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. 23249643Smm * Copyright (c) 2013 by Delphix. All rights reserved. 24168404Spjd */ 25168404Spjd 26169195Spjd/* Portions Copyright 2007 Jeremy Teo */ 27226943Smm/* Portions Copyright 2011 Martin Matuska <mm@FreeBSD.org> */ 28169195Spjd 29168404Spjd#ifdef _KERNEL 30168404Spjd#include <sys/types.h> 31168404Spjd#include <sys/param.h> 32168404Spjd#include <sys/time.h> 33168404Spjd#include <sys/systm.h> 34168404Spjd#include <sys/sysmacros.h> 35168404Spjd#include <sys/resource.h> 36168404Spjd#include <sys/mntent.h> 37185029Spjd#include <sys/u8_textprep.h> 38185029Spjd#include <sys/dsl_dataset.h> 39168404Spjd#include <sys/vfs.h> 40168404Spjd#include <sys/vnode.h> 41168404Spjd#include <sys/file.h> 42168404Spjd#include <sys/kmem.h> 43168404Spjd#include <sys/errno.h> 44168404Spjd#include <sys/unistd.h> 45168404Spjd#include <sys/atomic.h> 46168404Spjd#include <sys/zfs_dir.h> 47168404Spjd#include <sys/zfs_acl.h> 48168404Spjd#include <sys/zfs_ioctl.h> 49168404Spjd#include <sys/zfs_rlock.h> 50185029Spjd#include <sys/zfs_fuid.h> 51219089Spjd#include <sys/dnode.h> 52168404Spjd#include <sys/fs/zfs.h> 53185029Spjd#include <sys/kidmap.h> 54168404Spjd#endif /* _KERNEL */ 55168404Spjd 56168404Spjd#include <sys/dmu.h> 57168404Spjd#include <sys/refcount.h> 58168404Spjd#include <sys/stat.h> 59168404Spjd#include <sys/zap.h> 60168404Spjd#include <sys/zfs_znode.h> 61219089Spjd#include <sys/sa.h> 62219089Spjd#include <sys/zfs_sa.h> 63219089Spjd#include <sys/zfs_stat.h> 64168404Spjd#include <sys/refcount.h> 65168404Spjd 66185029Spjd#include "zfs_prop.h" 67219089Spjd#include "zfs_comutil.h" 68185029Spjd 69173268Slulf/* Used by fstat(1). */ 70173268SlulfSYSCTL_INT(_debug_sizeof, OID_AUTO, znode, CTLFLAG_RD, 0, sizeof(znode_t), 71173268Slulf "sizeof(znode_t)"); 72173268Slulf 73168404Spjd/* 74185029Spjd * Define ZNODE_STATS to turn on statistic gathering. By default, it is only 75185029Spjd * turned on when DEBUG is also defined. 76185029Spjd */ 77185029Spjd#ifdef DEBUG 78185029Spjd#define ZNODE_STATS 79185029Spjd#endif /* DEBUG */ 80185029Spjd 81185029Spjd#ifdef ZNODE_STATS 82185029Spjd#define ZNODE_STAT_ADD(stat) ((stat)++) 83185029Spjd#else 84185029Spjd#define ZNODE_STAT_ADD(stat) /* nothing */ 85185029Spjd#endif /* ZNODE_STATS */ 86185029Spjd 87185029Spjd/* 88168404Spjd * Functions needed for userland (ie: libzpool) are not put under 89168404Spjd * #ifdef_KERNEL; the rest of the functions have dependencies 90168404Spjd * (such as VFS logic) that will not compile easily in userland. 91168404Spjd */ 92168404Spjd#ifdef _KERNEL 93210470Smm/* 94210470Smm * Needed to close a small window in zfs_znode_move() that allows the zfsvfs to 95210470Smm * be freed before it can be safely accessed. 96210470Smm */ 97210470Smmkrwlock_t zfsvfs_lock; 98210470Smm 99185029Spjdstatic kmem_cache_t *znode_cache = NULL; 100168404Spjd 101168404Spjd/*ARGSUSED*/ 102168404Spjdstatic void 103185029Spjdznode_evict_error(dmu_buf_t *dbuf, void *user_ptr) 104168404Spjd{ 105185029Spjd /* 106185029Spjd * We should never drop all dbuf refs without first clearing 107185029Spjd * the eviction callback. 108185029Spjd */ 109185029Spjd panic("evicting znode %p\n", user_ptr); 110168404Spjd} 111168404Spjd 112168404Spjdextern struct vop_vector zfs_vnodeops; 113168404Spjdextern struct vop_vector zfs_fifoops; 114209962Smmextern struct vop_vector zfs_shareops; 115168404Spjd 116168404Spjdstatic int 117185029Spjdzfs_znode_cache_constructor(void *buf, void *arg, int kmflags) 118168404Spjd{ 119168404Spjd znode_t *zp = buf; 120168404Spjd 121185029Spjd POINTER_INVALIDATE(&zp->z_zfsvfs); 122185029Spjd 123185029Spjd list_link_init(&zp->z_link_node); 124185029Spjd 125168404Spjd mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL); 126168404Spjd rw_init(&zp->z_parent_lock, NULL, RW_DEFAULT, NULL); 127168404Spjd rw_init(&zp->z_name_lock, NULL, RW_DEFAULT, NULL); 128168404Spjd mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL); 129168404Spjd 130168404Spjd mutex_init(&zp->z_range_lock, NULL, MUTEX_DEFAULT, NULL); 131168404Spjd avl_create(&zp->z_range_avl, zfs_range_compare, 132168404Spjd sizeof (rl_t), offsetof(rl_t, r_node)); 133168404Spjd 134185029Spjd zp->z_dirlocks = NULL; 135211932Smm zp->z_acl_cached = NULL; 136254694Savg zp->z_vnode = NULL; 137219089Spjd zp->z_moved = 0; 138168404Spjd return (0); 139168404Spjd} 140168404Spjd 141168404Spjd/*ARGSUSED*/ 142168404Spjdstatic void 143185029Spjdzfs_znode_cache_destructor(void *buf, void *arg) 144168404Spjd{ 145168404Spjd znode_t *zp = buf; 146168404Spjd 147185029Spjd ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs)); 148185029Spjd ASSERT(ZTOV(zp) == NULL); 149185029Spjd vn_free(ZTOV(zp)); 150185029Spjd ASSERT(!list_link_active(&zp->z_link_node)); 151168404Spjd mutex_destroy(&zp->z_lock); 152168404Spjd rw_destroy(&zp->z_parent_lock); 153168404Spjd rw_destroy(&zp->z_name_lock); 154168404Spjd mutex_destroy(&zp->z_acl_lock); 155185029Spjd avl_destroy(&zp->z_range_avl); 156168404Spjd mutex_destroy(&zp->z_range_lock); 157168404Spjd 158185029Spjd ASSERT(zp->z_dirlocks == NULL); 159211932Smm ASSERT(zp->z_acl_cached == NULL); 160168404Spjd} 161168404Spjd 162185029Spjd#ifdef ZNODE_STATS 163185029Spjdstatic struct { 164185029Spjd uint64_t zms_zfsvfs_invalid; 165210470Smm uint64_t zms_zfsvfs_recheck1; 166185029Spjd uint64_t zms_zfsvfs_unmounted; 167210470Smm uint64_t zms_zfsvfs_recheck2; 168185029Spjd uint64_t zms_obj_held; 169185029Spjd uint64_t zms_vnode_locked; 170185029Spjd uint64_t zms_not_only_dnlc; 171185029Spjd} znode_move_stats; 172185029Spjd#endif /* ZNODE_STATS */ 173185029Spjd 174219089Spjd#ifdef sun 175185029Spjdstatic void 176185029Spjdzfs_znode_move_impl(znode_t *ozp, znode_t *nzp) 177185029Spjd{ 178185029Spjd vnode_t *vp; 179185029Spjd 180185029Spjd /* Copy fields. */ 181185029Spjd nzp->z_zfsvfs = ozp->z_zfsvfs; 182185029Spjd 183185029Spjd /* Swap vnodes. */ 184185029Spjd vp = nzp->z_vnode; 185185029Spjd nzp->z_vnode = ozp->z_vnode; 186185029Spjd ozp->z_vnode = vp; /* let destructor free the overwritten vnode */ 187185029Spjd ZTOV(ozp)->v_data = ozp; 188185029Spjd ZTOV(nzp)->v_data = nzp; 189185029Spjd 190185029Spjd nzp->z_id = ozp->z_id; 191185029Spjd ASSERT(ozp->z_dirlocks == NULL); /* znode not in use */ 192185029Spjd ASSERT(avl_numnodes(&ozp->z_range_avl) == 0); 193185029Spjd nzp->z_unlinked = ozp->z_unlinked; 194185029Spjd nzp->z_atime_dirty = ozp->z_atime_dirty; 195185029Spjd nzp->z_zn_prefetch = ozp->z_zn_prefetch; 196185029Spjd nzp->z_blksz = ozp->z_blksz; 197185029Spjd nzp->z_seq = ozp->z_seq; 198185029Spjd nzp->z_mapcnt = ozp->z_mapcnt; 199185029Spjd nzp->z_gen = ozp->z_gen; 200185029Spjd nzp->z_sync_cnt = ozp->z_sync_cnt; 201219089Spjd nzp->z_is_sa = ozp->z_is_sa; 202219089Spjd nzp->z_sa_hdl = ozp->z_sa_hdl; 203219089Spjd bcopy(ozp->z_atime, nzp->z_atime, sizeof (uint64_t) * 2); 204219089Spjd nzp->z_links = ozp->z_links; 205219089Spjd nzp->z_size = ozp->z_size; 206219089Spjd nzp->z_pflags = ozp->z_pflags; 207219089Spjd nzp->z_uid = ozp->z_uid; 208219089Spjd nzp->z_gid = ozp->z_gid; 209219089Spjd nzp->z_mode = ozp->z_mode; 210185029Spjd 211211932Smm /* 212211932Smm * Since this is just an idle znode and kmem is already dealing with 213211932Smm * memory pressure, release any cached ACL. 214211932Smm */ 215211932Smm if (ozp->z_acl_cached) { 216211932Smm zfs_acl_free(ozp->z_acl_cached); 217211932Smm ozp->z_acl_cached = NULL; 218211932Smm } 219211932Smm 220219089Spjd sa_set_userp(nzp->z_sa_hdl, nzp); 221185029Spjd 222185029Spjd /* 223185029Spjd * Invalidate the original znode by clearing fields that provide a 224185029Spjd * pointer back to the znode. Set the low bit of the vfs pointer to 225185029Spjd * ensure that zfs_znode_move() recognizes the znode as invalid in any 226185029Spjd * subsequent callback. 227185029Spjd */ 228219089Spjd ozp->z_sa_hdl = NULL; 229185029Spjd POINTER_INVALIDATE(&ozp->z_zfsvfs); 230219089Spjd 231219089Spjd /* 232219089Spjd * Mark the znode. 233219089Spjd */ 234219089Spjd nzp->z_moved = 1; 235219089Spjd ozp->z_moved = (uint8_t)-1; 236185029Spjd} 237185029Spjd 238185029Spjd/*ARGSUSED*/ 239185029Spjdstatic kmem_cbrc_t 240185029Spjdzfs_znode_move(void *buf, void *newbuf, size_t size, void *arg) 241185029Spjd{ 242185029Spjd znode_t *ozp = buf, *nzp = newbuf; 243185029Spjd zfsvfs_t *zfsvfs; 244185029Spjd vnode_t *vp; 245185029Spjd 246185029Spjd /* 247185029Spjd * The znode is on the file system's list of known znodes if the vfs 248185029Spjd * pointer is valid. We set the low bit of the vfs pointer when freeing 249185029Spjd * the znode to invalidate it, and the memory patterns written by kmem 250185029Spjd * (baddcafe and deadbeef) set at least one of the two low bits. A newly 251185029Spjd * created znode sets the vfs pointer last of all to indicate that the 252185029Spjd * znode is known and in a valid state to be moved by this function. 253185029Spjd */ 254185029Spjd zfsvfs = ozp->z_zfsvfs; 255185029Spjd if (!POINTER_IS_VALID(zfsvfs)) { 256185029Spjd ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_invalid); 257185029Spjd return (KMEM_CBRC_DONT_KNOW); 258185029Spjd } 259185029Spjd 260185029Spjd /* 261210470Smm * Close a small window in which it's possible that the filesystem could 262210470Smm * be unmounted and freed, and zfsvfs, though valid in the previous 263210470Smm * statement, could point to unrelated memory by the time we try to 264210470Smm * prevent the filesystem from being unmounted. 265185029Spjd */ 266210470Smm rw_enter(&zfsvfs_lock, RW_WRITER); 267210470Smm if (zfsvfs != ozp->z_zfsvfs) { 268210470Smm rw_exit(&zfsvfs_lock); 269210470Smm ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_recheck1); 270210470Smm return (KMEM_CBRC_DONT_KNOW); 271210470Smm } 272210470Smm 273210470Smm /* 274210470Smm * If the znode is still valid, then so is the file system. We know that 275210470Smm * no valid file system can be freed while we hold zfsvfs_lock, so we 276210470Smm * can safely ensure that the filesystem is not and will not be 277210470Smm * unmounted. The next statement is equivalent to ZFS_ENTER(). 278210470Smm */ 279209962Smm rrw_enter(&zfsvfs->z_teardown_lock, RW_READER, FTAG); 280209962Smm if (zfsvfs->z_unmounted) { 281209962Smm ZFS_EXIT(zfsvfs); 282210470Smm rw_exit(&zfsvfs_lock); 283185029Spjd ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_unmounted); 284185029Spjd return (KMEM_CBRC_DONT_KNOW); 285185029Spjd } 286210470Smm rw_exit(&zfsvfs_lock); 287185029Spjd 288185029Spjd mutex_enter(&zfsvfs->z_znodes_lock); 289185029Spjd /* 290185029Spjd * Recheck the vfs pointer in case the znode was removed just before 291185029Spjd * acquiring the lock. 292185029Spjd */ 293185029Spjd if (zfsvfs != ozp->z_zfsvfs) { 294185029Spjd mutex_exit(&zfsvfs->z_znodes_lock); 295185029Spjd ZFS_EXIT(zfsvfs); 296210470Smm ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_recheck2); 297185029Spjd return (KMEM_CBRC_DONT_KNOW); 298185029Spjd } 299185029Spjd 300185029Spjd /* 301185029Spjd * At this point we know that as long as we hold z_znodes_lock, the 302185029Spjd * znode cannot be freed and fields within the znode can be safely 303185029Spjd * accessed. Now, prevent a race with zfs_zget(). 304185029Spjd */ 305185029Spjd if (ZFS_OBJ_HOLD_TRYENTER(zfsvfs, ozp->z_id) == 0) { 306185029Spjd mutex_exit(&zfsvfs->z_znodes_lock); 307185029Spjd ZFS_EXIT(zfsvfs); 308185029Spjd ZNODE_STAT_ADD(znode_move_stats.zms_obj_held); 309185029Spjd return (KMEM_CBRC_LATER); 310185029Spjd } 311185029Spjd 312185029Spjd vp = ZTOV(ozp); 313185029Spjd if (mutex_tryenter(&vp->v_lock) == 0) { 314185029Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id); 315185029Spjd mutex_exit(&zfsvfs->z_znodes_lock); 316185029Spjd ZFS_EXIT(zfsvfs); 317185029Spjd ZNODE_STAT_ADD(znode_move_stats.zms_vnode_locked); 318185029Spjd return (KMEM_CBRC_LATER); 319185029Spjd } 320185029Spjd 321185029Spjd /* Only move znodes that are referenced _only_ by the DNLC. */ 322185029Spjd if (vp->v_count != 1 || !vn_in_dnlc(vp)) { 323185029Spjd mutex_exit(&vp->v_lock); 324185029Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id); 325185029Spjd mutex_exit(&zfsvfs->z_znodes_lock); 326185029Spjd ZFS_EXIT(zfsvfs); 327185029Spjd ZNODE_STAT_ADD(znode_move_stats.zms_not_only_dnlc); 328185029Spjd return (KMEM_CBRC_LATER); 329185029Spjd } 330185029Spjd 331185029Spjd /* 332185029Spjd * The znode is known and in a valid state to move. We're holding the 333185029Spjd * locks needed to execute the critical section. 334185029Spjd */ 335185029Spjd zfs_znode_move_impl(ozp, nzp); 336185029Spjd mutex_exit(&vp->v_lock); 337185029Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id); 338185029Spjd 339185029Spjd list_link_replace(&ozp->z_link_node, &nzp->z_link_node); 340185029Spjd mutex_exit(&zfsvfs->z_znodes_lock); 341185029Spjd ZFS_EXIT(zfsvfs); 342185029Spjd 343185029Spjd return (KMEM_CBRC_YES); 344185029Spjd} 345185029Spjd#endif /* sun */ 346185029Spjd 347168404Spjdvoid 348168404Spjdzfs_znode_init(void) 349168404Spjd{ 350168404Spjd /* 351168404Spjd * Initialize zcache 352168404Spjd */ 353210470Smm rw_init(&zfsvfs_lock, NULL, RW_DEFAULT, NULL); 354168404Spjd ASSERT(znode_cache == NULL); 355168404Spjd znode_cache = kmem_cache_create("zfs_znode_cache", 356254694Savg sizeof (znode_t), 0, zfs_znode_cache_constructor, 357168404Spjd zfs_znode_cache_destructor, NULL, NULL, NULL, 0); 358185029Spjd kmem_cache_set_move(znode_cache, zfs_znode_move); 359168404Spjd} 360168404Spjd 361168404Spjdvoid 362168404Spjdzfs_znode_fini(void) 363168404Spjd{ 364219089Spjd#ifdef sun 365168404Spjd /* 366219089Spjd * Cleanup vfs & vnode ops 367219089Spjd */ 368219089Spjd zfs_remove_op_tables(); 369219089Spjd#endif /* sun */ 370219089Spjd 371219089Spjd /* 372168404Spjd * Cleanup zcache 373168404Spjd */ 374168404Spjd if (znode_cache) 375168404Spjd kmem_cache_destroy(znode_cache); 376168404Spjd znode_cache = NULL; 377210470Smm rw_destroy(&zfsvfs_lock); 378168404Spjd} 379168404Spjd 380219089Spjd#ifdef sun 381219089Spjdstruct vnodeops *zfs_dvnodeops; 382219089Spjdstruct vnodeops *zfs_fvnodeops; 383219089Spjdstruct vnodeops *zfs_symvnodeops; 384219089Spjdstruct vnodeops *zfs_xdvnodeops; 385219089Spjdstruct vnodeops *zfs_evnodeops; 386219089Spjdstruct vnodeops *zfs_sharevnodeops; 387219089Spjd 388219089Spjdvoid 389219089Spjdzfs_remove_op_tables() 390219089Spjd{ 391219089Spjd /* 392219089Spjd * Remove vfs ops 393219089Spjd */ 394219089Spjd ASSERT(zfsfstype); 395219089Spjd (void) vfs_freevfsops_by_type(zfsfstype); 396219089Spjd zfsfstype = 0; 397219089Spjd 398219089Spjd /* 399219089Spjd * Remove vnode ops 400219089Spjd */ 401219089Spjd if (zfs_dvnodeops) 402219089Spjd vn_freevnodeops(zfs_dvnodeops); 403219089Spjd if (zfs_fvnodeops) 404219089Spjd vn_freevnodeops(zfs_fvnodeops); 405219089Spjd if (zfs_symvnodeops) 406219089Spjd vn_freevnodeops(zfs_symvnodeops); 407219089Spjd if (zfs_xdvnodeops) 408219089Spjd vn_freevnodeops(zfs_xdvnodeops); 409219089Spjd if (zfs_evnodeops) 410219089Spjd vn_freevnodeops(zfs_evnodeops); 411219089Spjd if (zfs_sharevnodeops) 412219089Spjd vn_freevnodeops(zfs_sharevnodeops); 413219089Spjd 414219089Spjd zfs_dvnodeops = NULL; 415219089Spjd zfs_fvnodeops = NULL; 416219089Spjd zfs_symvnodeops = NULL; 417219089Spjd zfs_xdvnodeops = NULL; 418219089Spjd zfs_evnodeops = NULL; 419219089Spjd zfs_sharevnodeops = NULL; 420219089Spjd} 421219089Spjd 422219089Spjdextern const fs_operation_def_t zfs_dvnodeops_template[]; 423219089Spjdextern const fs_operation_def_t zfs_fvnodeops_template[]; 424219089Spjdextern const fs_operation_def_t zfs_xdvnodeops_template[]; 425219089Spjdextern const fs_operation_def_t zfs_symvnodeops_template[]; 426219089Spjdextern const fs_operation_def_t zfs_evnodeops_template[]; 427219089Spjdextern const fs_operation_def_t zfs_sharevnodeops_template[]; 428219089Spjd 429168404Spjdint 430219089Spjdzfs_create_op_tables() 431219089Spjd{ 432219089Spjd int error; 433219089Spjd 434219089Spjd /* 435219089Spjd * zfs_dvnodeops can be set if mod_remove() calls mod_installfs() 436219089Spjd * due to a failure to remove the the 2nd modlinkage (zfs_modldrv). 437219089Spjd * In this case we just return as the ops vectors are already set up. 438219089Spjd */ 439219089Spjd if (zfs_dvnodeops) 440219089Spjd return (0); 441219089Spjd 442219089Spjd error = vn_make_ops(MNTTYPE_ZFS, zfs_dvnodeops_template, 443219089Spjd &zfs_dvnodeops); 444219089Spjd if (error) 445219089Spjd return (error); 446219089Spjd 447219089Spjd error = vn_make_ops(MNTTYPE_ZFS, zfs_fvnodeops_template, 448219089Spjd &zfs_fvnodeops); 449219089Spjd if (error) 450219089Spjd return (error); 451219089Spjd 452219089Spjd error = vn_make_ops(MNTTYPE_ZFS, zfs_symvnodeops_template, 453219089Spjd &zfs_symvnodeops); 454219089Spjd if (error) 455219089Spjd return (error); 456219089Spjd 457219089Spjd error = vn_make_ops(MNTTYPE_ZFS, zfs_xdvnodeops_template, 458219089Spjd &zfs_xdvnodeops); 459219089Spjd if (error) 460219089Spjd return (error); 461219089Spjd 462219089Spjd error = vn_make_ops(MNTTYPE_ZFS, zfs_evnodeops_template, 463219089Spjd &zfs_evnodeops); 464219089Spjd if (error) 465219089Spjd return (error); 466219089Spjd 467219089Spjd error = vn_make_ops(MNTTYPE_ZFS, zfs_sharevnodeops_template, 468219089Spjd &zfs_sharevnodeops); 469219089Spjd 470219089Spjd return (error); 471219089Spjd} 472219089Spjd#endif /* sun */ 473219089Spjd 474219089Spjdint 475209962Smmzfs_create_share_dir(zfsvfs_t *zfsvfs, dmu_tx_t *tx) 476168404Spjd{ 477209962Smm zfs_acl_ids_t acl_ids; 478209962Smm vattr_t vattr; 479209962Smm znode_t *sharezp; 480209962Smm znode_t *zp; 481209962Smm int error; 482168404Spjd 483209962Smm vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE; 484209962Smm vattr.va_type = VDIR; 485209962Smm vattr.va_mode = S_IFDIR|0555; 486209962Smm vattr.va_uid = crgetuid(kcred); 487209962Smm vattr.va_gid = crgetgid(kcred); 488168404Spjd 489209962Smm sharezp = kmem_cache_alloc(znode_cache, KM_SLEEP); 490219089Spjd ASSERT(!POINTER_IS_VALID(sharezp->z_zfsvfs)); 491219089Spjd sharezp->z_moved = 0; 492209962Smm sharezp->z_unlinked = 0; 493209962Smm sharezp->z_atime_dirty = 0; 494209962Smm sharezp->z_zfsvfs = zfsvfs; 495219089Spjd sharezp->z_is_sa = zfsvfs->z_use_sa; 496168404Spjd 497209962Smm VERIFY(0 == zfs_acl_ids_create(sharezp, IS_ROOT_NODE, &vattr, 498209962Smm kcred, NULL, &acl_ids)); 499219089Spjd zfs_mknode(sharezp, &vattr, tx, kcred, IS_ROOT_NODE, &zp, &acl_ids); 500209962Smm ASSERT3P(zp, ==, sharezp); 501209962Smm POINTER_INVALIDATE(&sharezp->z_zfsvfs); 502209962Smm error = zap_add(zfsvfs->z_os, MASTER_NODE_OBJ, 503209962Smm ZFS_SHARES_DIR, 8, 1, &sharezp->z_id, tx); 504209962Smm zfsvfs->z_shares_dir = sharezp->z_id; 505168404Spjd 506209962Smm zfs_acl_ids_free(&acl_ids); 507219089Spjd sa_handle_destroy(sharezp->z_sa_hdl); 508209962Smm kmem_cache_free(znode_cache, sharezp); 509168404Spjd 510209962Smm return (error); 511168404Spjd} 512168404Spjd 513168404Spjd/* 514168404Spjd * define a couple of values we need available 515168404Spjd * for both 64 and 32 bit environments. 516168404Spjd */ 517168404Spjd#ifndef NBITSMINOR64 518168404Spjd#define NBITSMINOR64 32 519168404Spjd#endif 520168404Spjd#ifndef MAXMAJ64 521168404Spjd#define MAXMAJ64 0xffffffffUL 522168404Spjd#endif 523168404Spjd#ifndef MAXMIN64 524168404Spjd#define MAXMIN64 0xffffffffUL 525168404Spjd#endif 526168404Spjd 527168404Spjd/* 528168404Spjd * Create special expldev for ZFS private use. 529168404Spjd * Can't use standard expldev since it doesn't do 530168404Spjd * what we want. The standard expldev() takes a 531168404Spjd * dev32_t in LP64 and expands it to a long dev_t. 532168404Spjd * We need an interface that takes a dev32_t in ILP32 533168404Spjd * and expands it to a long dev_t. 534168404Spjd */ 535168404Spjdstatic uint64_t 536168404Spjdzfs_expldev(dev_t dev) 537168404Spjd{ 538187830Sed return (((uint64_t)major(dev) << NBITSMINOR64) | minor(dev)); 539168404Spjd} 540168404Spjd/* 541168404Spjd * Special cmpldev for ZFS private use. 542168404Spjd * Can't use standard cmpldev since it takes 543168404Spjd * a long dev_t and compresses it to dev32_t in 544168404Spjd * LP64. We need to do a compaction of a long dev_t 545168404Spjd * to a dev32_t in ILP32. 546168404Spjd */ 547168404Spjddev_t 548168404Spjdzfs_cmpldev(uint64_t dev) 549168404Spjd{ 550168958Spjd return (makedev((dev >> NBITSMINOR64), (dev & MAXMIN64))); 551168404Spjd} 552168404Spjd 553185029Spjdstatic void 554219089Spjdzfs_znode_sa_init(zfsvfs_t *zfsvfs, znode_t *zp, 555219089Spjd dmu_buf_t *db, dmu_object_type_t obj_type, sa_handle_t *sa_hdl) 556185029Spjd{ 557185029Spjd ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs) || (zfsvfs == zp->z_zfsvfs)); 558185029Spjd ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id))); 559185029Spjd 560185029Spjd mutex_enter(&zp->z_lock); 561185029Spjd 562219089Spjd ASSERT(zp->z_sa_hdl == NULL); 563211932Smm ASSERT(zp->z_acl_cached == NULL); 564219089Spjd if (sa_hdl == NULL) { 565219089Spjd VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, zp, 566219089Spjd SA_HDL_SHARED, &zp->z_sa_hdl)); 567219089Spjd } else { 568219089Spjd zp->z_sa_hdl = sa_hdl; 569219089Spjd sa_set_userp(sa_hdl, zp); 570219089Spjd } 571185029Spjd 572219089Spjd zp->z_is_sa = (obj_type == DMU_OT_SA) ? B_TRUE : B_FALSE; 573185029Spjd 574185029Spjd /* 575185029Spjd * Slap on VROOT if we are the root znode 576185029Spjd */ 577185029Spjd if (zp->z_id == zfsvfs->z_root) 578185029Spjd ZTOV(zp)->v_flag |= VROOT; 579185029Spjd 580185029Spjd mutex_exit(&zp->z_lock); 581185029Spjd vn_exists(ZTOV(zp)); 582185029Spjd} 583185029Spjd 584185029Spjdvoid 585185029Spjdzfs_znode_dmu_fini(znode_t *zp) 586185029Spjd{ 587185029Spjd ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zp->z_zfsvfs, zp->z_id)) || 588185029Spjd zp->z_unlinked || 589185029Spjd RW_WRITE_HELD(&zp->z_zfsvfs->z_teardown_inactive_lock)); 590219089Spjd 591219089Spjd sa_handle_destroy(zp->z_sa_hdl); 592219089Spjd zp->z_sa_hdl = NULL; 593185029Spjd} 594185029Spjd 595225153Spjdstatic void 596225153Spjdzfs_vnode_forget(vnode_t *vp) 597225153Spjd{ 598225153Spjd 599243488Savg /* copied from insmntque_stddtr */ 600225153Spjd vp->v_data = NULL; 601243488Savg vp->v_op = &dead_vnodeops; 602243488Savg vgone(vp); 603243488Savg vput(vp); 604225153Spjd} 605225153Spjd 606168404Spjd/* 607168404Spjd * Construct a new znode/vnode and intialize. 608168404Spjd * 609168404Spjd * This does not do a call to dmu_set_user() that is 610168404Spjd * up to the caller to do, in case you don't want to 611168404Spjd * return the znode 612168404Spjd */ 613168404Spjdstatic znode_t * 614219089Spjdzfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz, 615219089Spjd dmu_object_type_t obj_type, sa_handle_t *hdl) 616168404Spjd{ 617168404Spjd znode_t *zp; 618168404Spjd vnode_t *vp; 619219089Spjd uint64_t mode; 620219089Spjd uint64_t parent; 621219089Spjd sa_bulk_attr_t bulk[9]; 622219089Spjd int count = 0; 623254694Savg int error; 624168404Spjd 625168404Spjd zp = kmem_cache_alloc(znode_cache, KM_SLEEP); 626168404Spjd 627262111Savg KASSERT(curthread->td_vp_reserv > 0, 628262111Savg ("zfs_znode_alloc: getnewvnode without any vnodes reserved")); 629254694Savg error = getnewvnode("zfs", zfsvfs->z_parent->z_vfs, &zfs_vnodeops, &vp); 630254694Savg if (error != 0) { 631254694Savg kmem_cache_free(znode_cache, zp); 632254694Savg return (NULL); 633254694Savg } 634254694Savg zp->z_vnode = vp; 635254694Savg vp->v_data = zp; 636254694Savg 637168404Spjd ASSERT(zp->z_dirlocks == NULL); 638185029Spjd ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs)); 639219089Spjd zp->z_moved = 0; 640168404Spjd 641185029Spjd /* 642185029Spjd * Defer setting z_zfsvfs until the znode is ready to be a candidate for 643185029Spjd * the zfs_znode_move() callback. 644185029Spjd */ 645219089Spjd zp->z_sa_hdl = NULL; 646168404Spjd zp->z_unlinked = 0; 647168404Spjd zp->z_atime_dirty = 0; 648168404Spjd zp->z_mapcnt = 0; 649185029Spjd zp->z_id = db->db_object; 650168404Spjd zp->z_blksz = blksz; 651168404Spjd zp->z_seq = 0x7A4653; 652168404Spjd zp->z_sync_cnt = 0; 653168404Spjd 654185029Spjd vp = ZTOV(zp); 655168404Spjd 656219089Spjd zfs_znode_sa_init(zfsvfs, zp, db, obj_type, hdl); 657185029Spjd 658219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8); 659219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, &zp->z_gen, 8); 660219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, 661219089Spjd &zp->z_size, 8); 662219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL, 663219089Spjd &zp->z_links, 8); 664219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, 665219089Spjd &zp->z_pflags, 8); 666219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8); 667219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, 668219089Spjd &zp->z_atime, 16); 669219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, 670219089Spjd &zp->z_uid, 8); 671219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, 672219089Spjd &zp->z_gid, 8); 673185029Spjd 674219089Spjd if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0) { 675219089Spjd if (hdl == NULL) 676219089Spjd sa_handle_destroy(zp->z_sa_hdl); 677225153Spjd zfs_vnode_forget(vp); 678225153Spjd zp->z_vnode = NULL; 679219089Spjd kmem_cache_free(znode_cache, zp); 680219089Spjd return (NULL); 681219089Spjd } 682168404Spjd 683219089Spjd zp->z_mode = mode; 684219089Spjd 685219089Spjd vp->v_type = IFTOVT((mode_t)mode); 686219089Spjd 687168404Spjd switch (vp->v_type) { 688168404Spjd case VDIR: 689168404Spjd zp->z_zn_prefetch = B_TRUE; /* z_prefetch default is enabled */ 690168404Spjd break; 691224251Sdelphij#ifdef sun 692224231Sdelphij case VBLK: 693224231Sdelphij case VCHR: 694224231Sdelphij { 695224231Sdelphij uint64_t rdev; 696224231Sdelphij VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_RDEV(zfsvfs), 697224231Sdelphij &rdev, sizeof (rdev)) == 0); 698224231Sdelphij 699224251Sdelphij vp->v_rdev = zfs_cmpldev(rdev); 700224231Sdelphij } 701224231Sdelphij break; 702224251Sdelphij#endif /* sun */ 703168404Spjd case VFIFO: 704224252Sdelphij#ifdef sun 705224252Sdelphij case VSOCK: 706224252Sdelphij case VDOOR: 707224252Sdelphij#endif /* sun */ 708168404Spjd vp->v_op = &zfs_fifoops; 709168404Spjd break; 710219089Spjd case VREG: 711219089Spjd if (parent == zfsvfs->z_shares_dir) { 712219089Spjd ASSERT(zp->z_uid == 0 && zp->z_gid == 0); 713209962Smm vp->v_op = &zfs_shareops; 714209962Smm } 715209962Smm break; 716224252Sdelphij#ifdef sun 717224252Sdelphij case VLNK: 718224252Sdelphij vn_setops(vp, zfs_symvnodeops); 719224252Sdelphij break; 720224252Sdelphij default: 721224252Sdelphij vn_setops(vp, zfs_evnodeops); 722224252Sdelphij break; 723224252Sdelphij#endif /* sun */ 724168404Spjd } 725168404Spjd 726185029Spjd mutex_enter(&zfsvfs->z_znodes_lock); 727185029Spjd list_insert_tail(&zfsvfs->z_all_znodes, zp); 728185029Spjd membar_producer(); 729168404Spjd /* 730185029Spjd * Everything else must be valid before assigning z_zfsvfs makes the 731185029Spjd * znode eligible for zfs_znode_move(). 732168404Spjd */ 733185029Spjd zp->z_zfsvfs = zfsvfs; 734185029Spjd mutex_exit(&zfsvfs->z_znodes_lock); 735168404Spjd 736254694Savg /* 737254694Savg * Acquire vnode lock before making it available to the world. 738254694Savg */ 739254694Savg vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 740254694Savg VN_LOCK_AREC(vp); 741254694Savg if (vp->v_type != VFIFO) 742254694Savg VN_LOCK_ASHARE(vp); 743254694Savg 744168404Spjd VFS_HOLD(zfsvfs->z_vfs); 745185029Spjd return (zp); 746168404Spjd} 747168404Spjd 748219089Spjdstatic uint64_t empty_xattr; 749219089Spjdstatic uint64_t pad[4]; 750219089Spjdstatic zfs_acl_phys_t acl_phys; 751168404Spjd/* 752168404Spjd * Create a new DMU object to hold a zfs znode. 753168404Spjd * 754168404Spjd * IN: dzp - parent directory for new znode 755168404Spjd * vap - file attributes for new znode 756168404Spjd * tx - dmu transaction id for zap operations 757168404Spjd * cr - credentials of caller 758168404Spjd * flag - flags: 759168404Spjd * IS_ROOT_NODE - new object will be root 760168404Spjd * IS_XATTR - new object is an attribute 761185029Spjd * bonuslen - length of bonus buffer 762185029Spjd * setaclp - File/Dir initial ACL 763185029Spjd * fuidp - Tracks fuid allocation. 764168404Spjd * 765185029Spjd * OUT: zpp - allocated znode 766168404Spjd * 767168404Spjd */ 768168404Spjdvoid 769185029Spjdzfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr, 770219089Spjd uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids) 771168404Spjd{ 772219089Spjd uint64_t crtime[2], atime[2], mtime[2], ctime[2]; 773219089Spjd uint64_t mode, size, links, parent, pflags; 774219089Spjd uint64_t dzp_pflags = 0; 775219089Spjd uint64_t rdev = 0; 776219089Spjd zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 777185029Spjd dmu_buf_t *db; 778168404Spjd timestruc_t now; 779185029Spjd uint64_t gen, obj; 780168404Spjd int err; 781219089Spjd int bonuslen; 782219089Spjd sa_handle_t *sa_hdl; 783219089Spjd dmu_object_type_t obj_type; 784219089Spjd sa_bulk_attr_t sa_attrs[ZPL_END]; 785219089Spjd int cnt = 0; 786219089Spjd zfs_acl_locator_cb_t locate = { 0 }; 787168404Spjd 788168404Spjd ASSERT(vap && (vap->va_mask & (AT_TYPE|AT_MODE)) == (AT_TYPE|AT_MODE)); 789168404Spjd 790209962Smm if (zfsvfs->z_replay) { 791185029Spjd obj = vap->va_nodeid; 792168404Spjd now = vap->va_ctime; /* see zfs_replay_create() */ 793168404Spjd gen = vap->va_nblocks; /* ditto */ 794168404Spjd } else { 795185029Spjd obj = 0; 796168404Spjd gethrestime(&now); 797168404Spjd gen = dmu_tx_get_txg(tx); 798168404Spjd } 799168404Spjd 800219089Spjd obj_type = zfsvfs->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE; 801219089Spjd bonuslen = (obj_type == DMU_OT_SA) ? 802219089Spjd DN_MAX_BONUSLEN : ZFS_OLD_ZNODE_PHYS_SIZE; 803219089Spjd 804168404Spjd /* 805168404Spjd * Create a new DMU object. 806168404Spjd */ 807168404Spjd /* 808168404Spjd * There's currently no mechanism for pre-reading the blocks that will 809219089Spjd * be needed to allocate a new object, so we accept the small chance 810168404Spjd * that there will be an i/o error and we will fail one of the 811168404Spjd * assertions below. 812168404Spjd */ 813168404Spjd if (vap->va_type == VDIR) { 814209962Smm if (zfsvfs->z_replay) { 815263398Sdelphij VERIFY0(zap_create_claim_norm(zfsvfs->z_os, obj, 816185029Spjd zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS, 817263398Sdelphij obj_type, bonuslen, tx)); 818168404Spjd } else { 819185029Spjd obj = zap_create_norm(zfsvfs->z_os, 820185029Spjd zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS, 821219089Spjd obj_type, bonuslen, tx); 822168404Spjd } 823168404Spjd } else { 824209962Smm if (zfsvfs->z_replay) { 825263398Sdelphij VERIFY0(dmu_object_claim(zfsvfs->z_os, obj, 826168404Spjd DMU_OT_PLAIN_FILE_CONTENTS, 0, 827263398Sdelphij obj_type, bonuslen, tx)); 828168404Spjd } else { 829185029Spjd obj = dmu_object_alloc(zfsvfs->z_os, 830168404Spjd DMU_OT_PLAIN_FILE_CONTENTS, 0, 831219089Spjd obj_type, bonuslen, tx); 832168404Spjd } 833168404Spjd } 834207334Spjd 835207334Spjd ZFS_OBJ_HOLD_ENTER(zfsvfs, obj); 836219089Spjd VERIFY(0 == sa_buf_hold(zfsvfs->z_os, obj, NULL, &db)); 837168404Spjd 838168404Spjd /* 839168404Spjd * If this is the root, fix up the half-initialized parent pointer 840168404Spjd * to reference the just-allocated physical data area. 841168404Spjd */ 842168404Spjd if (flag & IS_ROOT_NODE) { 843185029Spjd dzp->z_id = obj; 844219089Spjd } else { 845219089Spjd dzp_pflags = dzp->z_pflags; 846168404Spjd } 847168404Spjd 848168404Spjd /* 849168404Spjd * If parent is an xattr, so am I. 850168404Spjd */ 851219089Spjd if (dzp_pflags & ZFS_XATTR) { 852168404Spjd flag |= IS_XATTR; 853168404Spjd } 854168404Spjd 855185029Spjd if (zfsvfs->z_use_fuids) 856219089Spjd pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED; 857219089Spjd else 858219089Spjd pflags = 0; 859185029Spjd 860168404Spjd if (vap->va_type == VDIR) { 861219089Spjd size = 2; /* contents ("." and "..") */ 862219089Spjd links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1; 863219089Spjd } else { 864219089Spjd size = links = 0; 865168404Spjd } 866168404Spjd 867219089Spjd if (vap->va_type == VBLK || vap->va_type == VCHR) { 868219089Spjd rdev = zfs_expldev(vap->va_rdev); 869219089Spjd } 870219089Spjd 871219089Spjd parent = dzp->z_id; 872219089Spjd mode = acl_ids->z_mode; 873168404Spjd if (flag & IS_XATTR) 874219089Spjd pflags |= ZFS_XATTR; 875168404Spjd 876219089Spjd /* 877219089Spjd * No execs denied will be deterimed when zfs_mode_compute() is called. 878219089Spjd */ 879219089Spjd pflags |= acl_ids->z_aclp->z_hints & 880219089Spjd (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT| 881219089Spjd ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED); 882168404Spjd 883219089Spjd ZFS_TIME_ENCODE(&now, crtime); 884219089Spjd ZFS_TIME_ENCODE(&now, ctime); 885168404Spjd 886168404Spjd if (vap->va_mask & AT_ATIME) { 887219089Spjd ZFS_TIME_ENCODE(&vap->va_atime, atime); 888168404Spjd } else { 889219089Spjd ZFS_TIME_ENCODE(&now, atime); 890168404Spjd } 891168404Spjd 892168404Spjd if (vap->va_mask & AT_MTIME) { 893219089Spjd ZFS_TIME_ENCODE(&vap->va_mtime, mtime); 894168404Spjd } else { 895219089Spjd ZFS_TIME_ENCODE(&now, mtime); 896168404Spjd } 897168404Spjd 898219089Spjd /* Now add in all of the "SA" attributes */ 899219089Spjd VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, NULL, SA_HDL_SHARED, 900219089Spjd &sa_hdl)); 901219089Spjd 902219089Spjd /* 903219089Spjd * Setup the array of attributes to be replaced/set on the new file 904219089Spjd * 905219089Spjd * order for DMU_OT_ZNODE is critical since it needs to be constructed 906219089Spjd * in the old znode_phys_t format. Don't change this ordering 907219089Spjd */ 908219089Spjd 909219089Spjd if (obj_type == DMU_OT_ZNODE) { 910219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs), 911219089Spjd NULL, &atime, 16); 912219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs), 913219089Spjd NULL, &mtime, 16); 914219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs), 915219089Spjd NULL, &ctime, 16); 916219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs), 917219089Spjd NULL, &crtime, 16); 918219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs), 919219089Spjd NULL, &gen, 8); 920219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs), 921219089Spjd NULL, &mode, 8); 922219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs), 923219089Spjd NULL, &size, 8); 924219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs), 925219089Spjd NULL, &parent, 8); 926219089Spjd } else { 927219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs), 928219089Spjd NULL, &mode, 8); 929219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs), 930219089Spjd NULL, &size, 8); 931219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs), 932219089Spjd NULL, &gen, 8); 933219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL, 934219089Spjd &acl_ids->z_fuid, 8); 935219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL, 936219089Spjd &acl_ids->z_fgid, 8); 937219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs), 938219089Spjd NULL, &parent, 8); 939219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs), 940219089Spjd NULL, &pflags, 8); 941219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs), 942219089Spjd NULL, &atime, 16); 943219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs), 944219089Spjd NULL, &mtime, 16); 945219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs), 946219089Spjd NULL, &ctime, 16); 947219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs), 948219089Spjd NULL, &crtime, 16); 949219089Spjd } 950219089Spjd 951219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8); 952219089Spjd 953219089Spjd if (obj_type == DMU_OT_ZNODE) { 954219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zfsvfs), NULL, 955219089Spjd &empty_xattr, 8); 956219089Spjd } 957219089Spjd if (obj_type == DMU_OT_ZNODE || 958219089Spjd (vap->va_type == VBLK || vap->va_type == VCHR)) { 959219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zfsvfs), 960219089Spjd NULL, &rdev, 8); 961219089Spjd 962219089Spjd } 963219089Spjd if (obj_type == DMU_OT_ZNODE) { 964219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs), 965219089Spjd NULL, &pflags, 8); 966219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL, 967219089Spjd &acl_ids->z_fuid, 8); 968219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL, 969219089Spjd &acl_ids->z_fgid, 8); 970219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zfsvfs), NULL, pad, 971219089Spjd sizeof (uint64_t) * 4); 972219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zfsvfs), NULL, 973219089Spjd &acl_phys, sizeof (zfs_acl_phys_t)); 974219089Spjd } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) { 975219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zfsvfs), NULL, 976219089Spjd &acl_ids->z_aclp->z_acl_count, 8); 977219089Spjd locate.cb_aclp = acl_ids->z_aclp; 978219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zfsvfs), 979219089Spjd zfs_acl_data_locator, &locate, 980219089Spjd acl_ids->z_aclp->z_acl_bytes); 981219089Spjd mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags, 982219089Spjd acl_ids->z_fuid, acl_ids->z_fgid); 983219089Spjd } 984219089Spjd 985219089Spjd VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0); 986219089Spjd 987185029Spjd if (!(flag & IS_ROOT_NODE)) { 988219089Spjd *zpp = zfs_znode_alloc(zfsvfs, db, 0, obj_type, sa_hdl); 989219089Spjd ASSERT(*zpp != NULL); 990185029Spjd } else { 991185029Spjd /* 992185029Spjd * If we are creating the root node, the "parent" we 993185029Spjd * passed in is the znode for the root. 994185029Spjd */ 995185029Spjd *zpp = dzp; 996219089Spjd 997219089Spjd (*zpp)->z_sa_hdl = sa_hdl; 998185029Spjd } 999219089Spjd 1000219089Spjd (*zpp)->z_pflags = pflags; 1001219089Spjd (*zpp)->z_mode = mode; 1002219089Spjd 1003209962Smm if (vap->va_mask & AT_XVATTR) 1004219089Spjd zfs_xvattr_set(*zpp, (xvattr_t *)vap, tx); 1005219089Spjd 1006219089Spjd if (obj_type == DMU_OT_ZNODE || 1007219089Spjd acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) { 1008263398Sdelphij VERIFY0(zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx)); 1009219089Spjd } 1010185029Spjd if (!(flag & IS_ROOT_NODE)) { 1011185029Spjd vnode_t *vp; 1012168404Spjd 1013185029Spjd vp = ZTOV(*zpp); 1014185029Spjd vp->v_vflag |= VV_FORCEINSMQ; 1015185029Spjd err = insmntque(vp, zfsvfs->z_vfs); 1016185029Spjd vp->v_vflag &= ~VV_FORCEINSMQ; 1017185029Spjd KASSERT(err == 0, ("insmntque() failed: error %d", err)); 1018185029Spjd } 1019219089Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, obj); 1020185029Spjd} 1021168404Spjd 1022219089Spjd/* 1023252751Sdelphij * Update in-core attributes. It is assumed the caller will be doing an 1024252751Sdelphij * sa_bulk_update to push the changes out. 1025219089Spjd */ 1026185029Spjdvoid 1027219089Spjdzfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx) 1028185029Spjd{ 1029185029Spjd xoptattr_t *xoap; 1030168404Spjd 1031185029Spjd xoap = xva_getxoptattr(xvap); 1032185029Spjd ASSERT(xoap); 1033168404Spjd 1034185029Spjd if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) { 1035219089Spjd uint64_t times[2]; 1036219089Spjd ZFS_TIME_ENCODE(&xoap->xoa_createtime, times); 1037219089Spjd (void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(zp->z_zfsvfs), 1038219089Spjd ×, sizeof (times), tx); 1039185029Spjd XVA_SET_RTN(xvap, XAT_CREATETIME); 1040168404Spjd } 1041185029Spjd if (XVA_ISSET_REQ(xvap, XAT_READONLY)) { 1042219089Spjd ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly, 1043219089Spjd zp->z_pflags, tx); 1044185029Spjd XVA_SET_RTN(xvap, XAT_READONLY); 1045185029Spjd } 1046185029Spjd if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) { 1047219089Spjd ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden, 1048219089Spjd zp->z_pflags, tx); 1049185029Spjd XVA_SET_RTN(xvap, XAT_HIDDEN); 1050185029Spjd } 1051185029Spjd if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) { 1052219089Spjd ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system, 1053219089Spjd zp->z_pflags, tx); 1054185029Spjd XVA_SET_RTN(xvap, XAT_SYSTEM); 1055185029Spjd } 1056185029Spjd if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) { 1057219089Spjd ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive, 1058219089Spjd zp->z_pflags, tx); 1059185029Spjd XVA_SET_RTN(xvap, XAT_ARCHIVE); 1060185029Spjd } 1061185029Spjd if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { 1062219089Spjd ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable, 1063219089Spjd zp->z_pflags, tx); 1064185029Spjd XVA_SET_RTN(xvap, XAT_IMMUTABLE); 1065185029Spjd } 1066185029Spjd if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { 1067219089Spjd ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink, 1068219089Spjd zp->z_pflags, tx); 1069185029Spjd XVA_SET_RTN(xvap, XAT_NOUNLINK); 1070185029Spjd } 1071185029Spjd if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { 1072219089Spjd ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly, 1073219089Spjd zp->z_pflags, tx); 1074185029Spjd XVA_SET_RTN(xvap, XAT_APPENDONLY); 1075185029Spjd } 1076185029Spjd if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { 1077219089Spjd ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump, 1078219089Spjd zp->z_pflags, tx); 1079185029Spjd XVA_SET_RTN(xvap, XAT_NODUMP); 1080185029Spjd } 1081185029Spjd if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) { 1082219089Spjd ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque, 1083219089Spjd zp->z_pflags, tx); 1084185029Spjd XVA_SET_RTN(xvap, XAT_OPAQUE); 1085185029Spjd } 1086185029Spjd if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { 1087185029Spjd ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED, 1088219089Spjd xoap->xoa_av_quarantined, zp->z_pflags, tx); 1089185029Spjd XVA_SET_RTN(xvap, XAT_AV_QUARANTINED); 1090185029Spjd } 1091185029Spjd if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { 1092219089Spjd ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified, 1093219089Spjd zp->z_pflags, tx); 1094185029Spjd XVA_SET_RTN(xvap, XAT_AV_MODIFIED); 1095185029Spjd } 1096185029Spjd if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) { 1097219089Spjd zfs_sa_set_scanstamp(zp, xvap, tx); 1098185029Spjd XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP); 1099185029Spjd } 1100219089Spjd if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { 1101219089Spjd ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse, 1102219089Spjd zp->z_pflags, tx); 1103219089Spjd XVA_SET_RTN(xvap, XAT_REPARSE); 1104219089Spjd } 1105219089Spjd if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) { 1106219089Spjd ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline, 1107219089Spjd zp->z_pflags, tx); 1108219089Spjd XVA_SET_RTN(xvap, XAT_OFFLINE); 1109219089Spjd } 1110219089Spjd if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) { 1111219089Spjd ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse, 1112219089Spjd zp->z_pflags, tx); 1113219089Spjd XVA_SET_RTN(xvap, XAT_SPARSE); 1114219089Spjd } 1115168404Spjd} 1116168404Spjd 1117168404Spjdint 1118168404Spjdzfs_zget(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp) 1119168404Spjd{ 1120168404Spjd dmu_object_info_t doi; 1121168404Spjd dmu_buf_t *db; 1122168404Spjd znode_t *zp; 1123244626Savg vnode_t *vp; 1124244626Savg sa_handle_t *hdl; 1125244626Savg struct thread *td; 1126244626Savg int locked; 1127219089Spjd int err; 1128168404Spjd 1129244626Savg td = curthread; 1130242240Savg getnewvnode_reserve(1); 1131185029Spjdagain: 1132244626Savg *zpp = NULL; 1133168404Spjd ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num); 1134168404Spjd 1135219089Spjd err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db); 1136168404Spjd if (err) { 1137168404Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1138242240Savg getnewvnode_drop_reserve(); 1139168404Spjd return (err); 1140168404Spjd } 1141168404Spjd 1142168404Spjd dmu_object_info_from_db(db, &doi); 1143219089Spjd if (doi.doi_bonus_type != DMU_OT_SA && 1144219089Spjd (doi.doi_bonus_type != DMU_OT_ZNODE || 1145219089Spjd (doi.doi_bonus_type == DMU_OT_ZNODE && 1146219089Spjd doi.doi_bonus_size < sizeof (znode_phys_t)))) { 1147219089Spjd sa_buf_rele(db, NULL); 1148168404Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1149249643Smm#ifdef __FreeBSD__ 1150242240Savg getnewvnode_drop_reserve(); 1151249643Smm#endif 1152249643Smm return (SET_ERROR(EINVAL)); 1153168404Spjd } 1154168404Spjd 1155219089Spjd hdl = dmu_buf_get_user(db); 1156219089Spjd if (hdl != NULL) { 1157219089Spjd zp = sa_get_userdata(hdl); 1158168404Spjd 1159219089Spjd 1160185029Spjd /* 1161219089Spjd * Since "SA" does immediate eviction we 1162219089Spjd * should never find a sa handle that doesn't 1163219089Spjd * know about the znode. 1164185029Spjd */ 1165219089Spjd 1166219089Spjd ASSERT3P(zp, !=, NULL); 1167219089Spjd 1168219089Spjd mutex_enter(&zp->z_lock); 1169168404Spjd ASSERT3U(zp->z_id, ==, obj_num); 1170168404Spjd if (zp->z_unlinked) { 1171249643Smm err = SET_ERROR(ENOENT); 1172168404Spjd } else { 1173197458Spjd vp = ZTOV(zp); 1174185029Spjd *zpp = zp; 1175185029Spjd err = 0; 1176168404Spjd } 1177219089Spjd sa_buf_rele(db, NULL); 1178244626Savg 1179244626Savg /* Don't let the vnode disappear after ZFS_OBJ_HOLD_EXIT. */ 1180244626Savg if (err == 0) 1181244626Savg VN_HOLD(vp); 1182244626Savg 1183168404Spjd mutex_exit(&zp->z_lock); 1184168404Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1185244626Savg 1186244626Savg if (err == 0) { 1187244626Savg locked = VOP_ISLOCKED(vp); 1188244626Savg VI_LOCK(vp); 1189244626Savg if ((vp->v_iflag & VI_DOOMED) != 0 && 1190244626Savg locked != LK_EXCLUSIVE) { 1191244626Savg /* 1192244626Savg * The vnode is doomed and this thread doesn't 1193244626Savg * hold the exclusive lock on it, so the vnode 1194244626Savg * must be being reclaimed by another thread. 1195244626Savg * Otherwise the doomed vnode is being reclaimed 1196244626Savg * by this thread and zfs_zget is called from 1197244626Savg * ZIL internals. 1198244626Savg */ 1199244626Savg VI_UNLOCK(vp); 1200244626Savg VN_RELE(vp); 1201244626Savg goto again; 1202244626Savg } 1203244626Savg VI_UNLOCK(vp); 1204244626Savg } 1205242240Savg getnewvnode_drop_reserve(); 1206185029Spjd return (err); 1207168404Spjd } 1208168404Spjd 1209168404Spjd /* 1210168404Spjd * Not found create new znode/vnode 1211207334Spjd * but only if file exists. 1212207334Spjd * 1213207334Spjd * There is a small window where zfs_vget() could 1214207334Spjd * find this object while a file create is still in 1215219089Spjd * progress. This is checked for in zfs_znode_alloc() 1216219089Spjd * 1217219089Spjd * if zfs_znode_alloc() fails it will drop the hold on the 1218219089Spjd * bonus buffer. 1219168404Spjd */ 1220219089Spjd zp = zfs_znode_alloc(zfsvfs, db, doi.doi_data_block_size, 1221219089Spjd doi.doi_bonus_type, NULL); 1222219089Spjd if (zp == NULL) { 1223249643Smm err = SET_ERROR(ENOENT); 1224219089Spjd } else { 1225219089Spjd *zpp = zp; 1226219089Spjd } 1227219089Spjd if (err == 0) { 1228219089Spjd vnode_t *vp = ZTOV(zp); 1229185029Spjd 1230207334Spjd err = insmntque(vp, zfsvfs->z_vfs); 1231219089Spjd if (err == 0) 1232219089Spjd VOP_UNLOCK(vp, 0); 1233219089Spjd else { 1234219089Spjd zp->z_vnode = NULL; 1235219089Spjd zfs_znode_dmu_fini(zp); 1236219089Spjd zfs_znode_free(zp); 1237219089Spjd *zpp = NULL; 1238219089Spjd } 1239207334Spjd } 1240168404Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1241242240Savg getnewvnode_drop_reserve(); 1242207334Spjd return (err); 1243168404Spjd} 1244168404Spjd 1245185029Spjdint 1246185029Spjdzfs_rezget(znode_t *zp) 1247185029Spjd{ 1248185029Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1249185029Spjd dmu_object_info_t doi; 1250185029Spjd dmu_buf_t *db; 1251225166Smm vnode_t *vp; 1252185029Spjd uint64_t obj_num = zp->z_id; 1253219089Spjd uint64_t mode, size; 1254219089Spjd sa_bulk_attr_t bulk[8]; 1255185029Spjd int err; 1256219089Spjd int count = 0; 1257219089Spjd uint64_t gen; 1258185029Spjd 1259185029Spjd ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num); 1260185029Spjd 1261219089Spjd mutex_enter(&zp->z_acl_lock); 1262219089Spjd if (zp->z_acl_cached) { 1263219089Spjd zfs_acl_free(zp->z_acl_cached); 1264219089Spjd zp->z_acl_cached = NULL; 1265219089Spjd } 1266219089Spjd 1267219089Spjd mutex_exit(&zp->z_acl_lock); 1268219089Spjd ASSERT(zp->z_sa_hdl == NULL); 1269219089Spjd err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db); 1270185029Spjd if (err) { 1271185029Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1272185029Spjd return (err); 1273185029Spjd } 1274185029Spjd 1275185029Spjd dmu_object_info_from_db(db, &doi); 1276219089Spjd if (doi.doi_bonus_type != DMU_OT_SA && 1277219089Spjd (doi.doi_bonus_type != DMU_OT_ZNODE || 1278219089Spjd (doi.doi_bonus_type == DMU_OT_ZNODE && 1279219089Spjd doi.doi_bonus_size < sizeof (znode_phys_t)))) { 1280219089Spjd sa_buf_rele(db, NULL); 1281185029Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1282249643Smm return (SET_ERROR(EINVAL)); 1283185029Spjd } 1284185029Spjd 1285219089Spjd zfs_znode_sa_init(zfsvfs, zp, db, doi.doi_bonus_type, NULL); 1286219089Spjd size = zp->z_size; 1287219089Spjd 1288219089Spjd /* reload cached values */ 1289219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, 1290219089Spjd &gen, sizeof (gen)); 1291219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, 1292219089Spjd &zp->z_size, sizeof (zp->z_size)); 1293219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL, 1294219089Spjd &zp->z_links, sizeof (zp->z_links)); 1295219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, 1296219089Spjd &zp->z_pflags, sizeof (zp->z_pflags)); 1297219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, 1298219089Spjd &zp->z_atime, sizeof (zp->z_atime)); 1299219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, 1300219089Spjd &zp->z_uid, sizeof (zp->z_uid)); 1301219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, 1302219089Spjd &zp->z_gid, sizeof (zp->z_gid)); 1303219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, 1304219089Spjd &mode, sizeof (mode)); 1305219089Spjd 1306219089Spjd if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) { 1307219089Spjd zfs_znode_dmu_fini(zp); 1308185029Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1309249643Smm return (SET_ERROR(EIO)); 1310185029Spjd } 1311185029Spjd 1312219089Spjd zp->z_mode = mode; 1313219089Spjd 1314219089Spjd if (gen != zp->z_gen) { 1315219089Spjd zfs_znode_dmu_fini(zp); 1316219089Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1317249643Smm return (SET_ERROR(EIO)); 1318211932Smm } 1319211932Smm 1320219089Spjd /* 1321219089Spjd * XXXPJD: Not sure how is that possible, but under heavy 1322219089Spjd * zfs recv -F load it happens that z_gen is the same, but 1323219089Spjd * vnode type is different than znode type. This would mean 1324219089Spjd * that for example regular file was replaced with directory 1325219089Spjd * which has the same object number. 1326219089Spjd */ 1327225166Smm vp = ZTOV(zp); 1328225166Smm if (vp != NULL && 1329225166Smm vp->v_type != IFTOVT((mode_t)zp->z_mode)) { 1330219089Spjd zfs_znode_dmu_fini(zp); 1331219089Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1332219089Spjd return (EIO); 1333219089Spjd } 1334219089Spjd 1335219089Spjd zp->z_unlinked = (zp->z_links == 0); 1336185029Spjd zp->z_blksz = doi.doi_data_block_size; 1337225166Smm if (vp != NULL) { 1338225166Smm vn_pages_remove(vp, 0, 0); 1339225166Smm if (zp->z_size != size) 1340225166Smm vnode_pager_setsize(vp, zp->z_size); 1341225166Smm } 1342185029Spjd 1343185029Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1344185029Spjd 1345185029Spjd return (0); 1346185029Spjd} 1347185029Spjd 1348168404Spjdvoid 1349168404Spjdzfs_znode_delete(znode_t *zp, dmu_tx_t *tx) 1350168404Spjd{ 1351168404Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1352185029Spjd objset_t *os = zfsvfs->z_os; 1353185029Spjd uint64_t obj = zp->z_id; 1354219089Spjd uint64_t acl_obj = zfs_external_acl(zp); 1355168404Spjd 1356185029Spjd ZFS_OBJ_HOLD_ENTER(zfsvfs, obj); 1357219089Spjd if (acl_obj) { 1358219089Spjd VERIFY(!zp->z_is_sa); 1359185029Spjd VERIFY(0 == dmu_object_free(os, acl_obj, tx)); 1360219089Spjd } 1361185029Spjd VERIFY(0 == dmu_object_free(os, obj, tx)); 1362185029Spjd zfs_znode_dmu_fini(zp); 1363185029Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, obj); 1364185029Spjd zfs_znode_free(zp); 1365168404Spjd} 1366168404Spjd 1367168404Spjdvoid 1368168404Spjdzfs_zinactive(znode_t *zp) 1369168404Spjd{ 1370168404Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1371168404Spjd uint64_t z_id = zp->z_id; 1372168404Spjd 1373219089Spjd ASSERT(zp->z_sa_hdl); 1374168404Spjd 1375168404Spjd /* 1376168404Spjd * Don't allow a zfs_zget() while were trying to release this znode 1377168404Spjd */ 1378168404Spjd ZFS_OBJ_HOLD_ENTER(zfsvfs, z_id); 1379168404Spjd 1380168404Spjd mutex_enter(&zp->z_lock); 1381168404Spjd 1382168404Spjd /* 1383168404Spjd * If this was the last reference to a file with no links, 1384168404Spjd * remove the file from the file system. 1385168404Spjd */ 1386168404Spjd if (zp->z_unlinked) { 1387168404Spjd mutex_exit(&zp->z_lock); 1388168404Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id); 1389168404Spjd zfs_rmnode(zp); 1390168404Spjd return; 1391168404Spjd } 1392219089Spjd 1393168404Spjd mutex_exit(&zp->z_lock); 1394244626Savg zfs_znode_dmu_fini(zp); 1395168404Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id); 1396244626Savg zfs_znode_free(zp); 1397168404Spjd} 1398168404Spjd 1399168404Spjdvoid 1400168404Spjdzfs_znode_free(znode_t *zp) 1401168404Spjd{ 1402168404Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1403168404Spjd 1404219089Spjd ASSERT(zp->z_sa_hdl == NULL); 1405244626Savg zp->z_vnode = NULL; 1406168404Spjd mutex_enter(&zfsvfs->z_znodes_lock); 1407185029Spjd POINTER_INVALIDATE(&zp->z_zfsvfs); 1408168404Spjd list_remove(&zfsvfs->z_all_znodes, zp); 1409168404Spjd mutex_exit(&zfsvfs->z_znodes_lock); 1410168404Spjd 1411211932Smm if (zp->z_acl_cached) { 1412211932Smm zfs_acl_free(zp->z_acl_cached); 1413211932Smm zp->z_acl_cached = NULL; 1414211932Smm } 1415211932Smm 1416168404Spjd kmem_cache_free(znode_cache, zp); 1417185029Spjd 1418185029Spjd VFS_RELE(zfsvfs->z_vfs); 1419168404Spjd} 1420168404Spjd 1421168404Spjdvoid 1422219089Spjdzfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2], 1423219089Spjd uint64_t ctime[2], boolean_t have_tx) 1424168404Spjd{ 1425168404Spjd timestruc_t now; 1426168404Spjd 1427168404Spjd gethrestime(&now); 1428168404Spjd 1429219089Spjd if (have_tx) { /* will sa_bulk_update happen really soon? */ 1430168404Spjd zp->z_atime_dirty = 0; 1431168404Spjd zp->z_seq++; 1432168404Spjd } else { 1433168404Spjd zp->z_atime_dirty = 1; 1434168404Spjd } 1435168404Spjd 1436219089Spjd if (flag & AT_ATIME) { 1437219089Spjd ZFS_TIME_ENCODE(&now, zp->z_atime); 1438219089Spjd } 1439168404Spjd 1440185029Spjd if (flag & AT_MTIME) { 1441219089Spjd ZFS_TIME_ENCODE(&now, mtime); 1442219089Spjd if (zp->z_zfsvfs->z_use_fuids) { 1443219089Spjd zp->z_pflags |= (ZFS_ARCHIVE | 1444219089Spjd ZFS_AV_MODIFIED); 1445219089Spjd } 1446185029Spjd } 1447168404Spjd 1448185029Spjd if (flag & AT_CTIME) { 1449219089Spjd ZFS_TIME_ENCODE(&now, ctime); 1450185029Spjd if (zp->z_zfsvfs->z_use_fuids) 1451219089Spjd zp->z_pflags |= ZFS_ARCHIVE; 1452185029Spjd } 1453168404Spjd} 1454168404Spjd 1455168404Spjd/* 1456168404Spjd * Grow the block size for a file. 1457168404Spjd * 1458168404Spjd * IN: zp - znode of file to free data in. 1459168404Spjd * size - requested block size 1460168404Spjd * tx - open transaction. 1461168404Spjd * 1462168404Spjd * NOTE: this function assumes that the znode is write locked. 1463168404Spjd */ 1464168404Spjdvoid 1465168404Spjdzfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx) 1466168404Spjd{ 1467168404Spjd int error; 1468168404Spjd u_longlong_t dummy; 1469168404Spjd 1470168404Spjd if (size <= zp->z_blksz) 1471168404Spjd return; 1472168404Spjd /* 1473168404Spjd * If the file size is already greater than the current blocksize, 1474168404Spjd * we will not grow. If there is more than one block in a file, 1475168404Spjd * the blocksize cannot change. 1476168404Spjd */ 1477219089Spjd if (zp->z_blksz && zp->z_size > zp->z_blksz) 1478168404Spjd return; 1479168404Spjd 1480168404Spjd error = dmu_object_set_blocksize(zp->z_zfsvfs->z_os, zp->z_id, 1481168404Spjd size, 0, tx); 1482219089Spjd 1483168404Spjd if (error == ENOTSUP) 1484168404Spjd return; 1485243674Smm ASSERT0(error); 1486168404Spjd 1487168404Spjd /* What blocksize did we actually get? */ 1488219089Spjd dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy); 1489168404Spjd} 1490168404Spjd 1491219089Spjd#ifdef sun 1492168404Spjd/* 1493219089Spjd * This is a dummy interface used when pvn_vplist_dirty() should *not* 1494219089Spjd * be calling back into the fs for a putpage(). E.g.: when truncating 1495219089Spjd * a file, the pages being "thrown away* don't need to be written out. 1496219089Spjd */ 1497219089Spjd/* ARGSUSED */ 1498219089Spjdstatic int 1499219089Spjdzfs_no_putpage(vnode_t *vp, page_t *pp, u_offset_t *offp, size_t *lenp, 1500219089Spjd int flags, cred_t *cr) 1501219089Spjd{ 1502219089Spjd ASSERT(0); 1503219089Spjd return (0); 1504219089Spjd} 1505219089Spjd#endif /* sun */ 1506219089Spjd 1507219089Spjd/* 1508185029Spjd * Increase the file length 1509168404Spjd * 1510168404Spjd * IN: zp - znode of file to free data in. 1511185029Spjd * end - new end-of-file 1512168404Spjd * 1513252751Sdelphij * RETURN: 0 on success, error code on failure 1514168404Spjd */ 1515185029Spjdstatic int 1516185029Spjdzfs_extend(znode_t *zp, uint64_t end) 1517168404Spjd{ 1518185029Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1519168404Spjd dmu_tx_t *tx; 1520168404Spjd rl_t *rl; 1521185029Spjd uint64_t newblksz; 1522168404Spjd int error; 1523168404Spjd 1524168404Spjd /* 1525185029Spjd * We will change zp_size, lock the whole file. 1526168404Spjd */ 1527185029Spjd rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER); 1528168404Spjd 1529168404Spjd /* 1530168404Spjd * Nothing to do if file already at desired length. 1531168404Spjd */ 1532219089Spjd if (end <= zp->z_size) { 1533168404Spjd zfs_range_unlock(rl); 1534168404Spjd return (0); 1535168404Spjd } 1536168404Spjd tx = dmu_tx_create(zfsvfs->z_os); 1537219089Spjd dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); 1538219089Spjd zfs_sa_upgrade_txholds(tx, zp); 1539185029Spjd if (end > zp->z_blksz && 1540168404Spjd (!ISP2(zp->z_blksz) || zp->z_blksz < zfsvfs->z_max_blksz)) { 1541168404Spjd /* 1542168404Spjd * We are growing the file past the current block size. 1543168404Spjd */ 1544168404Spjd if (zp->z_blksz > zp->z_zfsvfs->z_max_blksz) { 1545168404Spjd ASSERT(!ISP2(zp->z_blksz)); 1546185029Spjd newblksz = MIN(end, SPA_MAXBLOCKSIZE); 1547168404Spjd } else { 1548185029Spjd newblksz = MIN(end, zp->z_zfsvfs->z_max_blksz); 1549168404Spjd } 1550185029Spjd dmu_tx_hold_write(tx, zp->z_id, 0, newblksz); 1551185029Spjd } else { 1552185029Spjd newblksz = 0; 1553168404Spjd } 1554168404Spjd 1555260777Savg error = dmu_tx_assign(tx, TXG_WAIT); 1556168404Spjd if (error) { 1557168404Spjd dmu_tx_abort(tx); 1558168404Spjd zfs_range_unlock(rl); 1559168404Spjd return (error); 1560168404Spjd } 1561168404Spjd 1562185029Spjd if (newblksz) 1563185029Spjd zfs_grow_blocksize(zp, newblksz, tx); 1564168404Spjd 1565219089Spjd zp->z_size = end; 1566168404Spjd 1567219089Spjd VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zp->z_zfsvfs), 1568219089Spjd &zp->z_size, sizeof (zp->z_size), tx)); 1569219089Spjd 1570219089Spjd vnode_pager_setsize(ZTOV(zp), end); 1571219089Spjd 1572185029Spjd zfs_range_unlock(rl); 1573168404Spjd 1574185029Spjd dmu_tx_commit(tx); 1575185029Spjd 1576185029Spjd return (0); 1577185029Spjd} 1578185029Spjd 1579185029Spjd/* 1580185029Spjd * Free space in a file. 1581185029Spjd * 1582185029Spjd * IN: zp - znode of file to free data in. 1583185029Spjd * off - start of section to free. 1584185029Spjd * len - length of section to free. 1585185029Spjd * 1586252751Sdelphij * RETURN: 0 on success, error code on failure 1587185029Spjd */ 1588185029Spjdstatic int 1589185029Spjdzfs_free_range(znode_t *zp, uint64_t off, uint64_t len) 1590185029Spjd{ 1591185029Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1592185029Spjd rl_t *rl; 1593185029Spjd int error; 1594185029Spjd 1595185029Spjd /* 1596185029Spjd * Lock the range being freed. 1597185029Spjd */ 1598185029Spjd rl = zfs_range_lock(zp, off, len, RL_WRITER); 1599185029Spjd 1600185029Spjd /* 1601185029Spjd * Nothing to do if file already at desired length. 1602185029Spjd */ 1603219089Spjd if (off >= zp->z_size) { 1604185029Spjd zfs_range_unlock(rl); 1605185029Spjd return (0); 1606168404Spjd } 1607168404Spjd 1608219089Spjd if (off + len > zp->z_size) 1609219089Spjd len = zp->z_size - off; 1610185029Spjd 1611185029Spjd error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, off, len); 1612185029Spjd 1613185029Spjd if (error == 0) { 1614185029Spjd /* 1615185029Spjd * In FreeBSD we cannot free block in the middle of a file, 1616219089Spjd * but only at the end of a file, so this code path should 1617219089Spjd * never happen. 1618185029Spjd */ 1619185029Spjd vnode_pager_setsize(ZTOV(zp), off); 1620168404Spjd } 1621168404Spjd 1622168404Spjd zfs_range_unlock(rl); 1623168404Spjd 1624185029Spjd return (error); 1625185029Spjd} 1626185029Spjd 1627185029Spjd/* 1628185029Spjd * Truncate a file 1629185029Spjd * 1630185029Spjd * IN: zp - znode of file to free data in. 1631185029Spjd * end - new end-of-file. 1632185029Spjd * 1633252751Sdelphij * RETURN: 0 on success, error code on failure 1634185029Spjd */ 1635185029Spjdstatic int 1636185029Spjdzfs_trunc(znode_t *zp, uint64_t end) 1637185029Spjd{ 1638185029Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1639185029Spjd vnode_t *vp = ZTOV(zp); 1640185029Spjd dmu_tx_t *tx; 1641185029Spjd rl_t *rl; 1642185029Spjd int error; 1643219089Spjd sa_bulk_attr_t bulk[2]; 1644219089Spjd int count = 0; 1645185029Spjd 1646185029Spjd /* 1647185029Spjd * We will change zp_size, lock the whole file. 1648185029Spjd */ 1649185029Spjd rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER); 1650185029Spjd 1651185029Spjd /* 1652185029Spjd * Nothing to do if file already at desired length. 1653185029Spjd */ 1654219089Spjd if (end >= zp->z_size) { 1655185029Spjd zfs_range_unlock(rl); 1656185029Spjd return (0); 1657185029Spjd } 1658185029Spjd 1659185029Spjd error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, end, -1); 1660185029Spjd if (error) { 1661185029Spjd zfs_range_unlock(rl); 1662185029Spjd return (error); 1663185029Spjd } 1664185029Spjd tx = dmu_tx_create(zfsvfs->z_os); 1665219089Spjd dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); 1666219089Spjd zfs_sa_upgrade_txholds(tx, zp); 1667260777Savg error = dmu_tx_assign(tx, TXG_WAIT); 1668185029Spjd if (error) { 1669185029Spjd dmu_tx_abort(tx); 1670185029Spjd zfs_range_unlock(rl); 1671185029Spjd return (error); 1672185029Spjd } 1673185029Spjd 1674219089Spjd zp->z_size = end; 1675219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), 1676219089Spjd NULL, &zp->z_size, sizeof (zp->z_size)); 1677185029Spjd 1678219089Spjd if (end == 0) { 1679219089Spjd zp->z_pflags &= ~ZFS_SPARSE; 1680219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), 1681219089Spjd NULL, &zp->z_pflags, 8); 1682219089Spjd } 1683219089Spjd VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0); 1684219089Spjd 1685168404Spjd dmu_tx_commit(tx); 1686168404Spjd 1687168404Spjd /* 1688168404Spjd * Clear any mapped pages in the truncated region. This has to 1689168404Spjd * happen outside of the transaction to avoid the possibility of 1690168404Spjd * a deadlock with someone trying to push a page that we are 1691168404Spjd * about to invalidate. 1692168404Spjd */ 1693185029Spjd vnode_pager_setsize(vp, end); 1694168404Spjd 1695209962Smm zfs_range_unlock(rl); 1696209962Smm 1697168404Spjd return (0); 1698168404Spjd} 1699168404Spjd 1700185029Spjd/* 1701185029Spjd * Free space in a file 1702185029Spjd * 1703185029Spjd * IN: zp - znode of file to free data in. 1704185029Spjd * off - start of range 1705185029Spjd * len - end of range (0 => EOF) 1706185029Spjd * flag - current file open mode flags. 1707185029Spjd * log - TRUE if this action should be logged 1708185029Spjd * 1709252751Sdelphij * RETURN: 0 on success, error code on failure 1710185029Spjd */ 1711185029Spjdint 1712185029Spjdzfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log) 1713185029Spjd{ 1714185029Spjd vnode_t *vp = ZTOV(zp); 1715185029Spjd dmu_tx_t *tx; 1716185029Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1717185029Spjd zilog_t *zilog = zfsvfs->z_log; 1718219089Spjd uint64_t mode; 1719219089Spjd uint64_t mtime[2], ctime[2]; 1720219089Spjd sa_bulk_attr_t bulk[3]; 1721219089Spjd int count = 0; 1722185029Spjd int error; 1723185029Spjd 1724219089Spjd if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), &mode, 1725219089Spjd sizeof (mode))) != 0) 1726219089Spjd return (error); 1727219089Spjd 1728219089Spjd if (off > zp->z_size) { 1729185029Spjd error = zfs_extend(zp, off+len); 1730185029Spjd if (error == 0 && log) 1731185029Spjd goto log; 1732185029Spjd else 1733185029Spjd return (error); 1734185029Spjd } 1735185029Spjd 1736219089Spjd /* 1737219089Spjd * Check for any locks in the region to be freed. 1738219089Spjd */ 1739219089Spjd 1740219089Spjd if (MANDLOCK(vp, (mode_t)mode)) { 1741219089Spjd uint64_t length = (len ? len : zp->z_size - off); 1742219089Spjd if (error = chklock(vp, FWRITE, off, length, flag, NULL)) 1743219089Spjd return (error); 1744219089Spjd } 1745219089Spjd 1746185029Spjd if (len == 0) { 1747185029Spjd error = zfs_trunc(zp, off); 1748185029Spjd } else { 1749185029Spjd if ((error = zfs_free_range(zp, off, len)) == 0 && 1750219089Spjd off + len > zp->z_size) 1751185029Spjd error = zfs_extend(zp, off+len); 1752185029Spjd } 1753185029Spjd if (error || !log) 1754185029Spjd return (error); 1755185029Spjdlog: 1756185029Spjd tx = dmu_tx_create(zfsvfs->z_os); 1757219089Spjd dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); 1758219089Spjd zfs_sa_upgrade_txholds(tx, zp); 1759260777Savg error = dmu_tx_assign(tx, TXG_WAIT); 1760185029Spjd if (error) { 1761185029Spjd dmu_tx_abort(tx); 1762185029Spjd return (error); 1763185029Spjd } 1764185029Spjd 1765219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, mtime, 16); 1766219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, ctime, 16); 1767219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), 1768219089Spjd NULL, &zp->z_pflags, 8); 1769219089Spjd zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, B_TRUE); 1770219089Spjd error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); 1771219089Spjd ASSERT(error == 0); 1772219089Spjd 1773185029Spjd zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len); 1774185029Spjd 1775185029Spjd dmu_tx_commit(tx); 1776185029Spjd return (0); 1777185029Spjd} 1778185029Spjd 1779168404Spjdvoid 1780185029Spjdzfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx) 1781168404Spjd{ 1782168404Spjd zfsvfs_t zfsvfs; 1783219089Spjd uint64_t moid, obj, sa_obj, version; 1784185029Spjd uint64_t sense = ZFS_CASE_SENSITIVE; 1785185029Spjd uint64_t norm = 0; 1786185029Spjd nvpair_t *elem; 1787168404Spjd int error; 1788207334Spjd int i; 1789168404Spjd znode_t *rootzp = NULL; 1790168404Spjd vattr_t vattr; 1791185029Spjd znode_t *zp; 1792209962Smm zfs_acl_ids_t acl_ids; 1793168404Spjd 1794168404Spjd /* 1795168404Spjd * First attempt to create master node. 1796168404Spjd */ 1797168404Spjd /* 1798168404Spjd * In an empty objset, there are no blocks to read and thus 1799168404Spjd * there can be no i/o errors (which we assert below). 1800168404Spjd */ 1801168404Spjd moid = MASTER_NODE_OBJ; 1802168404Spjd error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE, 1803168404Spjd DMU_OT_NONE, 0, tx); 1804168404Spjd ASSERT(error == 0); 1805168404Spjd 1806168404Spjd /* 1807168404Spjd * Set starting attributes. 1808168404Spjd */ 1809219089Spjd version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os))); 1810185029Spjd elem = NULL; 1811185029Spjd while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) { 1812185029Spjd /* For the moment we expect all zpl props to be uint64_ts */ 1813185029Spjd uint64_t val; 1814185029Spjd char *name; 1815168404Spjd 1816185029Spjd ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64); 1817185029Spjd VERIFY(nvpair_value_uint64(elem, &val) == 0); 1818185029Spjd name = nvpair_name(elem); 1819185029Spjd if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) { 1820209962Smm if (val < version) 1821209962Smm version = val; 1822185029Spjd } else { 1823185029Spjd error = zap_update(os, moid, name, 8, 1, &val, tx); 1824185029Spjd } 1825185029Spjd ASSERT(error == 0); 1826185029Spjd if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0) 1827185029Spjd norm = val; 1828185029Spjd else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0) 1829185029Spjd sense = val; 1830185029Spjd } 1831185029Spjd ASSERT(version != 0); 1832209962Smm error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx); 1833168404Spjd 1834168404Spjd /* 1835219089Spjd * Create zap object used for SA attribute registration 1836219089Spjd */ 1837219089Spjd 1838219089Spjd if (version >= ZPL_VERSION_SA) { 1839219089Spjd sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE, 1840219089Spjd DMU_OT_NONE, 0, tx); 1841219089Spjd error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx); 1842219089Spjd ASSERT(error == 0); 1843219089Spjd } else { 1844219089Spjd sa_obj = 0; 1845219089Spjd } 1846219089Spjd /* 1847168404Spjd * Create a delete queue. 1848168404Spjd */ 1849209962Smm obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx); 1850168404Spjd 1851209962Smm error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx); 1852168404Spjd ASSERT(error == 0); 1853168404Spjd 1854168404Spjd /* 1855168404Spjd * Create root znode. Create minimal znode/vnode/zfsvfs 1856168404Spjd * to allow zfs_mknode to work. 1857168404Spjd */ 1858185029Spjd VATTR_NULL(&vattr); 1859168404Spjd vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE; 1860168404Spjd vattr.va_type = VDIR; 1861168404Spjd vattr.va_mode = S_IFDIR|0755; 1862185029Spjd vattr.va_uid = crgetuid(cr); 1863185029Spjd vattr.va_gid = crgetgid(cr); 1864168404Spjd 1865219089Spjd bzero(&zfsvfs, sizeof (zfsvfs_t)); 1866219089Spjd 1867168404Spjd rootzp = kmem_cache_alloc(znode_cache, KM_SLEEP); 1868219089Spjd ASSERT(!POINTER_IS_VALID(rootzp->z_zfsvfs)); 1869219089Spjd rootzp->z_moved = 0; 1870168404Spjd rootzp->z_unlinked = 0; 1871168404Spjd rootzp->z_atime_dirty = 0; 1872219089Spjd rootzp->z_is_sa = USE_SA(version, os); 1873168404Spjd 1874168404Spjd zfsvfs.z_os = os; 1875168404Spjd zfsvfs.z_parent = &zfsvfs; 1876185029Spjd zfsvfs.z_version = version; 1877185029Spjd zfsvfs.z_use_fuids = USE_FUIDS(version, os); 1878219089Spjd zfsvfs.z_use_sa = USE_SA(version, os); 1879185029Spjd zfsvfs.z_norm = norm; 1880219089Spjd 1881219089Spjd error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END, 1882219089Spjd &zfsvfs.z_attr_table); 1883219089Spjd 1884219089Spjd ASSERT(error == 0); 1885219089Spjd 1886185029Spjd /* 1887185029Spjd * Fold case on file systems that are always or sometimes case 1888185029Spjd * insensitive. 1889185029Spjd */ 1890185029Spjd if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED) 1891185029Spjd zfsvfs.z_norm |= U8_TEXTPREP_TOUPPER; 1892168404Spjd 1893168404Spjd mutex_init(&zfsvfs.z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); 1894168404Spjd list_create(&zfsvfs.z_all_znodes, sizeof (znode_t), 1895168404Spjd offsetof(znode_t, z_link_node)); 1896168404Spjd 1897207334Spjd for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) 1898207334Spjd mutex_init(&zfsvfs.z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL); 1899207334Spjd 1900185029Spjd rootzp->z_zfsvfs = &zfsvfs; 1901209962Smm VERIFY(0 == zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr, 1902209962Smm cr, NULL, &acl_ids)); 1903219089Spjd zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, &acl_ids); 1904185029Spjd ASSERT3P(zp, ==, rootzp); 1905185029Spjd error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx); 1906168404Spjd ASSERT(error == 0); 1907209962Smm zfs_acl_ids_free(&acl_ids); 1908185029Spjd POINTER_INVALIDATE(&rootzp->z_zfsvfs); 1909168404Spjd 1910219089Spjd sa_handle_destroy(rootzp->z_sa_hdl); 1911209962Smm kmem_cache_free(znode_cache, rootzp); 1912209962Smm 1913209962Smm /* 1914209962Smm * Create shares directory 1915209962Smm */ 1916209962Smm 1917209962Smm error = zfs_create_share_dir(&zfsvfs, tx); 1918209962Smm 1919209962Smm ASSERT(error == 0); 1920209962Smm 1921207334Spjd for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) 1922207334Spjd mutex_destroy(&zfsvfs.z_hold_mtx[i]); 1923168404Spjd} 1924185029Spjd 1925168404Spjd#endif /* _KERNEL */ 1926219089Spjd 1927168404Spjdstatic int 1928219089Spjdzfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table) 1929168404Spjd{ 1930219089Spjd uint64_t sa_obj = 0; 1931219089Spjd int error; 1932219089Spjd 1933219089Spjd error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj); 1934219089Spjd if (error != 0 && error != ENOENT) 1935219089Spjd return (error); 1936219089Spjd 1937219089Spjd error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table); 1938219089Spjd return (error); 1939219089Spjd} 1940219089Spjd 1941219089Spjdstatic int 1942219089Spjdzfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp, 1943219089Spjd dmu_buf_t **db, void *tag) 1944219089Spjd{ 1945168404Spjd dmu_object_info_t doi; 1946168404Spjd int error; 1947168404Spjd 1948219089Spjd if ((error = sa_buf_hold(osp, obj, tag, db)) != 0) 1949168404Spjd return (error); 1950168404Spjd 1951219089Spjd dmu_object_info_from_db(*db, &doi); 1952219089Spjd if ((doi.doi_bonus_type != DMU_OT_SA && 1953219089Spjd doi.doi_bonus_type != DMU_OT_ZNODE) || 1954219089Spjd doi.doi_bonus_type == DMU_OT_ZNODE && 1955168404Spjd doi.doi_bonus_size < sizeof (znode_phys_t)) { 1956219089Spjd sa_buf_rele(*db, tag); 1957249643Smm return (SET_ERROR(ENOTSUP)); 1958168404Spjd } 1959168404Spjd 1960219089Spjd error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp); 1961219089Spjd if (error != 0) { 1962219089Spjd sa_buf_rele(*db, tag); 1963219089Spjd return (error); 1964219089Spjd } 1965168404Spjd 1966168404Spjd return (0); 1967168404Spjd} 1968168404Spjd 1969219089Spjdvoid 1970219089Spjdzfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db, void *tag) 1971168404Spjd{ 1972219089Spjd sa_handle_destroy(hdl); 1973219089Spjd sa_buf_rele(db, tag); 1974219089Spjd} 1975219089Spjd 1976219089Spjd/* 1977219089Spjd * Given an object number, return its parent object number and whether 1978219089Spjd * or not the object is an extended attribute directory. 1979219089Spjd */ 1980219089Spjdstatic int 1981243674Smmzfs_obj_to_pobj(objset_t *osp, sa_handle_t *hdl, sa_attr_type_t *sa_table, 1982243674Smm uint64_t *pobjp, int *is_xattrdir) 1983219089Spjd{ 1984219089Spjd uint64_t parent; 1985219089Spjd uint64_t pflags; 1986219089Spjd uint64_t mode; 1987243674Smm uint64_t parent_mode; 1988219089Spjd sa_bulk_attr_t bulk[3]; 1989243674Smm sa_handle_t *sa_hdl; 1990243674Smm dmu_buf_t *sa_db; 1991219089Spjd int count = 0; 1992219089Spjd int error; 1993219089Spjd 1994219089Spjd SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL, 1995219089Spjd &parent, sizeof (parent)); 1996219089Spjd SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL, 1997219089Spjd &pflags, sizeof (pflags)); 1998219089Spjd SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL, 1999219089Spjd &mode, sizeof (mode)); 2000219089Spjd 2001219089Spjd if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0) 2002219089Spjd return (error); 2003219089Spjd 2004243674Smm /* 2005243674Smm * When a link is removed its parent pointer is not changed and will 2006243674Smm * be invalid. There are two cases where a link is removed but the 2007243674Smm * file stays around, when it goes to the delete queue and when there 2008243674Smm * are additional links. 2009243674Smm */ 2010243674Smm error = zfs_grab_sa_handle(osp, parent, &sa_hdl, &sa_db, FTAG); 2011243674Smm if (error != 0) 2012243674Smm return (error); 2013243674Smm 2014243674Smm error = sa_lookup(sa_hdl, ZPL_MODE, &parent_mode, sizeof (parent_mode)); 2015243674Smm zfs_release_sa_handle(sa_hdl, sa_db, FTAG); 2016243674Smm if (error != 0) 2017243674Smm return (error); 2018243674Smm 2019219089Spjd *is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode); 2020219089Spjd 2021243674Smm /* 2022243674Smm * Extended attributes can be applied to files, directories, etc. 2023243674Smm * Otherwise the parent must be a directory. 2024243674Smm */ 2025243674Smm if (!*is_xattrdir && !S_ISDIR(parent_mode)) 2026249643Smm return (SET_ERROR(EINVAL)); 2027243674Smm 2028243674Smm *pobjp = parent; 2029243674Smm 2030219089Spjd return (0); 2031219089Spjd} 2032219089Spjd 2033219089Spjd/* 2034219089Spjd * Given an object number, return some zpl level statistics 2035219089Spjd */ 2036219089Spjdstatic int 2037219089Spjdzfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table, 2038219089Spjd zfs_stat_t *sb) 2039219089Spjd{ 2040219089Spjd sa_bulk_attr_t bulk[4]; 2041219089Spjd int count = 0; 2042219089Spjd 2043219089Spjd SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL, 2044219089Spjd &sb->zs_mode, sizeof (sb->zs_mode)); 2045219089Spjd SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL, 2046219089Spjd &sb->zs_gen, sizeof (sb->zs_gen)); 2047219089Spjd SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL, 2048219089Spjd &sb->zs_links, sizeof (sb->zs_links)); 2049219089Spjd SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL, 2050219089Spjd &sb->zs_ctime, sizeof (sb->zs_ctime)); 2051219089Spjd 2052219089Spjd return (sa_bulk_lookup(hdl, bulk, count)); 2053219089Spjd} 2054219089Spjd 2055219089Spjdstatic int 2056219089Spjdzfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl, 2057219089Spjd sa_attr_type_t *sa_table, char *buf, int len) 2058219089Spjd{ 2059219089Spjd sa_handle_t *sa_hdl; 2060219089Spjd sa_handle_t *prevhdl = NULL; 2061219089Spjd dmu_buf_t *prevdb = NULL; 2062219089Spjd dmu_buf_t *sa_db = NULL; 2063168404Spjd char *path = buf + len - 1; 2064168404Spjd int error; 2065168404Spjd 2066168404Spjd *path = '\0'; 2067219089Spjd sa_hdl = hdl; 2068168404Spjd 2069168404Spjd for (;;) { 2070168404Spjd uint64_t pobj; 2071168404Spjd char component[MAXNAMELEN + 2]; 2072168404Spjd size_t complen; 2073168404Spjd int is_xattrdir; 2074168404Spjd 2075219089Spjd if (prevdb) 2076219089Spjd zfs_release_sa_handle(prevhdl, prevdb, FTAG); 2077219089Spjd 2078243674Smm if ((error = zfs_obj_to_pobj(osp, sa_hdl, sa_table, &pobj, 2079168404Spjd &is_xattrdir)) != 0) 2080168404Spjd break; 2081168404Spjd 2082168404Spjd if (pobj == obj) { 2083168404Spjd if (path[0] != '/') 2084168404Spjd *--path = '/'; 2085168404Spjd break; 2086168404Spjd } 2087168404Spjd 2088168404Spjd component[0] = '/'; 2089168404Spjd if (is_xattrdir) { 2090168404Spjd (void) sprintf(component + 1, "<xattrdir>"); 2091168404Spjd } else { 2092185029Spjd error = zap_value_search(osp, pobj, obj, 2093185029Spjd ZFS_DIRENT_OBJ(-1ULL), component + 1); 2094168404Spjd if (error != 0) 2095168404Spjd break; 2096168404Spjd } 2097168404Spjd 2098168404Spjd complen = strlen(component); 2099168404Spjd path -= complen; 2100168404Spjd ASSERT(path >= buf); 2101168404Spjd bcopy(component, path, complen); 2102168404Spjd obj = pobj; 2103219089Spjd 2104219089Spjd if (sa_hdl != hdl) { 2105219089Spjd prevhdl = sa_hdl; 2106219089Spjd prevdb = sa_db; 2107219089Spjd } 2108219089Spjd error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db, FTAG); 2109219089Spjd if (error != 0) { 2110219089Spjd sa_hdl = prevhdl; 2111219089Spjd sa_db = prevdb; 2112219089Spjd break; 2113219089Spjd } 2114168404Spjd } 2115168404Spjd 2116219089Spjd if (sa_hdl != NULL && sa_hdl != hdl) { 2117219089Spjd ASSERT(sa_db != NULL); 2118219089Spjd zfs_release_sa_handle(sa_hdl, sa_db, FTAG); 2119219089Spjd } 2120219089Spjd 2121168404Spjd if (error == 0) 2122168404Spjd (void) memmove(buf, path, buf + len - path); 2123219089Spjd 2124168404Spjd return (error); 2125168404Spjd} 2126219089Spjd 2127219089Spjdint 2128219089Spjdzfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len) 2129219089Spjd{ 2130219089Spjd sa_attr_type_t *sa_table; 2131219089Spjd sa_handle_t *hdl; 2132219089Spjd dmu_buf_t *db; 2133219089Spjd int error; 2134219089Spjd 2135219089Spjd error = zfs_sa_setup(osp, &sa_table); 2136219089Spjd if (error != 0) 2137219089Spjd return (error); 2138219089Spjd 2139219089Spjd error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG); 2140219089Spjd if (error != 0) 2141219089Spjd return (error); 2142219089Spjd 2143219089Spjd error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len); 2144219089Spjd 2145219089Spjd zfs_release_sa_handle(hdl, db, FTAG); 2146219089Spjd return (error); 2147219089Spjd} 2148219089Spjd 2149219089Spjdint 2150219089Spjdzfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb, 2151219089Spjd char *buf, int len) 2152219089Spjd{ 2153219089Spjd char *path = buf + len - 1; 2154219089Spjd sa_attr_type_t *sa_table; 2155219089Spjd sa_handle_t *hdl; 2156219089Spjd dmu_buf_t *db; 2157219089Spjd int error; 2158219089Spjd 2159219089Spjd *path = '\0'; 2160219089Spjd 2161219089Spjd error = zfs_sa_setup(osp, &sa_table); 2162219089Spjd if (error != 0) 2163219089Spjd return (error); 2164219089Spjd 2165219089Spjd error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG); 2166219089Spjd if (error != 0) 2167219089Spjd return (error); 2168219089Spjd 2169219089Spjd error = zfs_obj_to_stats_impl(hdl, sa_table, sb); 2170219089Spjd if (error != 0) { 2171219089Spjd zfs_release_sa_handle(hdl, db, FTAG); 2172219089Spjd return (error); 2173219089Spjd } 2174219089Spjd 2175219089Spjd error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len); 2176219089Spjd 2177219089Spjd zfs_release_sa_handle(hdl, db, FTAG); 2178219089Spjd return (error); 2179219089Spjd} 2180