1fa9e4066Sahrens /* 2fa9e4066Sahrens * CDDL HEADER START 3fa9e4066Sahrens * 4fa9e4066Sahrens * The contents of this file are subject to the terms of the 5ea8dc4b6Seschrock * Common Development and Distribution License (the "License"). 6ea8dc4b6Seschrock * You may not use this file except in compliance with the License. 7fa9e4066Sahrens * 8fa9e4066Sahrens * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9fa9e4066Sahrens * or http://www.opensolaris.org/os/licensing. 10fa9e4066Sahrens * See the License for the specific language governing permissions 11fa9e4066Sahrens * and limitations under the License. 12fa9e4066Sahrens * 13fa9e4066Sahrens * When distributing Covered Code, include this CDDL HEADER in each 14fa9e4066Sahrens * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15fa9e4066Sahrens * If applicable, add the following below this CDDL HEADER, with the 16fa9e4066Sahrens * fields enclosed by brackets "[]" replaced with your own identifying 17fa9e4066Sahrens * information: Portions Copyright [yyyy] [name of copyright owner] 18fa9e4066Sahrens * 19fa9e4066Sahrens * CDDL HEADER END 20fa9e4066Sahrens */ 21fa9e4066Sahrens /* 22ac05c741SMark Maybee * Copyright 2009 Sun Microsystems, Inc. All rights reserved. 23fa9e4066Sahrens * Use is subject to license terms. 24fa9e4066Sahrens */ 25fa9e4066Sahrens 2675c76197Speteh /* Portions Copyright 2007 Jeremy Teo */ 2775c76197Speteh 2855434c77Sek #ifdef _KERNEL 29fa9e4066Sahrens #include <sys/types.h> 30fa9e4066Sahrens #include <sys/param.h> 31fa9e4066Sahrens #include <sys/time.h> 32fa9e4066Sahrens #include <sys/systm.h> 33fa9e4066Sahrens #include <sys/sysmacros.h> 34fa9e4066Sahrens #include <sys/resource.h> 35fa9e4066Sahrens #include <sys/mntent.h> 3672fc53bcSmarks #include <sys/mkdev.h> 37de8267e0Stimh #include <sys/u8_textprep.h> 38ab04eb8eStimh #include <sys/dsl_dataset.h> 39fa9e4066Sahrens #include <sys/vfs.h> 40aa59c4cbSrsb #include <sys/vfs_opreg.h> 41fa9e4066Sahrens #include <sys/vnode.h> 42fa9e4066Sahrens #include <sys/file.h> 43fa9e4066Sahrens #include <sys/kmem.h> 44fa9e4066Sahrens #include <sys/errno.h> 45fa9e4066Sahrens #include <sys/unistd.h> 46fa9e4066Sahrens #include <sys/mode.h> 47fa9e4066Sahrens #include <sys/atomic.h> 48fa9e4066Sahrens #include <vm/pvn.h> 49fa9e4066Sahrens #include "fs/fs_subr.h" 50fa9e4066Sahrens #include <sys/zfs_dir.h> 51fa9e4066Sahrens #include <sys/zfs_acl.h> 52fa9e4066Sahrens #include <sys/zfs_ioctl.h> 53104e2ed7Sperrin #include <sys/zfs_rlock.h> 54da6c28aaSamw #include <sys/zfs_fuid.h> 55fa9e4066Sahrens #include <sys/fs/zfs.h> 56da6c28aaSamw #include <sys/kidmap.h> 5755434c77Sek #endif /* _KERNEL */ 5855434c77Sek 5955434c77Sek #include <sys/dmu.h> 6055434c77Sek #include <sys/refcount.h> 6155434c77Sek #include <sys/stat.h> 6255434c77Sek #include <sys/zap.h> 6355434c77Sek #include <sys/zfs_znode.h> 64fa9e4066Sahrens 65de8267e0Stimh #include "zfs_prop.h" 66de8267e0Stimh 67b5fca8f8Stomee /* 68b5fca8f8Stomee * Define ZNODE_STATS to turn on statistic gathering. By default, it is only 69b5fca8f8Stomee * turned on when DEBUG is also defined. 70b5fca8f8Stomee */ 71b5fca8f8Stomee #ifdef DEBUG 72b5fca8f8Stomee #define ZNODE_STATS 73b5fca8f8Stomee #endif /* DEBUG */ 74b5fca8f8Stomee 75b5fca8f8Stomee #ifdef ZNODE_STATS 76b5fca8f8Stomee #define ZNODE_STAT_ADD(stat) ((stat)++) 77b5fca8f8Stomee #else 78b5fca8f8Stomee #define ZNODE_STAT_ADD(stat) /* nothing */ 79b5fca8f8Stomee #endif /* ZNODE_STATS */ 80b5fca8f8Stomee 81b5fca8f8Stomee #define POINTER_IS_VALID(p) (!((uintptr_t)(p) & 0x3)) 82b5fca8f8Stomee #define POINTER_INVALIDATE(pp) (*(pp) = (void *)((uintptr_t)(*(pp)) | 0x1)) 83b5fca8f8Stomee 8455434c77Sek /* 8555434c77Sek * Functions needed for userland (ie: libzpool) are not put under 8655434c77Sek * #ifdef_KERNEL; the rest of the functions have dependencies 8755434c77Sek * (such as VFS logic) that will not compile easily in userland. 8855434c77Sek */ 8955434c77Sek #ifdef _KERNEL 904e9583b2STom Erickson /* 914e9583b2STom Erickson * Needed to close a small window in zfs_znode_move() that allows the zfsvfs to 924e9583b2STom Erickson * be freed before it can be safely accessed. 934e9583b2STom Erickson */ 944e9583b2STom Erickson krwlock_t zfsvfs_lock; 954e9583b2STom Erickson 96b5fca8f8Stomee static kmem_cache_t *znode_cache = NULL; 97fa9e4066Sahrens 98fa9e4066Sahrens /*ARGSUSED*/ 99fa9e4066Sahrens static void 100874395d5Smaybee znode_evict_error(dmu_buf_t *dbuf, void *user_ptr) 101fa9e4066Sahrens { 102874395d5Smaybee /* 103874395d5Smaybee * We should never drop all dbuf refs without first clearing 104874395d5Smaybee * the eviction callback. 105874395d5Smaybee */ 106874395d5Smaybee panic("evicting znode %p\n", user_ptr); 107fa9e4066Sahrens } 108fa9e4066Sahrens 109fa9e4066Sahrens /*ARGSUSED*/ 110fa9e4066Sahrens static int 111b5fca8f8Stomee zfs_znode_cache_constructor(void *buf, void *arg, int kmflags) 112fa9e4066Sahrens { 113fa9e4066Sahrens znode_t *zp = buf; 114fa9e4066Sahrens 115b5fca8f8Stomee ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs)); 116b5fca8f8Stomee 117b5fca8f8Stomee zp->z_vnode = vn_alloc(kmflags); 118b5fca8f8Stomee if (zp->z_vnode == NULL) { 119b5fca8f8Stomee return (-1); 120b5fca8f8Stomee } 121b5fca8f8Stomee ZTOV(zp)->v_data = zp; 122b5fca8f8Stomee 123b5fca8f8Stomee list_link_init(&zp->z_link_node); 124b5fca8f8Stomee 125fa9e4066Sahrens mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL); 126104e2ed7Sperrin rw_init(&zp->z_parent_lock, NULL, RW_DEFAULT, NULL); 127af2c4821Smaybee rw_init(&zp->z_name_lock, NULL, RW_DEFAULT, NULL); 128fa9e4066Sahrens mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL); 129104e2ed7Sperrin 130104e2ed7Sperrin mutex_init(&zp->z_range_lock, NULL, MUTEX_DEFAULT, NULL); 131104e2ed7Sperrin avl_create(&zp->z_range_avl, zfs_range_compare, 132104e2ed7Sperrin sizeof (rl_t), offsetof(rl_t, r_node)); 133104e2ed7Sperrin 1344ccbb6e7Sahrens zp->z_dbuf = NULL; 135b5fca8f8Stomee zp->z_dirlocks = NULL; 136d47621a4STim Haley zp->z_acl_cached = NULL; 137fa9e4066Sahrens return (0); 138fa9e4066Sahrens } 139fa9e4066Sahrens 140fa9e4066Sahrens /*ARGSUSED*/ 141fa9e4066Sahrens static void 142b5fca8f8Stomee zfs_znode_cache_destructor(void *buf, void *arg) 143fa9e4066Sahrens { 144fa9e4066Sahrens znode_t *zp = buf; 145fa9e4066Sahrens 146b5fca8f8Stomee ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs)); 147b5fca8f8Stomee ASSERT(ZTOV(zp)->v_data == zp); 148b5fca8f8Stomee vn_free(ZTOV(zp)); 149b5fca8f8Stomee ASSERT(!list_link_active(&zp->z_link_node)); 150fa9e4066Sahrens mutex_destroy(&zp->z_lock); 151104e2ed7Sperrin rw_destroy(&zp->z_parent_lock); 152af2c4821Smaybee rw_destroy(&zp->z_name_lock); 153fa9e4066Sahrens mutex_destroy(&zp->z_acl_lock); 154104e2ed7Sperrin avl_destroy(&zp->z_range_avl); 155c25056deSgw mutex_destroy(&zp->z_range_lock); 156fa9e4066Sahrens 1574ccbb6e7Sahrens ASSERT(zp->z_dbuf == NULL); 158b5fca8f8Stomee ASSERT(zp->z_dirlocks == NULL); 159*4929fd5eSTim Haley ASSERT(zp->z_acl_cached == NULL); 160b5fca8f8Stomee } 161b5fca8f8Stomee 162b5fca8f8Stomee #ifdef ZNODE_STATS 163b5fca8f8Stomee static struct { 164b5fca8f8Stomee uint64_t zms_zfsvfs_invalid; 1654e9583b2STom Erickson uint64_t zms_zfsvfs_recheck1; 166b5fca8f8Stomee uint64_t zms_zfsvfs_unmounted; 1674e9583b2STom Erickson uint64_t zms_zfsvfs_recheck2; 168a66b2b35STom Erickson uint64_t zms_obj_held; 169b5fca8f8Stomee uint64_t zms_vnode_locked; 170a66b2b35STom Erickson uint64_t zms_not_only_dnlc; 171b5fca8f8Stomee } znode_move_stats; 172b5fca8f8Stomee #endif /* ZNODE_STATS */ 173b5fca8f8Stomee 174b5fca8f8Stomee static void 175b5fca8f8Stomee zfs_znode_move_impl(znode_t *ozp, znode_t *nzp) 176b5fca8f8Stomee { 177b5fca8f8Stomee vnode_t *vp; 178b5fca8f8Stomee 179b5fca8f8Stomee /* Copy fields. */ 180b5fca8f8Stomee nzp->z_zfsvfs = ozp->z_zfsvfs; 181b5fca8f8Stomee 182b5fca8f8Stomee /* Swap vnodes. */ 183b5fca8f8Stomee vp = nzp->z_vnode; 184b5fca8f8Stomee nzp->z_vnode = ozp->z_vnode; 185b5fca8f8Stomee ozp->z_vnode = vp; /* let destructor free the overwritten vnode */ 186b5fca8f8Stomee ZTOV(ozp)->v_data = ozp; 187b5fca8f8Stomee ZTOV(nzp)->v_data = nzp; 188b5fca8f8Stomee 189b5fca8f8Stomee nzp->z_id = ozp->z_id; 190b5fca8f8Stomee ASSERT(ozp->z_dirlocks == NULL); /* znode not in use */ 191b5fca8f8Stomee ASSERT(avl_numnodes(&ozp->z_range_avl) == 0); 192b5fca8f8Stomee nzp->z_unlinked = ozp->z_unlinked; 193b5fca8f8Stomee nzp->z_atime_dirty = ozp->z_atime_dirty; 194b5fca8f8Stomee nzp->z_zn_prefetch = ozp->z_zn_prefetch; 195b5fca8f8Stomee nzp->z_blksz = ozp->z_blksz; 196b5fca8f8Stomee nzp->z_seq = ozp->z_seq; 197b5fca8f8Stomee nzp->z_mapcnt = ozp->z_mapcnt; 198b5fca8f8Stomee nzp->z_last_itx = ozp->z_last_itx; 199b5fca8f8Stomee nzp->z_gen = ozp->z_gen; 200b5fca8f8Stomee nzp->z_sync_cnt = ozp->z_sync_cnt; 201b5fca8f8Stomee nzp->z_phys = ozp->z_phys; 202b5fca8f8Stomee nzp->z_dbuf = ozp->z_dbuf; 203*4929fd5eSTim Haley nzp->z_acl_cached = ozp->z_acl_cached; 204b5fca8f8Stomee 205b5fca8f8Stomee /* Update back pointers. */ 206b5fca8f8Stomee (void) dmu_buf_update_user(nzp->z_dbuf, ozp, nzp, &nzp->z_phys, 207b5fca8f8Stomee znode_evict_error); 208b5fca8f8Stomee 209b5fca8f8Stomee /* 210b5fca8f8Stomee * Invalidate the original znode by clearing fields that provide a 211b5fca8f8Stomee * pointer back to the znode. Set the low bit of the vfs pointer to 212b5fca8f8Stomee * ensure that zfs_znode_move() recognizes the znode as invalid in any 213b5fca8f8Stomee * subsequent callback. 214b5fca8f8Stomee */ 215b5fca8f8Stomee ozp->z_dbuf = NULL; 216*4929fd5eSTim Haley ozp->z_acl_cached = NULL; 217b5fca8f8Stomee POINTER_INVALIDATE(&ozp->z_zfsvfs); 218b5fca8f8Stomee } 219b5fca8f8Stomee 220b5fca8f8Stomee /*ARGSUSED*/ 221b5fca8f8Stomee static kmem_cbrc_t 222b5fca8f8Stomee zfs_znode_move(void *buf, void *newbuf, size_t size, void *arg) 223b5fca8f8Stomee { 224b5fca8f8Stomee znode_t *ozp = buf, *nzp = newbuf; 225b5fca8f8Stomee zfsvfs_t *zfsvfs; 226b5fca8f8Stomee vnode_t *vp; 227b5fca8f8Stomee 228b5fca8f8Stomee /* 229b5fca8f8Stomee * The znode is on the file system's list of known znodes if the vfs 230b5fca8f8Stomee * pointer is valid. We set the low bit of the vfs pointer when freeing 231b5fca8f8Stomee * the znode to invalidate it, and the memory patterns written by kmem 232b5fca8f8Stomee * (baddcafe and deadbeef) set at least one of the two low bits. A newly 233b5fca8f8Stomee * created znode sets the vfs pointer last of all to indicate that the 234b5fca8f8Stomee * znode is known and in a valid state to be moved by this function. 235b5fca8f8Stomee */ 236b5fca8f8Stomee zfsvfs = ozp->z_zfsvfs; 237b5fca8f8Stomee if (!POINTER_IS_VALID(zfsvfs)) { 238b5fca8f8Stomee ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_invalid); 239b5fca8f8Stomee return (KMEM_CBRC_DONT_KNOW); 240b5fca8f8Stomee } 241b5fca8f8Stomee 242b5fca8f8Stomee /* 2434e9583b2STom Erickson * Close a small window in which it's possible that the filesystem could 2444e9583b2STom Erickson * be unmounted and freed, and zfsvfs, though valid in the previous 2454e9583b2STom Erickson * statement, could point to unrelated memory by the time we try to 2464e9583b2STom Erickson * prevent the filesystem from being unmounted. 2474e9583b2STom Erickson */ 2484e9583b2STom Erickson rw_enter(&zfsvfs_lock, RW_WRITER); 2494e9583b2STom Erickson if (zfsvfs != ozp->z_zfsvfs) { 2504e9583b2STom Erickson rw_exit(&zfsvfs_lock); 2514e9583b2STom Erickson ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_recheck1); 2524e9583b2STom Erickson return (KMEM_CBRC_DONT_KNOW); 2534e9583b2STom Erickson } 2544e9583b2STom Erickson 2554e9583b2STom Erickson /* 2564e9583b2STom Erickson * If the znode is still valid, then so is the file system. We know that 2574e9583b2STom Erickson * no valid file system can be freed while we hold zfsvfs_lock, so we 2584e9583b2STom Erickson * can safely ensure that the filesystem is not and will not be 2594e9583b2STom Erickson * unmounted. The next statement is equivalent to ZFS_ENTER(). 260b5fca8f8Stomee */ 26114843421SMatthew Ahrens rrw_enter(&zfsvfs->z_teardown_lock, RW_READER, FTAG); 26214843421SMatthew Ahrens if (zfsvfs->z_unmounted) { 26314843421SMatthew Ahrens ZFS_EXIT(zfsvfs); 2644e9583b2STom Erickson rw_exit(&zfsvfs_lock); 265b5fca8f8Stomee ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_unmounted); 266b5fca8f8Stomee return (KMEM_CBRC_DONT_KNOW); 267b5fca8f8Stomee } 2684e9583b2STom Erickson rw_exit(&zfsvfs_lock); 269b5fca8f8Stomee 270b5fca8f8Stomee mutex_enter(&zfsvfs->z_znodes_lock); 271b5fca8f8Stomee /* 272b5fca8f8Stomee * Recheck the vfs pointer in case the znode was removed just before 273b5fca8f8Stomee * acquiring the lock. 274b5fca8f8Stomee */ 275b5fca8f8Stomee if (zfsvfs != ozp->z_zfsvfs) { 276b5fca8f8Stomee mutex_exit(&zfsvfs->z_znodes_lock); 277b5fca8f8Stomee ZFS_EXIT(zfsvfs); 2784e9583b2STom Erickson ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_recheck2); 279b5fca8f8Stomee return (KMEM_CBRC_DONT_KNOW); 280b5fca8f8Stomee } 281b5fca8f8Stomee 282b5fca8f8Stomee /* 283b5fca8f8Stomee * At this point we know that as long as we hold z_znodes_lock, the 284b5fca8f8Stomee * znode cannot be freed and fields within the znode can be safely 285a66b2b35STom Erickson * accessed. Now, prevent a race with zfs_zget(). 286b5fca8f8Stomee */ 287a66b2b35STom Erickson if (ZFS_OBJ_HOLD_TRYENTER(zfsvfs, ozp->z_id) == 0) { 288a66b2b35STom Erickson mutex_exit(&zfsvfs->z_znodes_lock); 289a66b2b35STom Erickson ZFS_EXIT(zfsvfs); 290a66b2b35STom Erickson ZNODE_STAT_ADD(znode_move_stats.zms_obj_held); 291a66b2b35STom Erickson return (KMEM_CBRC_LATER); 292a66b2b35STom Erickson } 293a66b2b35STom Erickson 294b5fca8f8Stomee vp = ZTOV(ozp); 295b5fca8f8Stomee if (mutex_tryenter(&vp->v_lock) == 0) { 296a66b2b35STom Erickson ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id); 297b5fca8f8Stomee mutex_exit(&zfsvfs->z_znodes_lock); 298b5fca8f8Stomee ZFS_EXIT(zfsvfs); 299b5fca8f8Stomee ZNODE_STAT_ADD(znode_move_stats.zms_vnode_locked); 300b5fca8f8Stomee return (KMEM_CBRC_LATER); 301b5fca8f8Stomee } 302a66b2b35STom Erickson 303b5fca8f8Stomee /* Only move znodes that are referenced _only_ by the DNLC. */ 304b5fca8f8Stomee if (vp->v_count != 1 || !vn_in_dnlc(vp)) { 305b5fca8f8Stomee mutex_exit(&vp->v_lock); 306a66b2b35STom Erickson ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id); 307b5fca8f8Stomee mutex_exit(&zfsvfs->z_znodes_lock); 308b5fca8f8Stomee ZFS_EXIT(zfsvfs); 309a66b2b35STom Erickson ZNODE_STAT_ADD(znode_move_stats.zms_not_only_dnlc); 310b5fca8f8Stomee return (KMEM_CBRC_LATER); 311b5fca8f8Stomee } 312b5fca8f8Stomee 313b5fca8f8Stomee /* 314b5fca8f8Stomee * The znode is known and in a valid state to move. We're holding the 315b5fca8f8Stomee * locks needed to execute the critical section. 316b5fca8f8Stomee */ 317b5fca8f8Stomee zfs_znode_move_impl(ozp, nzp); 318b5fca8f8Stomee mutex_exit(&vp->v_lock); 319a66b2b35STom Erickson ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id); 320b5fca8f8Stomee 321b5fca8f8Stomee list_link_replace(&ozp->z_link_node, &nzp->z_link_node); 322b5fca8f8Stomee mutex_exit(&zfsvfs->z_znodes_lock); 323b5fca8f8Stomee ZFS_EXIT(zfsvfs); 324b5fca8f8Stomee 325b5fca8f8Stomee return (KMEM_CBRC_YES); 326fa9e4066Sahrens } 327fa9e4066Sahrens 328fa9e4066Sahrens void 329fa9e4066Sahrens zfs_znode_init(void) 330fa9e4066Sahrens { 331fa9e4066Sahrens /* 332fa9e4066Sahrens * Initialize zcache 333fa9e4066Sahrens */ 3344e9583b2STom Erickson rw_init(&zfsvfs_lock, NULL, RW_DEFAULT, NULL); 335fa9e4066Sahrens ASSERT(znode_cache == NULL); 336fa9e4066Sahrens znode_cache = kmem_cache_create("zfs_znode_cache", 337fa9e4066Sahrens sizeof (znode_t), 0, zfs_znode_cache_constructor, 338fa9e4066Sahrens zfs_znode_cache_destructor, NULL, NULL, NULL, 0); 339b5fca8f8Stomee kmem_cache_set_move(znode_cache, zfs_znode_move); 340fa9e4066Sahrens } 341fa9e4066Sahrens 342fa9e4066Sahrens void 343fa9e4066Sahrens zfs_znode_fini(void) 344fa9e4066Sahrens { 345fa9e4066Sahrens /* 346fa9e4066Sahrens * Cleanup vfs & vnode ops 347fa9e4066Sahrens */ 348fa9e4066Sahrens zfs_remove_op_tables(); 349fa9e4066Sahrens 350fa9e4066Sahrens /* 351fa9e4066Sahrens * Cleanup zcache 352fa9e4066Sahrens */ 353fa9e4066Sahrens if (znode_cache) 354fa9e4066Sahrens kmem_cache_destroy(znode_cache); 355fa9e4066Sahrens znode_cache = NULL; 3564e9583b2STom Erickson rw_destroy(&zfsvfs_lock); 357fa9e4066Sahrens } 358fa9e4066Sahrens 359fa9e4066Sahrens struct vnodeops *zfs_dvnodeops; 360fa9e4066Sahrens struct vnodeops *zfs_fvnodeops; 361fa9e4066Sahrens struct vnodeops *zfs_symvnodeops; 362fa9e4066Sahrens struct vnodeops *zfs_xdvnodeops; 363fa9e4066Sahrens struct vnodeops *zfs_evnodeops; 364743a77edSAlan Wright struct vnodeops *zfs_sharevnodeops; 365fa9e4066Sahrens 366fa9e4066Sahrens void 367fa9e4066Sahrens zfs_remove_op_tables() 368fa9e4066Sahrens { 369fa9e4066Sahrens /* 370fa9e4066Sahrens * Remove vfs ops 371fa9e4066Sahrens */ 372fa9e4066Sahrens ASSERT(zfsfstype); 373fa9e4066Sahrens (void) vfs_freevfsops_by_type(zfsfstype); 374fa9e4066Sahrens zfsfstype = 0; 375fa9e4066Sahrens 376fa9e4066Sahrens /* 377fa9e4066Sahrens * Remove vnode ops 378fa9e4066Sahrens */ 379fa9e4066Sahrens if (zfs_dvnodeops) 380fa9e4066Sahrens vn_freevnodeops(zfs_dvnodeops); 381fa9e4066Sahrens if (zfs_fvnodeops) 382fa9e4066Sahrens vn_freevnodeops(zfs_fvnodeops); 383fa9e4066Sahrens if (zfs_symvnodeops) 384fa9e4066Sahrens vn_freevnodeops(zfs_symvnodeops); 385fa9e4066Sahrens if (zfs_xdvnodeops) 386fa9e4066Sahrens vn_freevnodeops(zfs_xdvnodeops); 387fa9e4066Sahrens if (zfs_evnodeops) 388fa9e4066Sahrens vn_freevnodeops(zfs_evnodeops); 389743a77edSAlan Wright if (zfs_sharevnodeops) 390743a77edSAlan Wright vn_freevnodeops(zfs_sharevnodeops); 391fa9e4066Sahrens 392fa9e4066Sahrens zfs_dvnodeops = NULL; 393fa9e4066Sahrens zfs_fvnodeops = NULL; 394fa9e4066Sahrens zfs_symvnodeops = NULL; 395fa9e4066Sahrens zfs_xdvnodeops = NULL; 396fa9e4066Sahrens zfs_evnodeops = NULL; 397743a77edSAlan Wright zfs_sharevnodeops = NULL; 398fa9e4066Sahrens } 399fa9e4066Sahrens 400fa9e4066Sahrens extern const fs_operation_def_t zfs_dvnodeops_template[]; 401fa9e4066Sahrens extern const fs_operation_def_t zfs_fvnodeops_template[]; 402fa9e4066Sahrens extern const fs_operation_def_t zfs_xdvnodeops_template[]; 403fa9e4066Sahrens extern const fs_operation_def_t zfs_symvnodeops_template[]; 404fa9e4066Sahrens extern const fs_operation_def_t zfs_evnodeops_template[]; 405743a77edSAlan Wright extern const fs_operation_def_t zfs_sharevnodeops_template[]; 406fa9e4066Sahrens 407fa9e4066Sahrens int 408fa9e4066Sahrens zfs_create_op_tables() 409fa9e4066Sahrens { 410fa9e4066Sahrens int error; 411fa9e4066Sahrens 412fa9e4066Sahrens /* 413fa9e4066Sahrens * zfs_dvnodeops can be set if mod_remove() calls mod_installfs() 414fa9e4066Sahrens * due to a failure to remove the the 2nd modlinkage (zfs_modldrv). 415fa9e4066Sahrens * In this case we just return as the ops vectors are already set up. 416fa9e4066Sahrens */ 417fa9e4066Sahrens if (zfs_dvnodeops) 418fa9e4066Sahrens return (0); 419fa9e4066Sahrens 420fa9e4066Sahrens error = vn_make_ops(MNTTYPE_ZFS, zfs_dvnodeops_template, 421fa9e4066Sahrens &zfs_dvnodeops); 422fa9e4066Sahrens if (error) 423fa9e4066Sahrens return (error); 424fa9e4066Sahrens 425fa9e4066Sahrens error = vn_make_ops(MNTTYPE_ZFS, zfs_fvnodeops_template, 426fa9e4066Sahrens &zfs_fvnodeops); 427fa9e4066Sahrens if (error) 428fa9e4066Sahrens return (error); 429fa9e4066Sahrens 430fa9e4066Sahrens error = vn_make_ops(MNTTYPE_ZFS, zfs_symvnodeops_template, 431fa9e4066Sahrens &zfs_symvnodeops); 432fa9e4066Sahrens if (error) 433fa9e4066Sahrens return (error); 434fa9e4066Sahrens 435fa9e4066Sahrens error = vn_make_ops(MNTTYPE_ZFS, zfs_xdvnodeops_template, 436fa9e4066Sahrens &zfs_xdvnodeops); 437fa9e4066Sahrens if (error) 438fa9e4066Sahrens return (error); 439fa9e4066Sahrens 440fa9e4066Sahrens error = vn_make_ops(MNTTYPE_ZFS, zfs_evnodeops_template, 441fa9e4066Sahrens &zfs_evnodeops); 442743a77edSAlan Wright if (error) 443743a77edSAlan Wright return (error); 444743a77edSAlan Wright 445743a77edSAlan Wright error = vn_make_ops(MNTTYPE_ZFS, zfs_sharevnodeops_template, 446743a77edSAlan Wright &zfs_sharevnodeops); 447743a77edSAlan Wright 448743a77edSAlan Wright return (error); 449743a77edSAlan Wright } 450743a77edSAlan Wright 4519e1320c0SMark Shellenbaum int 452743a77edSAlan Wright zfs_create_share_dir(zfsvfs_t *zfsvfs, dmu_tx_t *tx) 453743a77edSAlan Wright { 45489459e17SMark Shellenbaum zfs_acl_ids_t acl_ids; 455743a77edSAlan Wright vattr_t vattr; 456743a77edSAlan Wright znode_t *sharezp; 457743a77edSAlan Wright vnode_t *vp; 458743a77edSAlan Wright znode_t *zp; 459743a77edSAlan Wright int error; 460743a77edSAlan Wright 461743a77edSAlan Wright vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE; 462743a77edSAlan Wright vattr.va_type = VDIR; 463743a77edSAlan Wright vattr.va_mode = S_IFDIR|0555; 464743a77edSAlan Wright vattr.va_uid = crgetuid(kcred); 465743a77edSAlan Wright vattr.va_gid = crgetgid(kcred); 466743a77edSAlan Wright 467743a77edSAlan Wright sharezp = kmem_cache_alloc(znode_cache, KM_SLEEP); 468743a77edSAlan Wright sharezp->z_unlinked = 0; 469743a77edSAlan Wright sharezp->z_atime_dirty = 0; 470743a77edSAlan Wright sharezp->z_zfsvfs = zfsvfs; 471743a77edSAlan Wright 472743a77edSAlan Wright vp = ZTOV(sharezp); 473743a77edSAlan Wright vn_reinit(vp); 474743a77edSAlan Wright vp->v_type = VDIR; 475743a77edSAlan Wright 47689459e17SMark Shellenbaum VERIFY(0 == zfs_acl_ids_create(sharezp, IS_ROOT_NODE, &vattr, 47789459e17SMark Shellenbaum kcred, NULL, &acl_ids)); 478743a77edSAlan Wright zfs_mknode(sharezp, &vattr, tx, kcred, IS_ROOT_NODE, 47989459e17SMark Shellenbaum &zp, 0, &acl_ids); 480743a77edSAlan Wright ASSERT3P(zp, ==, sharezp); 481743a77edSAlan Wright ASSERT(!vn_in_dnlc(ZTOV(sharezp))); /* not valid to move */ 482743a77edSAlan Wright POINTER_INVALIDATE(&sharezp->z_zfsvfs); 483743a77edSAlan Wright error = zap_add(zfsvfs->z_os, MASTER_NODE_OBJ, 484743a77edSAlan Wright ZFS_SHARES_DIR, 8, 1, &sharezp->z_id, tx); 485743a77edSAlan Wright zfsvfs->z_shares_dir = sharezp->z_id; 486743a77edSAlan Wright 48789459e17SMark Shellenbaum zfs_acl_ids_free(&acl_ids); 488743a77edSAlan Wright ZTOV(sharezp)->v_count = 0; 489743a77edSAlan Wright dmu_buf_rele(sharezp->z_dbuf, NULL); 490743a77edSAlan Wright sharezp->z_dbuf = NULL; 491743a77edSAlan Wright kmem_cache_free(znode_cache, sharezp); 492fa9e4066Sahrens 493fa9e4066Sahrens return (error); 494fa9e4066Sahrens } 495fa9e4066Sahrens 49672fc53bcSmarks /* 49772fc53bcSmarks * define a couple of values we need available 49872fc53bcSmarks * for both 64 and 32 bit environments. 49972fc53bcSmarks */ 50072fc53bcSmarks #ifndef NBITSMINOR64 50172fc53bcSmarks #define NBITSMINOR64 32 50272fc53bcSmarks #endif 50372fc53bcSmarks #ifndef MAXMAJ64 50472fc53bcSmarks #define MAXMAJ64 0xffffffffUL 50572fc53bcSmarks #endif 50672fc53bcSmarks #ifndef MAXMIN64 50772fc53bcSmarks #define MAXMIN64 0xffffffffUL 50872fc53bcSmarks #endif 50972fc53bcSmarks 51072fc53bcSmarks /* 51172fc53bcSmarks * Create special expldev for ZFS private use. 51272fc53bcSmarks * Can't use standard expldev since it doesn't do 51372fc53bcSmarks * what we want. The standard expldev() takes a 51472fc53bcSmarks * dev32_t in LP64 and expands it to a long dev_t. 51572fc53bcSmarks * We need an interface that takes a dev32_t in ILP32 51672fc53bcSmarks * and expands it to a long dev_t. 51772fc53bcSmarks */ 51872fc53bcSmarks static uint64_t 51972fc53bcSmarks zfs_expldev(dev_t dev) 52072fc53bcSmarks { 52172fc53bcSmarks #ifndef _LP64 52272fc53bcSmarks major_t major = (major_t)dev >> NBITSMINOR32 & MAXMAJ32; 52372fc53bcSmarks return (((uint64_t)major << NBITSMINOR64) | 52472fc53bcSmarks ((minor_t)dev & MAXMIN32)); 52572fc53bcSmarks #else 52672fc53bcSmarks return (dev); 52772fc53bcSmarks #endif 52872fc53bcSmarks } 52972fc53bcSmarks 53072fc53bcSmarks /* 53172fc53bcSmarks * Special cmpldev for ZFS private use. 53272fc53bcSmarks * Can't use standard cmpldev since it takes 53372fc53bcSmarks * a long dev_t and compresses it to dev32_t in 53472fc53bcSmarks * LP64. We need to do a compaction of a long dev_t 53572fc53bcSmarks * to a dev32_t in ILP32. 53672fc53bcSmarks */ 53772fc53bcSmarks dev_t 53872fc53bcSmarks zfs_cmpldev(uint64_t dev) 53972fc53bcSmarks { 54072fc53bcSmarks #ifndef _LP64 54172fc53bcSmarks minor_t minor = (minor_t)dev & MAXMIN64; 54272fc53bcSmarks major_t major = (major_t)(dev >> NBITSMINOR64) & MAXMAJ64; 54372fc53bcSmarks 54472fc53bcSmarks if (major > MAXMAJ32 || minor > MAXMIN32) 54572fc53bcSmarks return (NODEV32); 54672fc53bcSmarks 54772fc53bcSmarks return (((dev32_t)major << NBITSMINOR32) | minor); 54872fc53bcSmarks #else 54972fc53bcSmarks return (dev); 55072fc53bcSmarks #endif 55172fc53bcSmarks } 55272fc53bcSmarks 5534ccbb6e7Sahrens static void 554b5fca8f8Stomee zfs_znode_dmu_init(zfsvfs_t *zfsvfs, znode_t *zp, dmu_buf_t *db) 5554ccbb6e7Sahrens { 5564ccbb6e7Sahrens znode_t *nzp; 5574ccbb6e7Sahrens 558b5fca8f8Stomee ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs) || (zfsvfs == zp->z_zfsvfs)); 559b5fca8f8Stomee ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id))); 5604ccbb6e7Sahrens 5614ccbb6e7Sahrens mutex_enter(&zp->z_lock); 5624ccbb6e7Sahrens 5634ccbb6e7Sahrens ASSERT(zp->z_dbuf == NULL); 5644ccbb6e7Sahrens zp->z_dbuf = db; 565874395d5Smaybee nzp = dmu_buf_set_user_ie(db, zp, &zp->z_phys, znode_evict_error); 5664ccbb6e7Sahrens 5674ccbb6e7Sahrens /* 5684ccbb6e7Sahrens * there should be no 5694ccbb6e7Sahrens * concurrent zgets on this object. 5704ccbb6e7Sahrens */ 5714ccbb6e7Sahrens if (nzp != NULL) 572903a11ebSrh panic("existing znode %p for dbuf %p", (void *)nzp, (void *)db); 5734ccbb6e7Sahrens 5744ccbb6e7Sahrens /* 5754ccbb6e7Sahrens * Slap on VROOT if we are the root znode 5764ccbb6e7Sahrens */ 5774ccbb6e7Sahrens if (zp->z_id == zfsvfs->z_root) 5784ccbb6e7Sahrens ZTOV(zp)->v_flag |= VROOT; 5794ccbb6e7Sahrens 5804ccbb6e7Sahrens mutex_exit(&zp->z_lock); 5814ccbb6e7Sahrens vn_exists(ZTOV(zp)); 5824ccbb6e7Sahrens } 5834ccbb6e7Sahrens 584874395d5Smaybee void 5854ccbb6e7Sahrens zfs_znode_dmu_fini(znode_t *zp) 5864ccbb6e7Sahrens { 5874ccbb6e7Sahrens dmu_buf_t *db = zp->z_dbuf; 588b5fca8f8Stomee ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zp->z_zfsvfs, zp->z_id)) || 589b5fca8f8Stomee zp->z_unlinked || 590874395d5Smaybee RW_WRITE_HELD(&zp->z_zfsvfs->z_teardown_inactive_lock)); 5914ccbb6e7Sahrens ASSERT(zp->z_dbuf != NULL); 5924ccbb6e7Sahrens zp->z_dbuf = NULL; 593874395d5Smaybee VERIFY(zp == dmu_buf_update_user(db, zp, NULL, NULL, NULL)); 5944ccbb6e7Sahrens dmu_buf_rele(db, NULL); 5954ccbb6e7Sahrens } 5964ccbb6e7Sahrens 597fa9e4066Sahrens /* 598fa9e4066Sahrens * Construct a new znode/vnode and intialize. 599fa9e4066Sahrens * 600fa9e4066Sahrens * This does not do a call to dmu_set_user() that is 601fa9e4066Sahrens * up to the caller to do, in case you don't want to 602fa9e4066Sahrens * return the znode 603fa9e4066Sahrens */ 604ea8dc4b6Seschrock static znode_t * 6054ccbb6e7Sahrens zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz) 606fa9e4066Sahrens { 607fa9e4066Sahrens znode_t *zp; 608fa9e4066Sahrens vnode_t *vp; 609fa9e4066Sahrens 610fa9e4066Sahrens zp = kmem_cache_alloc(znode_cache, KM_SLEEP); 611fa9e4066Sahrens 612fa9e4066Sahrens ASSERT(zp->z_dirlocks == NULL); 6134ccbb6e7Sahrens ASSERT(zp->z_dbuf == NULL); 614b5fca8f8Stomee ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs)); 615fa9e4066Sahrens 616b5fca8f8Stomee /* 617b5fca8f8Stomee * Defer setting z_zfsvfs until the znode is ready to be a candidate for 618b5fca8f8Stomee * the zfs_znode_move() callback. 619b5fca8f8Stomee */ 6204ccbb6e7Sahrens zp->z_phys = NULL; 621893a6d32Sahrens zp->z_unlinked = 0; 622fa9e4066Sahrens zp->z_atime_dirty = 0; 623fa9e4066Sahrens zp->z_mapcnt = 0; 624fa9e4066Sahrens zp->z_last_itx = 0; 6254ccbb6e7Sahrens zp->z_id = db->db_object; 626fa9e4066Sahrens zp->z_blksz = blksz; 627fa9e4066Sahrens zp->z_seq = 0x7A4653; 62867bd71c6Sperrin zp->z_sync_cnt = 0; 6294ccbb6e7Sahrens 6304ccbb6e7Sahrens vp = ZTOV(zp); 6314ccbb6e7Sahrens vn_reinit(vp); 6324ccbb6e7Sahrens 633b5fca8f8Stomee zfs_znode_dmu_init(zfsvfs, zp, db); 6344ccbb6e7Sahrens 635f18faf3fSek zp->z_gen = zp->z_phys->zp_gen; 636fa9e4066Sahrens 637fa9e4066Sahrens vp->v_vfsp = zfsvfs->z_parent->z_vfs; 638fa9e4066Sahrens vp->v_type = IFTOVT((mode_t)zp->z_phys->zp_mode); 639fa9e4066Sahrens 640fa9e4066Sahrens switch (vp->v_type) { 641fa9e4066Sahrens case VDIR: 642fa9e4066Sahrens if (zp->z_phys->zp_flags & ZFS_XATTR) { 643fa9e4066Sahrens vn_setops(vp, zfs_xdvnodeops); 644fa9e4066Sahrens vp->v_flag |= V_XATTRDIR; 6454ccbb6e7Sahrens } else { 646fa9e4066Sahrens vn_setops(vp, zfs_dvnodeops); 6474ccbb6e7Sahrens } 6487f6e3e7dSperrin zp->z_zn_prefetch = B_TRUE; /* z_prefetch default is enabled */ 649fa9e4066Sahrens break; 650fa9e4066Sahrens case VBLK: 651fa9e4066Sahrens case VCHR: 65272fc53bcSmarks vp->v_rdev = zfs_cmpldev(zp->z_phys->zp_rdev); 653fa9e4066Sahrens /*FALLTHROUGH*/ 654fa9e4066Sahrens case VFIFO: 655fa9e4066Sahrens case VSOCK: 656fa9e4066Sahrens case VDOOR: 657fa9e4066Sahrens vn_setops(vp, zfs_fvnodeops); 658fa9e4066Sahrens break; 659fa9e4066Sahrens case VREG: 660fa9e4066Sahrens vp->v_flag |= VMODSORT; 661743a77edSAlan Wright if (zp->z_phys->zp_parent == zfsvfs->z_shares_dir) 662743a77edSAlan Wright vn_setops(vp, zfs_sharevnodeops); 663743a77edSAlan Wright else 664743a77edSAlan Wright vn_setops(vp, zfs_fvnodeops); 665fa9e4066Sahrens break; 666fa9e4066Sahrens case VLNK: 667fa9e4066Sahrens vn_setops(vp, zfs_symvnodeops); 668fa9e4066Sahrens break; 669fa9e4066Sahrens default: 670fa9e4066Sahrens vn_setops(vp, zfs_evnodeops); 671fa9e4066Sahrens break; 672fa9e4066Sahrens } 673fa9e4066Sahrens 674b5fca8f8Stomee mutex_enter(&zfsvfs->z_znodes_lock); 675b5fca8f8Stomee list_insert_tail(&zfsvfs->z_all_znodes, zp); 676b5fca8f8Stomee membar_producer(); 677b5fca8f8Stomee /* 678b5fca8f8Stomee * Everything else must be valid before assigning z_zfsvfs makes the 679b5fca8f8Stomee * znode eligible for zfs_znode_move(). 680b5fca8f8Stomee */ 681b5fca8f8Stomee zp->z_zfsvfs = zfsvfs; 682b5fca8f8Stomee mutex_exit(&zfsvfs->z_znodes_lock); 683b5fca8f8Stomee 684874395d5Smaybee VFS_HOLD(zfsvfs->z_vfs); 685fa9e4066Sahrens return (zp); 686fa9e4066Sahrens } 687fa9e4066Sahrens 688fa9e4066Sahrens /* 689fa9e4066Sahrens * Create a new DMU object to hold a zfs znode. 690fa9e4066Sahrens * 691fa9e4066Sahrens * IN: dzp - parent directory for new znode 692fa9e4066Sahrens * vap - file attributes for new znode 693fa9e4066Sahrens * tx - dmu transaction id for zap operations 694fa9e4066Sahrens * cr - credentials of caller 695fa9e4066Sahrens * flag - flags: 696fa9e4066Sahrens * IS_ROOT_NODE - new object will be root 697fa9e4066Sahrens * IS_XATTR - new object is an attribute 698fa9e4066Sahrens * IS_REPLAY - intent log replay 699da6c28aaSamw * bonuslen - length of bonus buffer 700da6c28aaSamw * setaclp - File/Dir initial ACL 701da6c28aaSamw * fuidp - Tracks fuid allocation. 702fa9e4066Sahrens * 7034ccbb6e7Sahrens * OUT: zpp - allocated znode 704fa9e4066Sahrens * 705fa9e4066Sahrens */ 706fa9e4066Sahrens void 7074ccbb6e7Sahrens zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr, 70889459e17SMark Shellenbaum uint_t flag, znode_t **zpp, int bonuslen, zfs_acl_ids_t *acl_ids) 709fa9e4066Sahrens { 7104ccbb6e7Sahrens dmu_buf_t *db; 711fa9e4066Sahrens znode_phys_t *pzp; 712fa9e4066Sahrens zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 713fa9e4066Sahrens timestruc_t now; 7144ccbb6e7Sahrens uint64_t gen, obj; 715fa9e4066Sahrens int err; 716fa9e4066Sahrens 717fa9e4066Sahrens ASSERT(vap && (vap->va_mask & (AT_TYPE|AT_MODE)) == (AT_TYPE|AT_MODE)); 718fa9e4066Sahrens 7191209a471SNeil Perrin if (zfsvfs->z_replay) { 7204ccbb6e7Sahrens obj = vap->va_nodeid; 721fa9e4066Sahrens flag |= IS_REPLAY; 722fa9e4066Sahrens now = vap->va_ctime; /* see zfs_replay_create() */ 723fa9e4066Sahrens gen = vap->va_nblocks; /* ditto */ 724fa9e4066Sahrens } else { 7254ccbb6e7Sahrens obj = 0; 726fa9e4066Sahrens gethrestime(&now); 727fa9e4066Sahrens gen = dmu_tx_get_txg(tx); 728fa9e4066Sahrens } 729fa9e4066Sahrens 730fa9e4066Sahrens /* 731fa9e4066Sahrens * Create a new DMU object. 732fa9e4066Sahrens */ 733ea8dc4b6Seschrock /* 734ea8dc4b6Seschrock * There's currently no mechanism for pre-reading the blocks that will 735ea8dc4b6Seschrock * be to needed allocate a new object, so we accept the small chance 736ea8dc4b6Seschrock * that there will be an i/o error and we will fail one of the 737ea8dc4b6Seschrock * assertions below. 738ea8dc4b6Seschrock */ 739fa9e4066Sahrens if (vap->va_type == VDIR) { 740fa9e4066Sahrens if (flag & IS_REPLAY) { 7414ccbb6e7Sahrens err = zap_create_claim_norm(zfsvfs->z_os, obj, 742da6c28aaSamw zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS, 743fa9e4066Sahrens DMU_OT_ZNODE, sizeof (znode_phys_t) + bonuslen, tx); 744fa9e4066Sahrens ASSERT3U(err, ==, 0); 745fa9e4066Sahrens } else { 7464ccbb6e7Sahrens obj = zap_create_norm(zfsvfs->z_os, 747da6c28aaSamw zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS, 748fa9e4066Sahrens DMU_OT_ZNODE, sizeof (znode_phys_t) + bonuslen, tx); 749fa9e4066Sahrens } 750fa9e4066Sahrens } else { 751fa9e4066Sahrens if (flag & IS_REPLAY) { 7524ccbb6e7Sahrens err = dmu_object_claim(zfsvfs->z_os, obj, 753fa9e4066Sahrens DMU_OT_PLAIN_FILE_CONTENTS, 0, 754fa9e4066Sahrens DMU_OT_ZNODE, sizeof (znode_phys_t) + bonuslen, tx); 755fa9e4066Sahrens ASSERT3U(err, ==, 0); 756fa9e4066Sahrens } else { 7574ccbb6e7Sahrens obj = dmu_object_alloc(zfsvfs->z_os, 758fa9e4066Sahrens DMU_OT_PLAIN_FILE_CONTENTS, 0, 759fa9e4066Sahrens DMU_OT_ZNODE, sizeof (znode_phys_t) + bonuslen, tx); 760fa9e4066Sahrens } 761fa9e4066Sahrens } 7624ccbb6e7Sahrens VERIFY(0 == dmu_bonus_hold(zfsvfs->z_os, obj, NULL, &db)); 7634ccbb6e7Sahrens dmu_buf_will_dirty(db, tx); 764fa9e4066Sahrens 765fa9e4066Sahrens /* 766fa9e4066Sahrens * Initialize the znode physical data to zero. 767fa9e4066Sahrens */ 7684ccbb6e7Sahrens ASSERT(db->db_size >= sizeof (znode_phys_t)); 7694ccbb6e7Sahrens bzero(db->db_data, db->db_size); 7704ccbb6e7Sahrens pzp = db->db_data; 771fa9e4066Sahrens 772fa9e4066Sahrens /* 773fa9e4066Sahrens * If this is the root, fix up the half-initialized parent pointer 774fa9e4066Sahrens * to reference the just-allocated physical data area. 775fa9e4066Sahrens */ 776fa9e4066Sahrens if (flag & IS_ROOT_NODE) { 777874395d5Smaybee dzp->z_dbuf = db; 778fa9e4066Sahrens dzp->z_phys = pzp; 7794ccbb6e7Sahrens dzp->z_id = obj; 780fa9e4066Sahrens } 781fa9e4066Sahrens 782fa9e4066Sahrens /* 783fa9e4066Sahrens * If parent is an xattr, so am I. 784fa9e4066Sahrens */ 785fa9e4066Sahrens if (dzp->z_phys->zp_flags & ZFS_XATTR) 786fa9e4066Sahrens flag |= IS_XATTR; 787fa9e4066Sahrens 788fa9e4066Sahrens if (vap->va_type == VBLK || vap->va_type == VCHR) { 78972fc53bcSmarks pzp->zp_rdev = zfs_expldev(vap->va_rdev); 790fa9e4066Sahrens } 791fa9e4066Sahrens 792da6c28aaSamw if (zfsvfs->z_use_fuids) 793da6c28aaSamw pzp->zp_flags = ZFS_ARCHIVE | ZFS_AV_MODIFIED; 794da6c28aaSamw 795fa9e4066Sahrens if (vap->va_type == VDIR) { 796fa9e4066Sahrens pzp->zp_size = 2; /* contents ("." and "..") */ 797fa9e4066Sahrens pzp->zp_links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1; 798fa9e4066Sahrens } 799fa9e4066Sahrens 800fa9e4066Sahrens pzp->zp_parent = dzp->z_id; 801fa9e4066Sahrens if (flag & IS_XATTR) 802fa9e4066Sahrens pzp->zp_flags |= ZFS_XATTR; 803fa9e4066Sahrens 804fa9e4066Sahrens pzp->zp_gen = gen; 805fa9e4066Sahrens 806fa9e4066Sahrens ZFS_TIME_ENCODE(&now, pzp->zp_crtime); 807fa9e4066Sahrens ZFS_TIME_ENCODE(&now, pzp->zp_ctime); 808fa9e4066Sahrens 809fa9e4066Sahrens if (vap->va_mask & AT_ATIME) { 810fa9e4066Sahrens ZFS_TIME_ENCODE(&vap->va_atime, pzp->zp_atime); 811fa9e4066Sahrens } else { 812fa9e4066Sahrens ZFS_TIME_ENCODE(&now, pzp->zp_atime); 813fa9e4066Sahrens } 814fa9e4066Sahrens 815fa9e4066Sahrens if (vap->va_mask & AT_MTIME) { 816fa9e4066Sahrens ZFS_TIME_ENCODE(&vap->va_mtime, pzp->zp_mtime); 817fa9e4066Sahrens } else { 818fa9e4066Sahrens ZFS_TIME_ENCODE(&now, pzp->zp_mtime); 819fa9e4066Sahrens } 820fa9e4066Sahrens 821fa9e4066Sahrens pzp->zp_mode = MAKEIMODE(vap->va_type, vap->va_mode); 822874395d5Smaybee if (!(flag & IS_ROOT_NODE)) { 823b5fca8f8Stomee ZFS_OBJ_HOLD_ENTER(zfsvfs, obj); 824874395d5Smaybee *zpp = zfs_znode_alloc(zfsvfs, db, 0); 825874395d5Smaybee ZFS_OBJ_HOLD_EXIT(zfsvfs, obj); 826874395d5Smaybee } else { 827874395d5Smaybee /* 828874395d5Smaybee * If we are creating the root node, the "parent" we 829874395d5Smaybee * passed in is the znode for the root. 830874395d5Smaybee */ 831874395d5Smaybee *zpp = dzp; 832874395d5Smaybee } 83389459e17SMark Shellenbaum pzp->zp_uid = acl_ids->z_fuid; 83489459e17SMark Shellenbaum pzp->zp_gid = acl_ids->z_fgid; 83589459e17SMark Shellenbaum pzp->zp_mode = acl_ids->z_mode; 83689459e17SMark Shellenbaum VERIFY(0 == zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx)); 83789459e17SMark Shellenbaum if (vap->va_mask & AT_XVATTR) 83889459e17SMark Shellenbaum zfs_xvattr_set(*zpp, (xvattr_t *)vap); 839fa9e4066Sahrens } 840fa9e4066Sahrens 841da6c28aaSamw void 842da6c28aaSamw zfs_xvattr_set(znode_t *zp, xvattr_t *xvap) 843da6c28aaSamw { 844da6c28aaSamw xoptattr_t *xoap; 845da6c28aaSamw 846da6c28aaSamw xoap = xva_getxoptattr(xvap); 847da6c28aaSamw ASSERT(xoap); 848da6c28aaSamw 849da6c28aaSamw if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) { 850da6c28aaSamw ZFS_TIME_ENCODE(&xoap->xoa_createtime, zp->z_phys->zp_crtime); 851da6c28aaSamw XVA_SET_RTN(xvap, XAT_CREATETIME); 852da6c28aaSamw } 853da6c28aaSamw if (XVA_ISSET_REQ(xvap, XAT_READONLY)) { 854da6c28aaSamw ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly); 855da6c28aaSamw XVA_SET_RTN(xvap, XAT_READONLY); 856da6c28aaSamw } 857da6c28aaSamw if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) { 858da6c28aaSamw ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden); 859da6c28aaSamw XVA_SET_RTN(xvap, XAT_HIDDEN); 860da6c28aaSamw } 861da6c28aaSamw if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) { 862da6c28aaSamw ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system); 863da6c28aaSamw XVA_SET_RTN(xvap, XAT_SYSTEM); 864da6c28aaSamw } 865da6c28aaSamw if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) { 866da6c28aaSamw ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive); 867da6c28aaSamw XVA_SET_RTN(xvap, XAT_ARCHIVE); 868da6c28aaSamw } 869da6c28aaSamw if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { 870da6c28aaSamw ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable); 871da6c28aaSamw XVA_SET_RTN(xvap, XAT_IMMUTABLE); 872da6c28aaSamw } 873da6c28aaSamw if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { 874da6c28aaSamw ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink); 875da6c28aaSamw XVA_SET_RTN(xvap, XAT_NOUNLINK); 876da6c28aaSamw } 877da6c28aaSamw if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { 878da6c28aaSamw ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly); 879da6c28aaSamw XVA_SET_RTN(xvap, XAT_APPENDONLY); 880da6c28aaSamw } 881da6c28aaSamw if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { 882da6c28aaSamw ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump); 883da6c28aaSamw XVA_SET_RTN(xvap, XAT_NODUMP); 884da6c28aaSamw } 885da6c28aaSamw if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) { 886da6c28aaSamw ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque); 887da6c28aaSamw XVA_SET_RTN(xvap, XAT_OPAQUE); 888da6c28aaSamw } 889da6c28aaSamw if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { 890da6c28aaSamw ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED, 891da6c28aaSamw xoap->xoa_av_quarantined); 892da6c28aaSamw XVA_SET_RTN(xvap, XAT_AV_QUARANTINED); 893da6c28aaSamw } 894da6c28aaSamw if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { 895da6c28aaSamw ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified); 896da6c28aaSamw XVA_SET_RTN(xvap, XAT_AV_MODIFIED); 897da6c28aaSamw } 898da6c28aaSamw if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) { 899da6c28aaSamw (void) memcpy(zp->z_phys + 1, xoap->xoa_av_scanstamp, 900da6c28aaSamw sizeof (xoap->xoa_av_scanstamp)); 901da6c28aaSamw zp->z_phys->zp_flags |= ZFS_BONUS_SCANSTAMP; 902da6c28aaSamw XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP); 903da6c28aaSamw } 904da6c28aaSamw } 905da6c28aaSamw 906fa9e4066Sahrens int 907fa9e4066Sahrens zfs_zget(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp) 908fa9e4066Sahrens { 909fa9e4066Sahrens dmu_object_info_t doi; 910fa9e4066Sahrens dmu_buf_t *db; 911fa9e4066Sahrens znode_t *zp; 912ea8dc4b6Seschrock int err; 913fa9e4066Sahrens 914fa9e4066Sahrens *zpp = NULL; 915fa9e4066Sahrens 916fa9e4066Sahrens ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num); 917fa9e4066Sahrens 918ea8dc4b6Seschrock err = dmu_bonus_hold(zfsvfs->z_os, obj_num, NULL, &db); 919ea8dc4b6Seschrock if (err) { 920fa9e4066Sahrens ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 921ea8dc4b6Seschrock return (err); 922fa9e4066Sahrens } 923fa9e4066Sahrens 924fa9e4066Sahrens dmu_object_info_from_db(db, &doi); 925fa9e4066Sahrens if (doi.doi_bonus_type != DMU_OT_ZNODE || 926fa9e4066Sahrens doi.doi_bonus_size < sizeof (znode_phys_t)) { 927ea8dc4b6Seschrock dmu_buf_rele(db, NULL); 928fa9e4066Sahrens ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 929fa9e4066Sahrens return (EINVAL); 930fa9e4066Sahrens } 931fa9e4066Sahrens 932fa9e4066Sahrens zp = dmu_buf_get_user(db); 933fa9e4066Sahrens if (zp != NULL) { 934fa9e4066Sahrens mutex_enter(&zp->z_lock); 935fa9e4066Sahrens 9364ccbb6e7Sahrens /* 9374ccbb6e7Sahrens * Since we do immediate eviction of the z_dbuf, we 9384ccbb6e7Sahrens * should never find a dbuf with a znode that doesn't 9394ccbb6e7Sahrens * know about the dbuf. 9404ccbb6e7Sahrens */ 9414ccbb6e7Sahrens ASSERT3P(zp->z_dbuf, ==, db); 942fa9e4066Sahrens ASSERT3U(zp->z_id, ==, obj_num); 943893a6d32Sahrens if (zp->z_unlinked) { 9444ccbb6e7Sahrens err = ENOENT; 945fa9e4066Sahrens } else { 9464ccbb6e7Sahrens VN_HOLD(ZTOV(zp)); 9474ccbb6e7Sahrens *zpp = zp; 9484ccbb6e7Sahrens err = 0; 949fa9e4066Sahrens } 9504ccbb6e7Sahrens dmu_buf_rele(db, NULL); 951fa9e4066Sahrens mutex_exit(&zp->z_lock); 952ea8dc4b6Seschrock ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 9534ccbb6e7Sahrens return (err); 954fa9e4066Sahrens } 955fa9e4066Sahrens 956fa9e4066Sahrens /* 957fa9e4066Sahrens * Not found create new znode/vnode 958fa9e4066Sahrens */ 9594ccbb6e7Sahrens zp = zfs_znode_alloc(zfsvfs, db, doi.doi_data_block_size); 960ea8dc4b6Seschrock ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 961fa9e4066Sahrens *zpp = zp; 962fa9e4066Sahrens return (0); 963fa9e4066Sahrens } 964fa9e4066Sahrens 965f18faf3fSek int 966f18faf3fSek zfs_rezget(znode_t *zp) 967f18faf3fSek { 968f18faf3fSek zfsvfs_t *zfsvfs = zp->z_zfsvfs; 969f18faf3fSek dmu_object_info_t doi; 970f18faf3fSek dmu_buf_t *db; 971f18faf3fSek uint64_t obj_num = zp->z_id; 972f18faf3fSek int err; 973f18faf3fSek 974f18faf3fSek ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num); 975f18faf3fSek 976f18faf3fSek err = dmu_bonus_hold(zfsvfs->z_os, obj_num, NULL, &db); 977f18faf3fSek if (err) { 978f18faf3fSek ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 979f18faf3fSek return (err); 980f18faf3fSek } 981f18faf3fSek 982f18faf3fSek dmu_object_info_from_db(db, &doi); 983f18faf3fSek if (doi.doi_bonus_type != DMU_OT_ZNODE || 984f18faf3fSek doi.doi_bonus_size < sizeof (znode_phys_t)) { 985f18faf3fSek dmu_buf_rele(db, NULL); 986f18faf3fSek ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 987f18faf3fSek return (EINVAL); 988f18faf3fSek } 989f18faf3fSek 990f18faf3fSek if (((znode_phys_t *)db->db_data)->zp_gen != zp->z_gen) { 991f18faf3fSek dmu_buf_rele(db, NULL); 992f18faf3fSek ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 993f18faf3fSek return (EIO); 994f18faf3fSek } 995f18faf3fSek 996b5fca8f8Stomee zfs_znode_dmu_init(zfsvfs, zp, db); 997f18faf3fSek zp->z_unlinked = (zp->z_phys->zp_links == 0); 9986166ad1cSek zp->z_blksz = doi.doi_data_block_size; 999f18faf3fSek 1000f18faf3fSek ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1001f18faf3fSek 1002f18faf3fSek return (0); 1003f18faf3fSek } 1004f18faf3fSek 1005fa9e4066Sahrens void 1006fa9e4066Sahrens zfs_znode_delete(znode_t *zp, dmu_tx_t *tx) 1007fa9e4066Sahrens { 1008fa9e4066Sahrens zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1009cdb0ab79Smaybee objset_t *os = zfsvfs->z_os; 10104ccbb6e7Sahrens uint64_t obj = zp->z_id; 1011cdb0ab79Smaybee uint64_t acl_obj = zp->z_phys->zp_acl.z_acl_extern_obj; 1012fa9e4066Sahrens 10134ccbb6e7Sahrens ZFS_OBJ_HOLD_ENTER(zfsvfs, obj); 1014cdb0ab79Smaybee if (acl_obj) 1015cdb0ab79Smaybee VERIFY(0 == dmu_object_free(os, acl_obj, tx)); 1016cdb0ab79Smaybee VERIFY(0 == dmu_object_free(os, obj, tx)); 10174ccbb6e7Sahrens zfs_znode_dmu_fini(zp); 10184ccbb6e7Sahrens ZFS_OBJ_HOLD_EXIT(zfsvfs, obj); 1019874395d5Smaybee zfs_znode_free(zp); 1020fa9e4066Sahrens } 1021fa9e4066Sahrens 1022fa9e4066Sahrens void 1023fa9e4066Sahrens zfs_zinactive(znode_t *zp) 1024fa9e4066Sahrens { 1025fa9e4066Sahrens vnode_t *vp = ZTOV(zp); 1026fa9e4066Sahrens zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1027fa9e4066Sahrens uint64_t z_id = zp->z_id; 1028fa9e4066Sahrens 10294ccbb6e7Sahrens ASSERT(zp->z_dbuf && zp->z_phys); 1030fa9e4066Sahrens 1031fa9e4066Sahrens /* 1032fa9e4066Sahrens * Don't allow a zfs_zget() while were trying to release this znode 1033fa9e4066Sahrens */ 1034fa9e4066Sahrens ZFS_OBJ_HOLD_ENTER(zfsvfs, z_id); 1035fa9e4066Sahrens 1036fa9e4066Sahrens mutex_enter(&zp->z_lock); 1037fa9e4066Sahrens mutex_enter(&vp->v_lock); 1038fa9e4066Sahrens vp->v_count--; 1039fa9e4066Sahrens if (vp->v_count > 0 || vn_has_cached_data(vp)) { 1040fa9e4066Sahrens /* 1041fa9e4066Sahrens * If the hold count is greater than zero, somebody has 1042fa9e4066Sahrens * obtained a new reference on this znode while we were 1043fa9e4066Sahrens * processing it here, so we are done. If we still have 1044fa9e4066Sahrens * mapped pages then we are also done, since we don't 1045fa9e4066Sahrens * want to inactivate the znode until the pages get pushed. 1046fa9e4066Sahrens * 1047fa9e4066Sahrens * XXX - if vn_has_cached_data(vp) is true, but count == 0, 1048fa9e4066Sahrens * this seems like it would leave the znode hanging with 1049fa9e4066Sahrens * no chance to go inactive... 1050fa9e4066Sahrens */ 1051fa9e4066Sahrens mutex_exit(&vp->v_lock); 1052fa9e4066Sahrens mutex_exit(&zp->z_lock); 1053fa9e4066Sahrens ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id); 1054fa9e4066Sahrens return; 1055fa9e4066Sahrens } 1056fa9e4066Sahrens mutex_exit(&vp->v_lock); 1057fa9e4066Sahrens 1058fa9e4066Sahrens /* 1059fa9e4066Sahrens * If this was the last reference to a file with no links, 1060fa9e4066Sahrens * remove the file from the file system. 1061fa9e4066Sahrens */ 1062893a6d32Sahrens if (zp->z_unlinked) { 1063fa9e4066Sahrens mutex_exit(&zp->z_lock); 1064fa9e4066Sahrens ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id); 1065893a6d32Sahrens zfs_rmnode(zp); 1066fa9e4066Sahrens return; 1067fa9e4066Sahrens } 1068fa9e4066Sahrens mutex_exit(&zp->z_lock); 10694ccbb6e7Sahrens zfs_znode_dmu_fini(zp); 1070fa9e4066Sahrens ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id); 1071874395d5Smaybee zfs_znode_free(zp); 1072fa9e4066Sahrens } 1073fa9e4066Sahrens 1074fa9e4066Sahrens void 1075fa9e4066Sahrens zfs_znode_free(znode_t *zp) 1076fa9e4066Sahrens { 1077fa9e4066Sahrens zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1078fa9e4066Sahrens 1079874395d5Smaybee vn_invalid(ZTOV(zp)); 1080874395d5Smaybee 1081b5fca8f8Stomee ASSERT(ZTOV(zp)->v_count == 0); 1082b5fca8f8Stomee 1083fa9e4066Sahrens mutex_enter(&zfsvfs->z_znodes_lock); 1084b5fca8f8Stomee POINTER_INVALIDATE(&zp->z_zfsvfs); 1085fa9e4066Sahrens list_remove(&zfsvfs->z_all_znodes, zp); 1086fa9e4066Sahrens mutex_exit(&zfsvfs->z_znodes_lock); 1087fa9e4066Sahrens 1088d47621a4STim Haley if (zp->z_acl_cached) { 1089d47621a4STim Haley zfs_acl_free(zp->z_acl_cached); 1090d47621a4STim Haley zp->z_acl_cached = NULL; 1091d47621a4STim Haley } 1092d47621a4STim Haley 1093fa9e4066Sahrens kmem_cache_free(znode_cache, zp); 1094874395d5Smaybee 1095874395d5Smaybee VFS_RELE(zfsvfs->z_vfs); 1096fa9e4066Sahrens } 1097fa9e4066Sahrens 1098fa9e4066Sahrens void 1099fa9e4066Sahrens zfs_time_stamper_locked(znode_t *zp, uint_t flag, dmu_tx_t *tx) 1100fa9e4066Sahrens { 1101fa9e4066Sahrens timestruc_t now; 1102fa9e4066Sahrens 1103fa9e4066Sahrens ASSERT(MUTEX_HELD(&zp->z_lock)); 1104fa9e4066Sahrens 1105fa9e4066Sahrens gethrestime(&now); 1106fa9e4066Sahrens 1107fa9e4066Sahrens if (tx) { 1108fa9e4066Sahrens dmu_buf_will_dirty(zp->z_dbuf, tx); 1109fa9e4066Sahrens zp->z_atime_dirty = 0; 1110fa9e4066Sahrens zp->z_seq++; 1111fa9e4066Sahrens } else { 1112fa9e4066Sahrens zp->z_atime_dirty = 1; 1113fa9e4066Sahrens } 1114fa9e4066Sahrens 1115fa9e4066Sahrens if (flag & AT_ATIME) 1116fa9e4066Sahrens ZFS_TIME_ENCODE(&now, zp->z_phys->zp_atime); 1117fa9e4066Sahrens 1118da6c28aaSamw if (flag & AT_MTIME) { 1119fa9e4066Sahrens ZFS_TIME_ENCODE(&now, zp->z_phys->zp_mtime); 1120da6c28aaSamw if (zp->z_zfsvfs->z_use_fuids) 1121da6c28aaSamw zp->z_phys->zp_flags |= (ZFS_ARCHIVE | ZFS_AV_MODIFIED); 1122da6c28aaSamw } 1123fa9e4066Sahrens 1124da6c28aaSamw if (flag & AT_CTIME) { 1125fa9e4066Sahrens ZFS_TIME_ENCODE(&now, zp->z_phys->zp_ctime); 1126da6c28aaSamw if (zp->z_zfsvfs->z_use_fuids) 1127da6c28aaSamw zp->z_phys->zp_flags |= ZFS_ARCHIVE; 1128da6c28aaSamw } 1129fa9e4066Sahrens } 1130fa9e4066Sahrens 1131fa9e4066Sahrens /* 1132fa9e4066Sahrens * Update the requested znode timestamps with the current time. 1133fa9e4066Sahrens * If we are in a transaction, then go ahead and mark the znode 1134fa9e4066Sahrens * dirty in the transaction so the timestamps will go to disk. 1135fa9e4066Sahrens * Otherwise, we will get pushed next time the znode is updated 1136fa9e4066Sahrens * in a transaction, or when this znode eventually goes inactive. 1137fa9e4066Sahrens * 1138fa9e4066Sahrens * Why is this OK? 1139fa9e4066Sahrens * 1 - Only the ACCESS time is ever updated outside of a transaction. 1140fa9e4066Sahrens * 2 - Multiple consecutive updates will be collapsed into a single 1141fa9e4066Sahrens * znode update by the transaction grouping semantics of the DMU. 1142fa9e4066Sahrens */ 1143fa9e4066Sahrens void 1144fa9e4066Sahrens zfs_time_stamper(znode_t *zp, uint_t flag, dmu_tx_t *tx) 1145fa9e4066Sahrens { 1146fa9e4066Sahrens mutex_enter(&zp->z_lock); 1147fa9e4066Sahrens zfs_time_stamper_locked(zp, flag, tx); 1148fa9e4066Sahrens mutex_exit(&zp->z_lock); 1149fa9e4066Sahrens } 1150fa9e4066Sahrens 1151fa9e4066Sahrens /* 1152104e2ed7Sperrin * Grow the block size for a file. 1153fa9e4066Sahrens * 1154fa9e4066Sahrens * IN: zp - znode of file to free data in. 1155fa9e4066Sahrens * size - requested block size 1156fa9e4066Sahrens * tx - open transaction. 1157fa9e4066Sahrens * 1158fa9e4066Sahrens * NOTE: this function assumes that the znode is write locked. 1159fa9e4066Sahrens */ 1160104e2ed7Sperrin void 1161fa9e4066Sahrens zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx) 1162fa9e4066Sahrens { 1163fa9e4066Sahrens int error; 1164fa9e4066Sahrens u_longlong_t dummy; 1165fa9e4066Sahrens 1166fa9e4066Sahrens if (size <= zp->z_blksz) 1167104e2ed7Sperrin return; 1168fa9e4066Sahrens /* 1169fa9e4066Sahrens * If the file size is already greater than the current blocksize, 1170fa9e4066Sahrens * we will not grow. If there is more than one block in a file, 1171fa9e4066Sahrens * the blocksize cannot change. 1172fa9e4066Sahrens */ 1173fa9e4066Sahrens if (zp->z_blksz && zp->z_phys->zp_size > zp->z_blksz) 1174104e2ed7Sperrin return; 1175fa9e4066Sahrens 1176fa9e4066Sahrens error = dmu_object_set_blocksize(zp->z_zfsvfs->z_os, zp->z_id, 1177fa9e4066Sahrens size, 0, tx); 1178fa9e4066Sahrens if (error == ENOTSUP) 1179104e2ed7Sperrin return; 1180fa9e4066Sahrens ASSERT3U(error, ==, 0); 1181fa9e4066Sahrens 1182fa9e4066Sahrens /* What blocksize did we actually get? */ 1183fa9e4066Sahrens dmu_object_size_from_db(zp->z_dbuf, &zp->z_blksz, &dummy); 1184fa9e4066Sahrens } 1185fa9e4066Sahrens 1186fa9e4066Sahrens /* 1187fa9e4066Sahrens * This is a dummy interface used when pvn_vplist_dirty() should *not* 1188fa9e4066Sahrens * be calling back into the fs for a putpage(). E.g.: when truncating 1189fa9e4066Sahrens * a file, the pages being "thrown away* don't need to be written out. 1190fa9e4066Sahrens */ 1191fa9e4066Sahrens /* ARGSUSED */ 1192fa9e4066Sahrens static int 1193fa9e4066Sahrens zfs_no_putpage(vnode_t *vp, page_t *pp, u_offset_t *offp, size_t *lenp, 1194fa9e4066Sahrens int flags, cred_t *cr) 1195fa9e4066Sahrens { 1196fa9e4066Sahrens ASSERT(0); 1197fa9e4066Sahrens return (0); 1198fa9e4066Sahrens } 1199fa9e4066Sahrens 1200fa9e4066Sahrens /* 1201cdb0ab79Smaybee * Increase the file length 1202fa9e4066Sahrens * 1203fa9e4066Sahrens * IN: zp - znode of file to free data in. 1204cdb0ab79Smaybee * end - new end-of-file 1205fa9e4066Sahrens * 1206fa9e4066Sahrens * RETURN: 0 if success 1207fa9e4066Sahrens * error code if failure 1208fa9e4066Sahrens */ 1209cdb0ab79Smaybee static int 1210cdb0ab79Smaybee zfs_extend(znode_t *zp, uint64_t end) 1211fa9e4066Sahrens { 12125730cc9aSmaybee zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1213cdb0ab79Smaybee dmu_tx_t *tx; 12145730cc9aSmaybee rl_t *rl; 1215cdb0ab79Smaybee uint64_t newblksz; 1216104e2ed7Sperrin int error; 1217fa9e4066Sahrens 12185730cc9aSmaybee /* 1219cdb0ab79Smaybee * We will change zp_size, lock the whole file. 12205730cc9aSmaybee */ 1221cdb0ab79Smaybee rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER); 12225730cc9aSmaybee 1223fa9e4066Sahrens /* 1224fa9e4066Sahrens * Nothing to do if file already at desired length. 1225fa9e4066Sahrens */ 1226cdb0ab79Smaybee if (end <= zp->z_phys->zp_size) { 1227c5c6ffa0Smaybee zfs_range_unlock(rl); 1228fa9e4066Sahrens return (0); 1229fa9e4066Sahrens } 1230cdb0ab79Smaybee top: 12315730cc9aSmaybee tx = dmu_tx_create(zfsvfs->z_os); 12325730cc9aSmaybee dmu_tx_hold_bonus(tx, zp->z_id); 1233cdb0ab79Smaybee if (end > zp->z_blksz && 12345730cc9aSmaybee (!ISP2(zp->z_blksz) || zp->z_blksz < zfsvfs->z_max_blksz)) { 1235fa9e4066Sahrens /* 1236fa9e4066Sahrens * We are growing the file past the current block size. 1237fa9e4066Sahrens */ 1238fa9e4066Sahrens if (zp->z_blksz > zp->z_zfsvfs->z_max_blksz) { 1239fa9e4066Sahrens ASSERT(!ISP2(zp->z_blksz)); 1240cdb0ab79Smaybee newblksz = MIN(end, SPA_MAXBLOCKSIZE); 1241fa9e4066Sahrens } else { 1242cdb0ab79Smaybee newblksz = MIN(end, zp->z_zfsvfs->z_max_blksz); 1243fa9e4066Sahrens } 1244cdb0ab79Smaybee dmu_tx_hold_write(tx, zp->z_id, 0, newblksz); 1245cdb0ab79Smaybee } else { 1246cdb0ab79Smaybee newblksz = 0; 12475730cc9aSmaybee } 12485730cc9aSmaybee 12491209a471SNeil Perrin error = dmu_tx_assign(tx, TXG_NOWAIT); 12505730cc9aSmaybee if (error) { 12511209a471SNeil Perrin if (error == ERESTART) { 12528a2f1b91Sahrens dmu_tx_wait(tx); 1253cdb0ab79Smaybee dmu_tx_abort(tx); 1254cdb0ab79Smaybee goto top; 1255cdb0ab79Smaybee } 12565730cc9aSmaybee dmu_tx_abort(tx); 1257c5c6ffa0Smaybee zfs_range_unlock(rl); 12585730cc9aSmaybee return (error); 1259fa9e4066Sahrens } 1260cdb0ab79Smaybee dmu_buf_will_dirty(zp->z_dbuf, tx); 12615730cc9aSmaybee 1262cdb0ab79Smaybee if (newblksz) 1263cdb0ab79Smaybee zfs_grow_blocksize(zp, newblksz, tx); 12645730cc9aSmaybee 1265cdb0ab79Smaybee zp->z_phys->zp_size = end; 12665730cc9aSmaybee 1267cdb0ab79Smaybee zfs_range_unlock(rl); 12685730cc9aSmaybee 1269cdb0ab79Smaybee dmu_tx_commit(tx); 12705730cc9aSmaybee 1271cdb0ab79Smaybee return (0); 1272cdb0ab79Smaybee } 1273cdb0ab79Smaybee 1274cdb0ab79Smaybee /* 1275cdb0ab79Smaybee * Free space in a file. 1276cdb0ab79Smaybee * 1277cdb0ab79Smaybee * IN: zp - znode of file to free data in. 1278cdb0ab79Smaybee * off - start of section to free. 1279cdb0ab79Smaybee * len - length of section to free. 1280cdb0ab79Smaybee * 1281cdb0ab79Smaybee * RETURN: 0 if success 1282cdb0ab79Smaybee * error code if failure 1283cdb0ab79Smaybee */ 1284cdb0ab79Smaybee static int 1285cdb0ab79Smaybee zfs_free_range(znode_t *zp, uint64_t off, uint64_t len) 1286cdb0ab79Smaybee { 1287cdb0ab79Smaybee zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1288cdb0ab79Smaybee rl_t *rl; 1289cdb0ab79Smaybee int error; 1290cdb0ab79Smaybee 1291cdb0ab79Smaybee /* 1292cdb0ab79Smaybee * Lock the range being freed. 1293cdb0ab79Smaybee */ 1294cdb0ab79Smaybee rl = zfs_range_lock(zp, off, len, RL_WRITER); 1295cdb0ab79Smaybee 1296cdb0ab79Smaybee /* 1297cdb0ab79Smaybee * Nothing to do if file already at desired length. 1298cdb0ab79Smaybee */ 1299cdb0ab79Smaybee if (off >= zp->z_phys->zp_size) { 1300cdb0ab79Smaybee zfs_range_unlock(rl); 1301cdb0ab79Smaybee return (0); 13025730cc9aSmaybee } 13035730cc9aSmaybee 1304cdb0ab79Smaybee if (off + len > zp->z_phys->zp_size) 1305cdb0ab79Smaybee len = zp->z_phys->zp_size - off; 1306cdb0ab79Smaybee 1307cdb0ab79Smaybee error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, off, len); 1308cdb0ab79Smaybee 1309c5c6ffa0Smaybee zfs_range_unlock(rl); 13105730cc9aSmaybee 1311cdb0ab79Smaybee return (error); 1312cdb0ab79Smaybee } 1313cdb0ab79Smaybee 1314cdb0ab79Smaybee /* 1315cdb0ab79Smaybee * Truncate a file 1316cdb0ab79Smaybee * 1317cdb0ab79Smaybee * IN: zp - znode of file to free data in. 1318cdb0ab79Smaybee * end - new end-of-file. 1319cdb0ab79Smaybee * 1320cdb0ab79Smaybee * RETURN: 0 if success 1321cdb0ab79Smaybee * error code if failure 1322cdb0ab79Smaybee */ 1323cdb0ab79Smaybee static int 1324cdb0ab79Smaybee zfs_trunc(znode_t *zp, uint64_t end) 1325cdb0ab79Smaybee { 1326cdb0ab79Smaybee zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1327cdb0ab79Smaybee vnode_t *vp = ZTOV(zp); 1328cdb0ab79Smaybee dmu_tx_t *tx; 1329cdb0ab79Smaybee rl_t *rl; 1330cdb0ab79Smaybee int error; 1331cdb0ab79Smaybee 1332cdb0ab79Smaybee /* 1333cdb0ab79Smaybee * We will change zp_size, lock the whole file. 1334cdb0ab79Smaybee */ 1335cdb0ab79Smaybee rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER); 1336cdb0ab79Smaybee 1337cdb0ab79Smaybee /* 1338cdb0ab79Smaybee * Nothing to do if file already at desired length. 1339cdb0ab79Smaybee */ 1340cdb0ab79Smaybee if (end >= zp->z_phys->zp_size) { 1341cdb0ab79Smaybee zfs_range_unlock(rl); 1342cdb0ab79Smaybee return (0); 1343cdb0ab79Smaybee } 1344cdb0ab79Smaybee 1345cdb0ab79Smaybee error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, end, -1); 1346cdb0ab79Smaybee if (error) { 1347cdb0ab79Smaybee zfs_range_unlock(rl); 1348cdb0ab79Smaybee return (error); 1349cdb0ab79Smaybee } 1350cdb0ab79Smaybee top: 1351cdb0ab79Smaybee tx = dmu_tx_create(zfsvfs->z_os); 1352cdb0ab79Smaybee dmu_tx_hold_bonus(tx, zp->z_id); 13531209a471SNeil Perrin error = dmu_tx_assign(tx, TXG_NOWAIT); 1354cdb0ab79Smaybee if (error) { 13551209a471SNeil Perrin if (error == ERESTART) { 1356cdb0ab79Smaybee dmu_tx_wait(tx); 1357cdb0ab79Smaybee dmu_tx_abort(tx); 1358cdb0ab79Smaybee goto top; 1359cdb0ab79Smaybee } 1360cdb0ab79Smaybee dmu_tx_abort(tx); 1361cdb0ab79Smaybee zfs_range_unlock(rl); 1362cdb0ab79Smaybee return (error); 1363cdb0ab79Smaybee } 1364cdb0ab79Smaybee dmu_buf_will_dirty(zp->z_dbuf, tx); 1365cdb0ab79Smaybee 1366cdb0ab79Smaybee zp->z_phys->zp_size = end; 1367cdb0ab79Smaybee 13685730cc9aSmaybee dmu_tx_commit(tx); 13695730cc9aSmaybee 1370fa9e4066Sahrens /* 13715730cc9aSmaybee * Clear any mapped pages in the truncated region. This has to 13725730cc9aSmaybee * happen outside of the transaction to avoid the possibility of 13735730cc9aSmaybee * a deadlock with someone trying to push a page that we are 13745730cc9aSmaybee * about to invalidate. 1375fa9e4066Sahrens */ 1376cdb0ab79Smaybee if (vn_has_cached_data(vp)) { 1377fa9e4066Sahrens page_t *pp; 1378cdb0ab79Smaybee uint64_t start = end & PAGEMASK; 1379cdb0ab79Smaybee int poff = end & PAGEOFFSET; 1380fa9e4066Sahrens 13815730cc9aSmaybee if (poff != 0 && (pp = page_lookup(vp, start, SE_SHARED))) { 1382fa9e4066Sahrens /* 1383fa9e4066Sahrens * We need to zero a partial page. 1384fa9e4066Sahrens */ 13855730cc9aSmaybee pagezero(pp, poff, PAGESIZE - poff); 1386fa9e4066Sahrens start += PAGESIZE; 1387fa9e4066Sahrens page_unlock(pp); 1388fa9e4066Sahrens } 1389fa9e4066Sahrens error = pvn_vplist_dirty(vp, start, zfs_no_putpage, 13905730cc9aSmaybee B_INVAL | B_TRUNC, NULL); 1391fa9e4066Sahrens ASSERT(error == 0); 1392fa9e4066Sahrens } 1393ac05c741SMark Maybee 1394ac05c741SMark Maybee zfs_range_unlock(rl); 1395fa9e4066Sahrens 1396fa9e4066Sahrens return (0); 1397fa9e4066Sahrens } 1398fa9e4066Sahrens 1399cdb0ab79Smaybee /* 1400cdb0ab79Smaybee * Free space in a file 1401cdb0ab79Smaybee * 1402cdb0ab79Smaybee * IN: zp - znode of file to free data in. 1403cdb0ab79Smaybee * off - start of range 1404cdb0ab79Smaybee * len - end of range (0 => EOF) 1405cdb0ab79Smaybee * flag - current file open mode flags. 1406cdb0ab79Smaybee * log - TRUE if this action should be logged 1407cdb0ab79Smaybee * 1408cdb0ab79Smaybee * RETURN: 0 if success 1409cdb0ab79Smaybee * error code if failure 1410cdb0ab79Smaybee */ 1411cdb0ab79Smaybee int 1412cdb0ab79Smaybee zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log) 1413cdb0ab79Smaybee { 1414cdb0ab79Smaybee vnode_t *vp = ZTOV(zp); 1415cdb0ab79Smaybee dmu_tx_t *tx; 1416cdb0ab79Smaybee zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1417cdb0ab79Smaybee zilog_t *zilog = zfsvfs->z_log; 1418cdb0ab79Smaybee int error; 1419cdb0ab79Smaybee 1420cdb0ab79Smaybee if (off > zp->z_phys->zp_size) { 1421cdb0ab79Smaybee error = zfs_extend(zp, off+len); 1422cdb0ab79Smaybee if (error == 0 && log) 1423cdb0ab79Smaybee goto log; 1424cdb0ab79Smaybee else 1425cdb0ab79Smaybee return (error); 1426cdb0ab79Smaybee } 1427cdb0ab79Smaybee 1428cdb0ab79Smaybee /* 1429cdb0ab79Smaybee * Check for any locks in the region to be freed. 1430cdb0ab79Smaybee */ 1431cdb0ab79Smaybee if (MANDLOCK(vp, (mode_t)zp->z_phys->zp_mode)) { 1432cdb0ab79Smaybee uint64_t length = (len ? len : zp->z_phys->zp_size - off); 1433cdb0ab79Smaybee if (error = chklock(vp, FWRITE, off, length, flag, NULL)) 1434cdb0ab79Smaybee return (error); 1435cdb0ab79Smaybee } 1436cdb0ab79Smaybee 1437cdb0ab79Smaybee if (len == 0) { 1438cdb0ab79Smaybee error = zfs_trunc(zp, off); 1439cdb0ab79Smaybee } else { 1440cdb0ab79Smaybee if ((error = zfs_free_range(zp, off, len)) == 0 && 1441cdb0ab79Smaybee off + len > zp->z_phys->zp_size) 1442cdb0ab79Smaybee error = zfs_extend(zp, off+len); 1443cdb0ab79Smaybee } 1444cdb0ab79Smaybee if (error || !log) 1445cdb0ab79Smaybee return (error); 1446cdb0ab79Smaybee log: 1447cdb0ab79Smaybee tx = dmu_tx_create(zfsvfs->z_os); 1448cdb0ab79Smaybee dmu_tx_hold_bonus(tx, zp->z_id); 14491209a471SNeil Perrin error = dmu_tx_assign(tx, TXG_NOWAIT); 1450cdb0ab79Smaybee if (error) { 14511209a471SNeil Perrin if (error == ERESTART) { 1452cdb0ab79Smaybee dmu_tx_wait(tx); 1453cdb0ab79Smaybee dmu_tx_abort(tx); 1454cdb0ab79Smaybee goto log; 1455cdb0ab79Smaybee } 1456cdb0ab79Smaybee dmu_tx_abort(tx); 1457cdb0ab79Smaybee return (error); 1458cdb0ab79Smaybee } 1459cdb0ab79Smaybee 1460cdb0ab79Smaybee zfs_time_stamper(zp, CONTENT_MODIFIED, tx); 1461cdb0ab79Smaybee zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len); 1462cdb0ab79Smaybee 1463cdb0ab79Smaybee dmu_tx_commit(tx); 1464cdb0ab79Smaybee return (0); 1465cdb0ab79Smaybee } 1466cdb0ab79Smaybee 1467fa9e4066Sahrens void 1468de8267e0Stimh zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx) 1469fa9e4066Sahrens { 1470fa9e4066Sahrens zfsvfs_t zfsvfs; 147114843421SMatthew Ahrens uint64_t moid, obj, version; 1472de8267e0Stimh uint64_t sense = ZFS_CASE_SENSITIVE; 1473de8267e0Stimh uint64_t norm = 0; 1474de8267e0Stimh nvpair_t *elem; 1475fa9e4066Sahrens int error; 1476fa9e4066Sahrens znode_t *rootzp = NULL; 1477fa9e4066Sahrens vnode_t *vp; 1478fa9e4066Sahrens vattr_t vattr; 14794ccbb6e7Sahrens znode_t *zp; 148089459e17SMark Shellenbaum zfs_acl_ids_t acl_ids; 1481fa9e4066Sahrens 1482fa9e4066Sahrens /* 1483fa9e4066Sahrens * First attempt to create master node. 1484fa9e4066Sahrens */ 1485ea8dc4b6Seschrock /* 1486ea8dc4b6Seschrock * In an empty objset, there are no blocks to read and thus 1487ea8dc4b6Seschrock * there can be no i/o errors (which we assert below). 1488ea8dc4b6Seschrock */ 1489fa9e4066Sahrens moid = MASTER_NODE_OBJ; 1490fa9e4066Sahrens error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE, 1491fa9e4066Sahrens DMU_OT_NONE, 0, tx); 1492fa9e4066Sahrens ASSERT(error == 0); 1493fa9e4066Sahrens 1494fa9e4066Sahrens /* 1495fa9e4066Sahrens * Set starting attributes. 1496fa9e4066Sahrens */ 149714843421SMatthew Ahrens if (spa_version(dmu_objset_spa(os)) >= SPA_VERSION_USERSPACE) 1498088f3894Sahrens version = ZPL_VERSION; 149914843421SMatthew Ahrens else if (spa_version(dmu_objset_spa(os)) >= SPA_VERSION_FUID) 150014843421SMatthew Ahrens version = ZPL_VERSION_USERSPACE - 1; 1501088f3894Sahrens else 1502088f3894Sahrens version = ZPL_VERSION_FUID - 1; 1503de8267e0Stimh elem = NULL; 1504de8267e0Stimh while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) { 1505de8267e0Stimh /* For the moment we expect all zpl props to be uint64_ts */ 1506de8267e0Stimh uint64_t val; 1507de8267e0Stimh char *name; 1508de8267e0Stimh 1509de8267e0Stimh ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64); 1510b3d911cbStimh VERIFY(nvpair_value_uint64(elem, &val) == 0); 1511de8267e0Stimh name = nvpair_name(elem); 1512de8267e0Stimh if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) { 151314843421SMatthew Ahrens if (val < version) 151414843421SMatthew Ahrens version = val; 1515de8267e0Stimh } else { 1516de8267e0Stimh error = zap_update(os, moid, name, 8, 1, &val, tx); 1517de8267e0Stimh } 1518de8267e0Stimh ASSERT(error == 0); 1519de8267e0Stimh if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0) 1520de8267e0Stimh norm = val; 1521de8267e0Stimh else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0) 1522de8267e0Stimh sense = val; 1523de8267e0Stimh } 1524de8267e0Stimh ASSERT(version != 0); 152514843421SMatthew Ahrens error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx); 1526fa9e4066Sahrens 1527fa9e4066Sahrens /* 1528fa9e4066Sahrens * Create a delete queue. 1529fa9e4066Sahrens */ 153014843421SMatthew Ahrens obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx); 1531fa9e4066Sahrens 153214843421SMatthew Ahrens error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx); 1533fa9e4066Sahrens ASSERT(error == 0); 1534fa9e4066Sahrens 1535fa9e4066Sahrens /* 1536fa9e4066Sahrens * Create root znode. Create minimal znode/vnode/zfsvfs 1537fa9e4066Sahrens * to allow zfs_mknode to work. 1538fa9e4066Sahrens */ 1539fa9e4066Sahrens vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE; 1540fa9e4066Sahrens vattr.va_type = VDIR; 1541fa9e4066Sahrens vattr.va_mode = S_IFDIR|0755; 1542ecd6cf80Smarks vattr.va_uid = crgetuid(cr); 1543ecd6cf80Smarks vattr.va_gid = crgetgid(cr); 1544fa9e4066Sahrens 1545fa9e4066Sahrens rootzp = kmem_cache_alloc(znode_cache, KM_SLEEP); 1546893a6d32Sahrens rootzp->z_unlinked = 0; 1547fa9e4066Sahrens rootzp->z_atime_dirty = 0; 1548fa9e4066Sahrens 1549fa9e4066Sahrens vp = ZTOV(rootzp); 1550fa9e4066Sahrens vn_reinit(vp); 1551fa9e4066Sahrens vp->v_type = VDIR; 1552fa9e4066Sahrens 1553fa9e4066Sahrens bzero(&zfsvfs, sizeof (zfsvfs_t)); 1554fa9e4066Sahrens 1555fa9e4066Sahrens zfsvfs.z_os = os; 1556fa9e4066Sahrens zfsvfs.z_parent = &zfsvfs; 1557da6c28aaSamw zfsvfs.z_version = version; 1558da6c28aaSamw zfsvfs.z_use_fuids = USE_FUIDS(version, os); 1559da6c28aaSamw zfsvfs.z_norm = norm; 1560de8267e0Stimh /* 1561de8267e0Stimh * Fold case on file systems that are always or sometimes case 1562de8267e0Stimh * insensitive. 1563de8267e0Stimh */ 1564de8267e0Stimh if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED) 1565de8267e0Stimh zfsvfs.z_norm |= U8_TEXTPREP_TOUPPER; 1566fa9e4066Sahrens 1567fa9e4066Sahrens mutex_init(&zfsvfs.z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); 1568fa9e4066Sahrens list_create(&zfsvfs.z_all_znodes, sizeof (znode_t), 1569fa9e4066Sahrens offsetof(znode_t, z_link_node)); 1570fa9e4066Sahrens 1571b5fca8f8Stomee ASSERT(!POINTER_IS_VALID(rootzp->z_zfsvfs)); 1572b5fca8f8Stomee rootzp->z_zfsvfs = &zfsvfs; 157389459e17SMark Shellenbaum VERIFY(0 == zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr, 157489459e17SMark Shellenbaum cr, NULL, &acl_ids)); 157589459e17SMark Shellenbaum zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, 0, &acl_ids); 1576874395d5Smaybee ASSERT3P(zp, ==, rootzp); 1577b5fca8f8Stomee ASSERT(!vn_in_dnlc(ZTOV(rootzp))); /* not valid to move */ 15784ccbb6e7Sahrens error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx); 1579fa9e4066Sahrens ASSERT(error == 0); 158089459e17SMark Shellenbaum zfs_acl_ids_free(&acl_ids); 1581b5fca8f8Stomee POINTER_INVALIDATE(&rootzp->z_zfsvfs); 1582fa9e4066Sahrens 1583fa9e4066Sahrens ZTOV(rootzp)->v_count = 0; 1584874395d5Smaybee dmu_buf_rele(rootzp->z_dbuf, NULL); 1585874395d5Smaybee rootzp->z_dbuf = NULL; 1586fa9e4066Sahrens kmem_cache_free(znode_cache, rootzp); 1587743a77edSAlan Wright 1588743a77edSAlan Wright /* 1589743a77edSAlan Wright * Create shares directory 1590743a77edSAlan Wright */ 1591743a77edSAlan Wright 1592743a77edSAlan Wright error = zfs_create_share_dir(&zfsvfs, tx); 159389459e17SMark Shellenbaum 1594743a77edSAlan Wright ASSERT(error == 0); 1595fa9e4066Sahrens } 159655434c77Sek 1597da6c28aaSamw #endif /* _KERNEL */ 159855434c77Sek /* 159955434c77Sek * Given an object number, return its parent object number and whether 160055434c77Sek * or not the object is an extended attribute directory. 160155434c77Sek */ 160255434c77Sek static int 160355434c77Sek zfs_obj_to_pobj(objset_t *osp, uint64_t obj, uint64_t *pobjp, int *is_xattrdir) 160455434c77Sek { 160555434c77Sek dmu_buf_t *db; 160655434c77Sek dmu_object_info_t doi; 160755434c77Sek znode_phys_t *zp; 160855434c77Sek int error; 160955434c77Sek 161055434c77Sek if ((error = dmu_bonus_hold(osp, obj, FTAG, &db)) != 0) 161155434c77Sek return (error); 161255434c77Sek 161355434c77Sek dmu_object_info_from_db(db, &doi); 161455434c77Sek if (doi.doi_bonus_type != DMU_OT_ZNODE || 161555434c77Sek doi.doi_bonus_size < sizeof (znode_phys_t)) { 161655434c77Sek dmu_buf_rele(db, FTAG); 161755434c77Sek return (EINVAL); 161855434c77Sek } 161955434c77Sek 162055434c77Sek zp = db->db_data; 162155434c77Sek *pobjp = zp->zp_parent; 162255434c77Sek *is_xattrdir = ((zp->zp_flags & ZFS_XATTR) != 0) && 162355434c77Sek S_ISDIR(zp->zp_mode); 162455434c77Sek dmu_buf_rele(db, FTAG); 162555434c77Sek 162655434c77Sek return (0); 162755434c77Sek } 162855434c77Sek 162955434c77Sek int 163055434c77Sek zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len) 163155434c77Sek { 163255434c77Sek char *path = buf + len - 1; 163355434c77Sek int error; 163455434c77Sek 163555434c77Sek *path = '\0'; 163655434c77Sek 163755434c77Sek for (;;) { 163855434c77Sek uint64_t pobj; 163955434c77Sek char component[MAXNAMELEN + 2]; 164055434c77Sek size_t complen; 164155434c77Sek int is_xattrdir; 164255434c77Sek 164355434c77Sek if ((error = zfs_obj_to_pobj(osp, obj, &pobj, 164455434c77Sek &is_xattrdir)) != 0) 164555434c77Sek break; 164655434c77Sek 164755434c77Sek if (pobj == obj) { 164855434c77Sek if (path[0] != '/') 164955434c77Sek *--path = '/'; 165055434c77Sek break; 165155434c77Sek } 165255434c77Sek 165355434c77Sek component[0] = '/'; 165455434c77Sek if (is_xattrdir) { 165555434c77Sek (void) sprintf(component + 1, "<xattrdir>"); 165655434c77Sek } else { 1657e7437265Sahrens error = zap_value_search(osp, pobj, obj, 1658e7437265Sahrens ZFS_DIRENT_OBJ(-1ULL), component + 1); 165955434c77Sek if (error != 0) 166055434c77Sek break; 166155434c77Sek } 166255434c77Sek 166355434c77Sek complen = strlen(component); 166455434c77Sek path -= complen; 166555434c77Sek ASSERT(path >= buf); 166655434c77Sek bcopy(component, path, complen); 166755434c77Sek obj = pobj; 166855434c77Sek } 166955434c77Sek 167055434c77Sek if (error == 0) 167155434c77Sek (void) memmove(buf, path, buf + len - path); 167255434c77Sek return (error); 167355434c77Sek } 1674