd return (EBUSY); 1926185029Spjd } else { 1927185029Spjd if (vfsp->vfs_count > 2 || 1928185029Spjd zfsvfs->z_ctldir->v_count > 1) 1929185029Spjd return (EBUSY); 1930185029Spjd } 1931185029Spjd } else { 1932168404Spjd MNT_ILOCK(vfsp); 1933168404Spjd vfsp->mnt_kern_flag |= MNTK_UNMOUNTF; 1934168404Spjd MNT_IUNLOCK(vfsp); 1935185029Spjd } 1936168404Spjd 1937185029Spjd VERIFY(zfsvfs_teardown(zfsvfs, B_TRUE) == 0); 1938185029Spjd os = zfsvfs->z_os; 1939185029Spjd 1940185029Spjd /* 1941185029Spjd * z_os will be NULL if there was an error in 1942185029Spjd * attempting to reopen zfsvfs. 1943185029Spjd */ 1944185029Spjd if (os != NULL) { 1945168404Spjd /* 1946185029Spjd * Unset the objset user_ptr. 1947168404Spjd */ 1948219089Spjd mutex_enter(&os->os_user_ptr_lock); 1949185029Spjd dmu_objset_set_user(os, NULL); 1950219089Spjd mutex_exit(&os->os_user_ptr_lock); 1951185029Spjd 1952185029Spjd /* 1953185029Spjd * Finally release the objset 1954185029Spjd */ 1955219089Spjd dmu_objset_disown(os, zfsvfs); 1956168404Spjd } 1957168404Spjd 1958185029Spjd /* 1959185029Spjd * We can now safely destroy the '.zfs' directory node. 1960185029Spjd */ 1961185029Spjd if (zfsvfs->z_ctldir != NULL) 1962185029Spjd zfsctl_destroy(zfsvfs); 1963185029Spjd if (zfsvfs->z_issnap) { 1964185029Spjd vnode_t *svp = vfsp->mnt_vnodecovered; 1965185029Spjd 1966197515Spjd if (svp->v_count >= 2) 1967192211Skmacy VN_RELE(svp); 1968185029Spjd } 1969168404Spjd zfs_freevfs(vfsp); 1970168404Spjd 1971168404Spjd return (0); 1972168404Spjd} 1973168404Spjd 1974168404Spjdstatic int 1975168404Spjdzfs_vget(vfs_t *vfsp, ino_t ino, int flags, vnode_t **vpp) 1976168404Spjd{ 1977168404Spjd zfsvfs_t *zfsvfs = vfsp->vfs_data; 1978168404Spjd znode_t *zp; 1979168404Spjd int err; 1980168404Spjd 1981197167Spjd /* 1982215397Savg * zfs_zget() can't operate on virtual entries like .zfs/ or 1983211855Spjd * .zfs/snapshot/ directories, that's why we return EOPNOTSUPP. 1984211855Spjd * This will make NFS to switch to LOOKUP instead of using VGET. 1985197167Spjd */ 1986197167Spjd if (ino == ZFSCTL_INO_ROOT || ino == ZFSCTL_INO_SNAPDIR) 1987197167Spjd return (EOPNOTSUPP); 1988197167Spjd 1989168404Spjd ZFS_ENTER(zfsvfs); 1990168404Spjd err = zfs_zget(zfsvfs, ino, &zp); 1991168404Spjd if (err == 0 && zp->z_unlinked) { 1992168404Spjd VN_RELE(ZTOV(zp)); 1993168404Spjd err = EINVAL; 1994168404Spjd } 1995219089Spjd if (err == 0) 1996219089Spjd *vpp = ZTOV(zp); 1997206667Spjd ZFS_EXIT(zfsvfs); 1998219089Spjd if (err == 0) 1999219089Spjd err = zfs_vnode_lock(*vpp, flags); 2000168404Spjd if (err != 0) 2001168404Spjd *vpp = NULL; 2002171063Sdfr return (err); 2003168404Spjd} 2004168404Spjd 2005168404Spjdstatic int 2006196982Spjdzfs_checkexp(vfs_t *vfsp, struct sockaddr *nam, int *extflagsp, 2007196982Spjd struct ucred **credanonp, int *numsecflavors, int **secflavors) 2008196982Spjd{ 2009196982Spjd zfsvfs_t *zfsvfs = vfsp->vfs_data; 2010196982Spjd 2011196982Spjd /* 2012196982Spjd * If this is regular file system vfsp is the same as 2013196982Spjd * zfsvfs->z_parent->z_vfs, but if it is snapshot, 2014196982Spjd * zfsvfs->z_parent->z_vfs represents parent file system 2015196982Spjd * which we have to use here, because only this file system 2016196982Spjd * has mnt_export configured. 2017196982Spjd */ 2018196982Spjd return (vfs_stdcheckexp(zfsvfs->z_parent->z_vfs, nam, extflagsp, 2019196982Spjd credanonp, numsecflavors, secflavors)); 2020196982Spjd} 2021196982Spjd 2022197151SpjdCTASSERT(SHORT_FID_LEN <= sizeof(struct fid)); 2023197151SpjdCTASSERT(LONG_FID_LEN <= sizeof(struct fid)); 2024196982Spjd 2025196982Spjdstatic int 2026222167Srmacklemzfs_fhtovp(vfs_t *vfsp, fid_t *fidp, int flags, vnode_t **vpp) 2027168404Spjd{ 2028168404Spjd zfsvfs_t *zfsvfs = vfsp->vfs_data; 2029168404Spjd znode_t *zp; 2030168404Spjd uint64_t object = 0; 2031168404Spjd uint64_t fid_gen = 0; 2032168404Spjd uint64_t gen_mask; 2033168404Spjd uint64_t zp_gen; 2034219089Spjd int i, err; 2035168404Spjd 2036168404Spjd *vpp = NULL; 2037168404Spjd 2038168404Spjd ZFS_ENTER(zfsvfs); 2039168404Spjd 2040196979Spjd /* 2041197177Spjd * On FreeBSD we can get snapshot's mount point or its parent file 2042197177Spjd * system mount point depending if snapshot is already mounted or not. 2043196979Spjd */ 2044197177Spjd if (zfsvfs->z_parent == zfsvfs && fidp->fid_len == LONG_FID_LEN) { 2045168404Spjd zfid_long_t *zlfid = (zfid_long_t *)fidp; 2046168404Spjd uint64_t objsetid = 0; 2047168404Spjd uint64_t setgen = 0; 2048168404Spjd 2049168404Spjd for (i = 0; i < sizeof (zlfid->zf_setid); i++) 2050168404Spjd objsetid |= ((uint64_t)zlfid->zf_setid[i]) << (8 * i); 2051168404Spjd 2052168404Spjd for (i = 0; i < sizeof (zlfid->zf_setgen); i++) 2053168404Spjd setgen |= ((uint64_t)zlfid->zf_setgen[i]) << (8 * i); 2054168404Spjd 2055168404Spjd ZFS_EXIT(zfsvfs); 2056168404Spjd 2057168404Spjd err = zfsctl_lookup_objset(vfsp, objsetid, &zfsvfs); 2058168404Spjd if (err) 2059168404Spjd return (EINVAL); 2060168404Spjd ZFS_ENTER(zfsvfs); 2061168404Spjd } 2062168404Spjd 2063168404Spjd if (fidp->fid_len == SHORT_FID_LEN || fidp->fid_len == LONG_FID_LEN) { 2064168404Spjd zfid_short_t *zfid = (zfid_short_t *)fidp; 2065168404Spjd 2066168404Spjd for (i = 0; i < sizeof (zfid->zf_object); i++) 2067168404Spjd object |= ((uint64_t)zfid->zf_object[i]) << (8 * i); 2068168404Spjd 2069168404Spjd for (i = 0; i < sizeof (zfid->zf_gen); i++) 2070168404Spjd fid_gen |= ((uint64_t)zfid->zf_gen[i]) << (8 * i); 2071168404Spjd } else { 2072168404Spjd ZFS_EXIT(zfsvfs); 2073168404Spjd return (EINVAL); 2074168404Spjd } 2075168404Spjd 2076168404Spjd /* A zero fid_gen means we are in the .zfs control directories */ 2077168404Spjd if (fid_gen == 0 && 2078168404Spjd (object == ZFSCTL_INO_ROOT || object == ZFSCTL_INO_SNAPDIR)) { 2079168404Spjd *vpp = zfsvfs->z_ctldir; 2080168404Spjd ASSERT(*vpp != NULL); 2081168404Spjd if (object == ZFSCTL_INO_SNAPDIR) { 2082168404Spjd VERIFY(zfsctl_root_lookup(*vpp, "snapshot", vpp, NULL, 2083185029Spjd 0, NULL, NULL, NULL, NULL, NULL) == 0); 2084168404Spjd } else { 2085168404Spjd VN_HOLD(*vpp); 2086168404Spjd } 2087206667Spjd ZFS_EXIT(zfsvfs); 2088222199Srmacklem err = zfs_vnode_lock(*vpp, flags | LK_RETRY); 2089219089Spjd if (err != 0) 2090219089Spjd *vpp = NULL; 2091219089Spjd return (err); 2092168404Spjd } 2093168404Spjd 2094168404Spjd gen_mask = -1ULL >> (64 - 8 * i); 2095168404Spjd 2096168404Spjd dprintf("getting %llu [%u mask %llx]\n", object, fid_gen, gen_mask); 2097168404Spjd if (err = zfs_zget(zfsvfs, object, &zp)) { 2098168404Spjd ZFS_EXIT(zfsvfs); 2099168404Spjd return (err); 2100168404Spjd } 2101219089Spjd (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zfsvfs), &zp_gen, 2102219089Spjd sizeof (uint64_t)); 2103219089Spjd zp_gen = zp_gen & gen_mask; 2104168404Spjd if (zp_gen == 0) 2105168404Spjd zp_gen = 1; 2106168404Spjd if (zp->z_unlinked || zp_gen != fid_gen) { 2107168404Spjd dprintf("znode gen (%u) != fid gen (%u)\n", zp_gen, fid_gen); 2108168404Spjd VN_RELE(ZTOV(zp)); 2109168404Spjd ZFS_EXIT(zfsvfs); 2110168404Spjd return (EINVAL); 2111168404Spjd } 2112168404Spjd 2113219089Spjd *vpp = ZTOV(zp); 2114206667Spjd ZFS_EXIT(zfsvfs); 2115222199Srmacklem err = zfs_vnode_lock(*vpp, flags | LK_RETRY); 2116219089Spjd if (err == 0) 2117219089Spjd vnode_create_vobject(*vpp, zp->z_size, curthread); 2118219089Spjd else 2119219089Spjd *vpp = NULL; 2120219089Spjd return (err); 2121168404Spjd} 2122168404Spjd 2123185029Spjd/* 2124185029Spjd * Block out VOPs and close zfsvfs_t::z_os 2125185029Spjd * 2126185029Spjd * Note, if successful, then we return with the 'z_teardown_lock' and 2127185029Spjd * 'z_teardown_inactive_lock' write held. 2128185029Spjd */ 2129185029Spjdint 2130219089Spjdzfs_suspend_fs(zfsvfs_t *zfsvfs) 2131168404Spjd{ 2132185029Spjd int error; 2133168404Spjd 2134185029Spjd if ((error = zfsvfs_teardown(zfsvfs, B_FALSE)) != 0) 2135185029Spjd return (error); 2136219089Spjd dmu_objset_disown(zfsvfs->z_os, zfsvfs); 2137168404Spjd 2138185029Spjd return (0); 2139185029Spjd} 2140168404Spjd 2141185029Spjd/* 2142185029Spjd * Reopen zfsvfs_t::z_os and release VOPs. 2143185029Spjd */ 2144185029Spjdint 2145219089Spjdzfs_resume_fs(zfsvfs_t *zfsvfs, const char *osname) 2146185029Spjd{ 2147185029Spjd int err; 2148168404Spjd 2149185029Spjd ASSERT(RRW_WRITE_HELD(&zfsvfs->z_teardown_lock)); 2150185029Spjd ASSERT(RW_WRITE_HELD(&zfsvfs->z_teardown_inactive_lock)); 2151185029Spjd 2152219089Spjd err = dmu_objset_own(osname, DMU_OST_ZFS, B_FALSE, zfsvfs, 2153219089Spjd &zfsvfs->z_os); 2154185029Spjd if (err) { 2155185029Spjd zfsvfs->z_os = NULL; 2156185029Spjd } else { 2157185029Spjd znode_t *zp; 2158219089Spjd uint64_t sa_obj = 0; 2159185029Spjd 2160219089Spjd /* 2161219089Spjd * Make sure version hasn't changed 2162219089Spjd */ 2163219089Spjd 2164219089Spjd err = zfs_get_zplprop(zfsvfs->z_os, ZFS_PROP_VERSION, 2165219089Spjd &zfsvfs->z_version); 2166219089Spjd 2167219089Spjd if (err) 2168219089Spjd goto bail; 2169219089Spjd 2170219089Spjd err = zap_lookup(zfsvfs->z_os, MASTER_NODE_OBJ, 2171219089Spjd ZFS_SA_ATTRS, 8, 1, &sa_obj); 2172219089Spjd 2173219089Spjd if (err && zfsvfs->z_version >= ZPL_VERSION_SA) 2174219089Spjd goto bail; 2175219089Spjd 2176219089Spjd if ((err = sa_setup(zfsvfs->z_os, sa_obj, 2177219089Spjd zfs_attr_table, ZPL_END, &zfsvfs->z_attr_table)) != 0) 2178219089Spjd goto bail; 2179219089Spjd 2180219089Spjd if (zfsvfs->z_version >= ZPL_VERSION_SA) 2181219089Spjd sa_register_update_callback(zfsvfs->z_os, 2182219089Spjd zfs_sa_upgrade); 2183219089Spjd 2184185029Spjd VERIFY(zfsvfs_setup(zfsvfs, B_FALSE) == 0); 2185185029Spjd 2186219089Spjd zfs_set_fuid_feature(zfsvfs); 2187219089Spjd 2188185029Spjd /* 2189185029Spjd * Attempt to re-establish all the active znodes with 2190185029Spjd * their dbufs. If a zfs_rezget() fails, then we'll let 2191185029Spjd * any potential callers discover that via ZFS_ENTER_VERIFY_VP 2192185029Spjd * when they try to use their znode. 2193185029Spjd */ 2194185029Spjd mutex_enter(&zfsvfs->z_znodes_lock); 2195185029Spjd for (zp = list_head(&zfsvfs->z_all_znodes); zp; 2196185029Spjd zp = list_next(&zfsvfs->z_all_znodes, zp)) { 2197185029Spjd (void) zfs_rezget(zp); 2198185029Spjd } 2199185029Spjd mutex_exit(&zfsvfs->z_znodes_lock); 2200168404Spjd } 2201168404Spjd 2202219089Spjdbail: 2203185029Spjd /* release the VOPs */ 2204185029Spjd rw_exit(&zfsvfs->z_teardown_inactive_lock); 2205185029Spjd rrw_exit(&zfsvfs->z_teardown_lock, FTAG); 2206185029Spjd 2207185029Spjd if (err) { 2208185029Spjd /* 2209219089Spjd * Since we couldn't reopen zfsvfs::z_os, or 2210219089Spjd * setup the sa framework force unmount this file system. 2211185029Spjd */ 2212185029Spjd if (vn_vfswlock(zfsvfs->z_vfs->vfs_vnodecovered) == 0) 2213185029Spjd (void) dounmount(zfsvfs->z_vfs, MS_FORCE, curthread); 2214168404Spjd } 2215185029Spjd return (err); 2216168404Spjd} 2217168404Spjd 2218168404Spjdstatic void 2219168404Spjdzfs_freevfs(vfs_t *vfsp) 2220168404Spjd{ 2221168404Spjd zfsvfs_t *zfsvfs = vfsp->vfs_data; 2222168404Spjd 2223215260Smm#ifdef sun 2224209962Smm /* 2225209962Smm * If this is a snapshot, we have an extra VFS_HOLD on our parent 2226219089Spjd * from zfs_mount(). Release it here. If we came through 2227219089Spjd * zfs_mountroot() instead, we didn't grab an extra hold, so 2228219089Spjd * skip the VFS_RELE for rootvfs. 2229209962Smm */ 2230219089Spjd if (zfsvfs->z_issnap && (vfsp != rootvfs)) 2231209962Smm VFS_RELE(zfsvfs->z_parent->z_vfs); 2232215260Smm#endif /* sun */ 2233168404Spjd 2234209962Smm zfsvfs_free(zfsvfs); 2235185029Spjd 2236168404Spjd atomic_add_32(&zfs_active_fs_count, -1); 2237168404Spjd} 2238168404Spjd 2239172135Spjd#ifdef __i386__ 2240172135Spjdstatic int desiredvnodes_backup; 2241172135Spjd#endif 2242172135Spjd 2243172135Spjdstatic void 2244172135Spjdzfs_vnodes_adjust(void) 2245172135Spjd{ 2246172135Spjd#ifdef __i386__ 2247185029Spjd int newdesiredvnodes; 2248172135Spjd 2249172135Spjd desiredvnodes_backup = desiredvnodes; 2250172135Spjd 2251172135Spjd /* 2252172135Spjd * We calculate newdesiredvnodes the same way it is done in 2253172135Spjd * vntblinit(). If it is equal to desiredvnodes, it means that 2254172135Spjd * it wasn't tuned by the administrator and we can tune it down. 2255172135Spjd */ 2256185029Spjd newdesiredvnodes = min(maxproc + cnt.v_page_count / 4, 2 * 2257185029Spjd vm_kmem_size / (5 * (sizeof(struct vm_object) + 2258185029Spjd sizeof(struct vnode)))); 2259185029Spjd if (newdesiredvnodes == desiredvnodes) 2260185029Spjd desiredvnodes = (3 * newdesiredvnodes) / 4; 2261172135Spjd#endif 2262172135Spjd} 2263172135Spjd 2264172135Spjdstatic void 2265172135Spjdzfs_vnodes_adjust_back(void) 2266172135Spjd{ 2267172135Spjd 2268172135Spjd#ifdef __i386__ 2269172135Spjd desiredvnodes = desiredvnodes_backup; 2270172135Spjd#endif 2271172135Spjd} 2272172135Spjd 2273168404Spjdvoid 2274168404Spjdzfs_init(void) 2275168404Spjd{ 2276168404Spjd 2277202129Sdelphij printf("ZFS filesystem version " ZPL_VERSION_STRING "\n"); 2278168404Spjd 2279168404Spjd /* 2280219089Spjd * Initialize .zfs directory structures 2281168404Spjd */ 2282219089Spjd zfsctl_init(); 2283168404Spjd 2284168404Spjd /* 2285219089Spjd * Initialize znode cache, vnode ops, etc... 2286168404Spjd */ 2287219089Spjd zfs_znode_init(); 2288172135Spjd 2289172135Spjd /* 2290219089Spjd * Reduce number of vnodes. Originally number of vnodes is calculated 2291172135Spjd * with UFS inode in mind. We reduce it here, because it's too big for 2292172135Spjd * ZFS/i386. 2293172135Spjd */ 2294172135Spjd zfs_vnodes_adjust(); 2295209962Smm 2296209962Smm dmu_objset_register_type(DMU_OST_ZFS, zfs_space_delta_cb); 2297168404Spjd} 2298168404Spjd 2299168404Spjdvoid 2300168404Spjdzfs_fini(void) 2301168404Spjd{ 2302168404Spjd zfsctl_fini(); 2303168404Spjd zfs_znode_fini(); 2304172135Spjd zfs_vnodes_adjust_back(); 2305168404Spjd} 2306168404Spjd 2307168404Spjdint 2308168404Spjdzfs_busy(void) 2309168404Spjd{ 2310168404Spjd return (zfs_active_fs_count != 0); 2311168404Spjd} 2312185029Spjd 2313185029Spjdint 2314209962Smmzfs_set_version(zfsvfs_t *zfsvfs, uint64_t newvers) 2315185029Spjd{ 2316185029Spjd int error; 2317209962Smm objset_t *os = zfsvfs->z_os; 2318185029Spjd dmu_tx_t *tx; 2319185029Spjd 2320185029Spjd if (newvers < ZPL_VERSION_INITIAL || newvers > ZPL_VERSION) 2321185029Spjd return (EINVAL); 2322185029Spjd 2323209962Smm if (newvers < zfsvfs->z_version) 2324209962Smm return (EINVAL); 2325185029Spjd 2326219089Spjd if (zfs_spa_version_map(newvers) > 2327219089Spjd spa_version(dmu_objset_spa(zfsvfs->z_os))) 2328219089Spjd return (ENOTSUP); 2329219089Spjd 2330185029Spjd tx = dmu_tx_create(os); 2331209962Smm dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_FALSE, ZPL_VERSION_STR); 2332219089Spjd if (newvers >= ZPL_VERSION_SA && !zfsvfs->z_use_sa) { 2333219089Spjd dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, B_TRUE, 2334219089Spjd ZFS_SA_ATTRS); 2335219089Spjd dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); 2336219089Spjd } 2337185029Spjd error = dmu_tx_assign(tx, TXG_WAIT); 2338185029Spjd if (error) { 2339185029Spjd dmu_tx_abort(tx); 2340209962Smm return (error); 2341185029Spjd } 2342219089Spjd 2343209962Smm error = zap_update(os, MASTER_NODE_OBJ, ZPL_VERSION_STR, 2344209962Smm 8, 1, &newvers, tx); 2345185029Spjd 2346209962Smm if (error) { 2347209962Smm dmu_tx_commit(tx); 2348209962Smm return (error); 2349209962Smm } 2350209962Smm 2351219089Spjd if (newvers >= ZPL_VERSION_SA && !zfsvfs->z_use_sa) { 2352219089Spjd uint64_t sa_obj; 2353219089Spjd 2354219089Spjd ASSERT3U(spa_version(dmu_objset_spa(zfsvfs->z_os)), >=, 2355219089Spjd SPA_VERSION_SA); 2356219089Spjd sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE, 2357219089Spjd DMU_OT_NONE, 0, tx); 2358219089Spjd 2359219089Spjd error = zap_add(os, MASTER_NODE_OBJ, 2360219089Spjd ZFS_SA_ATTRS, 8, 1, &sa_obj, tx); 2361219089Spjd ASSERT3U(error, ==, 0); 2362219089Spjd 2363219089Spjd VERIFY(0 == sa_set_sa_object(os, sa_obj)); 2364219089Spjd sa_register_update_callback(os, zfs_sa_upgrade); 2365219089Spjd } 2366219089Spjd 2367219089Spjd spa_history_log_internal(LOG_DS_UPGRADE, 2368219089Spjd dmu_objset_spa(os), tx, "oldver=%llu newver=%llu dataset = %llu", 2369209962Smm zfsvfs->z_version, newvers, dmu_objset_id(os)); 2370209962Smm 2371185029Spjd dmu_tx_commit(tx); 2372185029Spjd 2373209962Smm zfsvfs->z_version = newvers; 2374209962Smm 2375219089Spjd zfs_set_fuid_feature(zfsvfs); 2376209962Smm 2377209962Smm return (0); 2378185029Spjd} 2379219089Spjd 2380185029Spjd/* 2381185029Spjd * Read a property stored within the master node. 2382185029Spjd */ 2383185029Spjdint 2384185029Spjdzfs_get_zplprop(objset_t *os, zfs_prop_t prop, uint64_t *value) 2385185029Spjd{ 2386185029Spjd const char *pname; 2387185029Spjd int error = ENOENT; 2388185029Spjd 2389185029Spjd /* 2390185029Spjd * Look up the file system's value for the property. For the 2391185029Spjd * version property, we look up a slightly different string. 2392185029Spjd */ 2393185029Spjd if (prop == ZFS_PROP_VERSION) 2394185029Spjd pname = ZPL_VERSION_STR; 2395185029Spjd else 2396185029Spjd pname = zfs_prop_to_name(prop); 2397185029Spjd 2398185029Spjd if (os != NULL) 2399185029Spjd error = zap_lookup(os, MASTER_NODE_OBJ, pname, 8, 1, value); 2400185029Spjd 2401185029Spjd if (error == ENOENT) { 2402185029Spjd /* No value set, use the default value */ 2403185029Spjd switch (prop) { 2404185029Spjd case ZFS_PROP_VERSION: 2405185029Spjd *value = ZPL_VERSION; 2406185029Spjd break; 2407185029Spjd case ZFS_PROP_NORMALIZE: 2408185029Spjd case ZFS_PROP_UTF8ONLY: 2409185029Spjd *value = 0; 2410185029Spjd break; 2411185029Spjd case ZFS_PROP_CASE: 2412185029Spjd *value = ZFS_CASE_SENSITIVE; 2413185029Spjd break; 2414185029Spjd default: 2415185029Spjd return (error); 2416185029Spjd } 2417185029Spjd error = 0; 2418185029Spjd } 2419185029Spjd return (error); 2420185029Spjd} 2421229565Smm 2422229565Smm#ifdef _KERNEL 2423229565Smmvoid 2424229565Smmzfsvfs_update_fromname(const char *oldname, const char *newname) 2425229565Smm{ 2426229565Smm char tmpbuf[MAXPATHLEN]; 2427229565Smm struct mount *mp; 2428229565Smm char *fromname; 2429229565Smm size_t oldlen; 2430229565Smm 2431229565Smm oldlen = strlen(oldname); 2432229565Smm 2433229565Smm mtx_lock(&mountlist_mtx); 2434229565Smm TAILQ_FOREACH(mp, &mountlist, mnt_list) { 2435229565Smm fromname = mp->mnt_stat.f_mntfromname; 2436229565Smm if (strcmp(fromname, oldname) == 0) { 2437229565Smm (void)strlcpy(fromname, newname, 2438229565Smm sizeof(mp->mnt_stat.f_mntfromname)); 2439229565Smm continue; 2440229565Smm } 2441229565Smm if (strncmp(fromname, oldname, oldlen) == 0 && 2442229565Smm (fromname[oldlen] == '/' || fromname[oldlen] == '@')) { 2443229565Smm (void)snprintf(tmpbuf, sizeof(tmpbuf), "%s%s", 2444229565Smm newname, fromname + oldlen); 2445229565Smm (void)strlcpy(fromname, tmpbuf, 2446229565Smm sizeof(mp->mnt_stat.f_mntfromname)); 2447229565Smm continue; 2448229565Smm } 2449229565Smm } 2450229565Smm mtx_unlock(&mountlist_mtx); 2451229565Smm} 2452229565Smm#endif 2453