zfs_znode.c revision 243520
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. 23240415Smm * Copyright (c) 2012 by Delphix. All rights reserved. 24168404Spjd */ 25168404Spjd 26169195Spjd/* Portions Copyright 2007 Jeremy Teo */ 27226732Smm/* 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 116168404Spjd/* 117168404Spjd * XXX: We cannot use this function as a cache constructor, because 118168404Spjd * there is one global cache for all file systems and we need 119168404Spjd * to pass vfsp here, which is not possible, because argument 120168404Spjd * 'cdrarg' is defined at kmem_cache_create() time. 121168404Spjd */ 122219089Spjd/*ARGSUSED*/ 123168404Spjdstatic int 124185029Spjdzfs_znode_cache_constructor(void *buf, void *arg, int kmflags) 125168404Spjd{ 126168404Spjd znode_t *zp = buf; 127169196Spjd vnode_t *vp; 128185029Spjd vfs_t *vfsp = arg; 129168404Spjd int error; 130168404Spjd 131185029Spjd POINTER_INVALIDATE(&zp->z_zfsvfs); 132185029Spjd ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs)); 133185029Spjd 134199156Spjd if (vfsp != NULL) { 135199156Spjd error = getnewvnode("zfs", vfsp, &zfs_vnodeops, &vp); 136199156Spjd if (error != 0 && (kmflags & KM_NOSLEEP)) 137199156Spjd return (-1); 138199156Spjd ASSERT(error == 0); 139199156Spjd vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 140199156Spjd zp->z_vnode = vp; 141199156Spjd vp->v_data = (caddr_t)zp; 142199156Spjd VN_LOCK_AREC(vp); 143219089Spjd VN_LOCK_ASHARE(vp); 144199156Spjd } else { 145199156Spjd zp->z_vnode = NULL; 146199156Spjd } 147185029Spjd 148185029Spjd list_link_init(&zp->z_link_node); 149185029Spjd 150168404Spjd mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL); 151168404Spjd rw_init(&zp->z_parent_lock, NULL, RW_DEFAULT, NULL); 152168404Spjd rw_init(&zp->z_name_lock, NULL, RW_DEFAULT, NULL); 153168404Spjd mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL); 154168404Spjd 155168404Spjd mutex_init(&zp->z_range_lock, NULL, MUTEX_DEFAULT, NULL); 156168404Spjd avl_create(&zp->z_range_avl, zfs_range_compare, 157168404Spjd sizeof (rl_t), offsetof(rl_t, r_node)); 158168404Spjd 159185029Spjd zp->z_dirlocks = NULL; 160211932Smm zp->z_acl_cached = NULL; 161219089Spjd zp->z_moved = 0; 162168404Spjd return (0); 163168404Spjd} 164168404Spjd 165168404Spjd/*ARGSUSED*/ 166168404Spjdstatic void 167185029Spjdzfs_znode_cache_destructor(void *buf, void *arg) 168168404Spjd{ 169168404Spjd znode_t *zp = buf; 170168404Spjd 171185029Spjd ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs)); 172185029Spjd ASSERT(ZTOV(zp) == NULL); 173185029Spjd vn_free(ZTOV(zp)); 174185029Spjd ASSERT(!list_link_active(&zp->z_link_node)); 175168404Spjd mutex_destroy(&zp->z_lock); 176168404Spjd rw_destroy(&zp->z_parent_lock); 177168404Spjd rw_destroy(&zp->z_name_lock); 178168404Spjd mutex_destroy(&zp->z_acl_lock); 179185029Spjd avl_destroy(&zp->z_range_avl); 180168404Spjd mutex_destroy(&zp->z_range_lock); 181168404Spjd 182185029Spjd ASSERT(zp->z_dirlocks == NULL); 183211932Smm ASSERT(zp->z_acl_cached == NULL); 184168404Spjd} 185168404Spjd 186185029Spjd#ifdef ZNODE_STATS 187185029Spjdstatic struct { 188185029Spjd uint64_t zms_zfsvfs_invalid; 189210470Smm uint64_t zms_zfsvfs_recheck1; 190185029Spjd uint64_t zms_zfsvfs_unmounted; 191210470Smm uint64_t zms_zfsvfs_recheck2; 192185029Spjd uint64_t zms_obj_held; 193185029Spjd uint64_t zms_vnode_locked; 194185029Spjd uint64_t zms_not_only_dnlc; 195185029Spjd} znode_move_stats; 196185029Spjd#endif /* ZNODE_STATS */ 197185029Spjd 198219089Spjd#ifdef sun 199185029Spjdstatic void 200185029Spjdzfs_znode_move_impl(znode_t *ozp, znode_t *nzp) 201185029Spjd{ 202185029Spjd vnode_t *vp; 203185029Spjd 204185029Spjd /* Copy fields. */ 205185029Spjd nzp->z_zfsvfs = ozp->z_zfsvfs; 206185029Spjd 207185029Spjd /* Swap vnodes. */ 208185029Spjd vp = nzp->z_vnode; 209185029Spjd nzp->z_vnode = ozp->z_vnode; 210185029Spjd ozp->z_vnode = vp; /* let destructor free the overwritten vnode */ 211185029Spjd ZTOV(ozp)->v_data = ozp; 212185029Spjd ZTOV(nzp)->v_data = nzp; 213185029Spjd 214185029Spjd nzp->z_id = ozp->z_id; 215185029Spjd ASSERT(ozp->z_dirlocks == NULL); /* znode not in use */ 216185029Spjd ASSERT(avl_numnodes(&ozp->z_range_avl) == 0); 217185029Spjd nzp->z_unlinked = ozp->z_unlinked; 218185029Spjd nzp->z_atime_dirty = ozp->z_atime_dirty; 219185029Spjd nzp->z_zn_prefetch = ozp->z_zn_prefetch; 220185029Spjd nzp->z_blksz = ozp->z_blksz; 221185029Spjd nzp->z_seq = ozp->z_seq; 222185029Spjd nzp->z_mapcnt = ozp->z_mapcnt; 223185029Spjd nzp->z_gen = ozp->z_gen; 224185029Spjd nzp->z_sync_cnt = ozp->z_sync_cnt; 225219089Spjd nzp->z_is_sa = ozp->z_is_sa; 226219089Spjd nzp->z_sa_hdl = ozp->z_sa_hdl; 227219089Spjd bcopy(ozp->z_atime, nzp->z_atime, sizeof (uint64_t) * 2); 228219089Spjd nzp->z_links = ozp->z_links; 229219089Spjd nzp->z_size = ozp->z_size; 230219089Spjd nzp->z_pflags = ozp->z_pflags; 231219089Spjd nzp->z_uid = ozp->z_uid; 232219089Spjd nzp->z_gid = ozp->z_gid; 233219089Spjd nzp->z_mode = ozp->z_mode; 234185029Spjd 235211932Smm /* 236211932Smm * Since this is just an idle znode and kmem is already dealing with 237211932Smm * memory pressure, release any cached ACL. 238211932Smm */ 239211932Smm if (ozp->z_acl_cached) { 240211932Smm zfs_acl_free(ozp->z_acl_cached); 241211932Smm ozp->z_acl_cached = NULL; 242211932Smm } 243211932Smm 244219089Spjd sa_set_userp(nzp->z_sa_hdl, nzp); 245185029Spjd 246185029Spjd /* 247185029Spjd * Invalidate the original znode by clearing fields that provide a 248185029Spjd * pointer back to the znode. Set the low bit of the vfs pointer to 249185029Spjd * ensure that zfs_znode_move() recognizes the znode as invalid in any 250185029Spjd * subsequent callback. 251185029Spjd */ 252219089Spjd ozp->z_sa_hdl = NULL; 253185029Spjd POINTER_INVALIDATE(&ozp->z_zfsvfs); 254219089Spjd 255219089Spjd /* 256219089Spjd * Mark the znode. 257219089Spjd */ 258219089Spjd nzp->z_moved = 1; 259219089Spjd ozp->z_moved = (uint8_t)-1; 260185029Spjd} 261185029Spjd 262185029Spjd/*ARGSUSED*/ 263185029Spjdstatic kmem_cbrc_t 264185029Spjdzfs_znode_move(void *buf, void *newbuf, size_t size, void *arg) 265185029Spjd{ 266185029Spjd znode_t *ozp = buf, *nzp = newbuf; 267185029Spjd zfsvfs_t *zfsvfs; 268185029Spjd vnode_t *vp; 269185029Spjd 270185029Spjd /* 271185029Spjd * The znode is on the file system's list of known znodes if the vfs 272185029Spjd * pointer is valid. We set the low bit of the vfs pointer when freeing 273185029Spjd * the znode to invalidate it, and the memory patterns written by kmem 274185029Spjd * (baddcafe and deadbeef) set at least one of the two low bits. A newly 275185029Spjd * created znode sets the vfs pointer last of all to indicate that the 276185029Spjd * znode is known and in a valid state to be moved by this function. 277185029Spjd */ 278185029Spjd zfsvfs = ozp->z_zfsvfs; 279185029Spjd if (!POINTER_IS_VALID(zfsvfs)) { 280185029Spjd ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_invalid); 281185029Spjd return (KMEM_CBRC_DONT_KNOW); 282185029Spjd } 283185029Spjd 284185029Spjd /* 285210470Smm * Close a small window in which it's possible that the filesystem could 286210470Smm * be unmounted and freed, and zfsvfs, though valid in the previous 287210470Smm * statement, could point to unrelated memory by the time we try to 288210470Smm * prevent the filesystem from being unmounted. 289185029Spjd */ 290210470Smm rw_enter(&zfsvfs_lock, RW_WRITER); 291210470Smm if (zfsvfs != ozp->z_zfsvfs) { 292210470Smm rw_exit(&zfsvfs_lock); 293210470Smm ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_recheck1); 294210470Smm return (KMEM_CBRC_DONT_KNOW); 295210470Smm } 296210470Smm 297210470Smm /* 298210470Smm * If the znode is still valid, then so is the file system. We know that 299210470Smm * no valid file system can be freed while we hold zfsvfs_lock, so we 300210470Smm * can safely ensure that the filesystem is not and will not be 301210470Smm * unmounted. The next statement is equivalent to ZFS_ENTER(). 302210470Smm */ 303209962Smm rrw_enter(&zfsvfs->z_teardown_lock, RW_READER, FTAG); 304209962Smm if (zfsvfs->z_unmounted) { 305209962Smm ZFS_EXIT(zfsvfs); 306210470Smm rw_exit(&zfsvfs_lock); 307185029Spjd ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_unmounted); 308185029Spjd return (KMEM_CBRC_DONT_KNOW); 309185029Spjd } 310210470Smm rw_exit(&zfsvfs_lock); 311185029Spjd 312185029Spjd mutex_enter(&zfsvfs->z_znodes_lock); 313185029Spjd /* 314185029Spjd * Recheck the vfs pointer in case the znode was removed just before 315185029Spjd * acquiring the lock. 316185029Spjd */ 317185029Spjd if (zfsvfs != ozp->z_zfsvfs) { 318185029Spjd mutex_exit(&zfsvfs->z_znodes_lock); 319185029Spjd ZFS_EXIT(zfsvfs); 320210470Smm ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_recheck2); 321185029Spjd return (KMEM_CBRC_DONT_KNOW); 322185029Spjd } 323185029Spjd 324185029Spjd /* 325185029Spjd * At this point we know that as long as we hold z_znodes_lock, the 326185029Spjd * znode cannot be freed and fields within the znode can be safely 327185029Spjd * accessed. Now, prevent a race with zfs_zget(). 328185029Spjd */ 329185029Spjd if (ZFS_OBJ_HOLD_TRYENTER(zfsvfs, ozp->z_id) == 0) { 330185029Spjd mutex_exit(&zfsvfs->z_znodes_lock); 331185029Spjd ZFS_EXIT(zfsvfs); 332185029Spjd ZNODE_STAT_ADD(znode_move_stats.zms_obj_held); 333185029Spjd return (KMEM_CBRC_LATER); 334185029Spjd } 335185029Spjd 336185029Spjd vp = ZTOV(ozp); 337185029Spjd if (mutex_tryenter(&vp->v_lock) == 0) { 338185029Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id); 339185029Spjd mutex_exit(&zfsvfs->z_znodes_lock); 340185029Spjd ZFS_EXIT(zfsvfs); 341185029Spjd ZNODE_STAT_ADD(znode_move_stats.zms_vnode_locked); 342185029Spjd return (KMEM_CBRC_LATER); 343185029Spjd } 344185029Spjd 345185029Spjd /* Only move znodes that are referenced _only_ by the DNLC. */ 346185029Spjd if (vp->v_count != 1 || !vn_in_dnlc(vp)) { 347185029Spjd mutex_exit(&vp->v_lock); 348185029Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id); 349185029Spjd mutex_exit(&zfsvfs->z_znodes_lock); 350185029Spjd ZFS_EXIT(zfsvfs); 351185029Spjd ZNODE_STAT_ADD(znode_move_stats.zms_not_only_dnlc); 352185029Spjd return (KMEM_CBRC_LATER); 353185029Spjd } 354185029Spjd 355185029Spjd /* 356185029Spjd * The znode is known and in a valid state to move. We're holding the 357185029Spjd * locks needed to execute the critical section. 358185029Spjd */ 359185029Spjd zfs_znode_move_impl(ozp, nzp); 360185029Spjd mutex_exit(&vp->v_lock); 361185029Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id); 362185029Spjd 363185029Spjd list_link_replace(&ozp->z_link_node, &nzp->z_link_node); 364185029Spjd mutex_exit(&zfsvfs->z_znodes_lock); 365185029Spjd ZFS_EXIT(zfsvfs); 366185029Spjd 367185029Spjd return (KMEM_CBRC_YES); 368185029Spjd} 369185029Spjd#endif /* sun */ 370185029Spjd 371168404Spjdvoid 372168404Spjdzfs_znode_init(void) 373168404Spjd{ 374168404Spjd /* 375168404Spjd * Initialize zcache 376168404Spjd */ 377210470Smm rw_init(&zfsvfs_lock, NULL, RW_DEFAULT, NULL); 378168404Spjd ASSERT(znode_cache == NULL); 379168404Spjd znode_cache = kmem_cache_create("zfs_znode_cache", 380168404Spjd sizeof (znode_t), 0, /* zfs_znode_cache_constructor */ NULL, 381168404Spjd zfs_znode_cache_destructor, NULL, NULL, NULL, 0); 382185029Spjd kmem_cache_set_move(znode_cache, zfs_znode_move); 383168404Spjd} 384168404Spjd 385168404Spjdvoid 386168404Spjdzfs_znode_fini(void) 387168404Spjd{ 388219089Spjd#ifdef sun 389168404Spjd /* 390219089Spjd * Cleanup vfs & vnode ops 391219089Spjd */ 392219089Spjd zfs_remove_op_tables(); 393219089Spjd#endif /* sun */ 394219089Spjd 395219089Spjd /* 396168404Spjd * Cleanup zcache 397168404Spjd */ 398168404Spjd if (znode_cache) 399168404Spjd kmem_cache_destroy(znode_cache); 400168404Spjd znode_cache = NULL; 401210470Smm rw_destroy(&zfsvfs_lock); 402168404Spjd} 403168404Spjd 404219089Spjd#ifdef sun 405219089Spjdstruct vnodeops *zfs_dvnodeops; 406219089Spjdstruct vnodeops *zfs_fvnodeops; 407219089Spjdstruct vnodeops *zfs_symvnodeops; 408219089Spjdstruct vnodeops *zfs_xdvnodeops; 409219089Spjdstruct vnodeops *zfs_evnodeops; 410219089Spjdstruct vnodeops *zfs_sharevnodeops; 411219089Spjd 412219089Spjdvoid 413219089Spjdzfs_remove_op_tables() 414219089Spjd{ 415219089Spjd /* 416219089Spjd * Remove vfs ops 417219089Spjd */ 418219089Spjd ASSERT(zfsfstype); 419219089Spjd (void) vfs_freevfsops_by_type(zfsfstype); 420219089Spjd zfsfstype = 0; 421219089Spjd 422219089Spjd /* 423219089Spjd * Remove vnode ops 424219089Spjd */ 425219089Spjd if (zfs_dvnodeops) 426219089Spjd vn_freevnodeops(zfs_dvnodeops); 427219089Spjd if (zfs_fvnodeops) 428219089Spjd vn_freevnodeops(zfs_fvnodeops); 429219089Spjd if (zfs_symvnodeops) 430219089Spjd vn_freevnodeops(zfs_symvnodeops); 431219089Spjd if (zfs_xdvnodeops) 432219089Spjd vn_freevnodeops(zfs_xdvnodeops); 433219089Spjd if (zfs_evnodeops) 434219089Spjd vn_freevnodeops(zfs_evnodeops); 435219089Spjd if (zfs_sharevnodeops) 436219089Spjd vn_freevnodeops(zfs_sharevnodeops); 437219089Spjd 438219089Spjd zfs_dvnodeops = NULL; 439219089Spjd zfs_fvnodeops = NULL; 440219089Spjd zfs_symvnodeops = NULL; 441219089Spjd zfs_xdvnodeops = NULL; 442219089Spjd zfs_evnodeops = NULL; 443219089Spjd zfs_sharevnodeops = NULL; 444219089Spjd} 445219089Spjd 446219089Spjdextern const fs_operation_def_t zfs_dvnodeops_template[]; 447219089Spjdextern const fs_operation_def_t zfs_fvnodeops_template[]; 448219089Spjdextern const fs_operation_def_t zfs_xdvnodeops_template[]; 449219089Spjdextern const fs_operation_def_t zfs_symvnodeops_template[]; 450219089Spjdextern const fs_operation_def_t zfs_evnodeops_template[]; 451219089Spjdextern const fs_operation_def_t zfs_sharevnodeops_template[]; 452219089Spjd 453168404Spjdint 454219089Spjdzfs_create_op_tables() 455219089Spjd{ 456219089Spjd int error; 457219089Spjd 458219089Spjd /* 459219089Spjd * zfs_dvnodeops can be set if mod_remove() calls mod_installfs() 460219089Spjd * due to a failure to remove the the 2nd modlinkage (zfs_modldrv). 461219089Spjd * In this case we just return as the ops vectors are already set up. 462219089Spjd */ 463219089Spjd if (zfs_dvnodeops) 464219089Spjd return (0); 465219089Spjd 466219089Spjd error = vn_make_ops(MNTTYPE_ZFS, zfs_dvnodeops_template, 467219089Spjd &zfs_dvnodeops); 468219089Spjd if (error) 469219089Spjd return (error); 470219089Spjd 471219089Spjd error = vn_make_ops(MNTTYPE_ZFS, zfs_fvnodeops_template, 472219089Spjd &zfs_fvnodeops); 473219089Spjd if (error) 474219089Spjd return (error); 475219089Spjd 476219089Spjd error = vn_make_ops(MNTTYPE_ZFS, zfs_symvnodeops_template, 477219089Spjd &zfs_symvnodeops); 478219089Spjd if (error) 479219089Spjd return (error); 480219089Spjd 481219089Spjd error = vn_make_ops(MNTTYPE_ZFS, zfs_xdvnodeops_template, 482219089Spjd &zfs_xdvnodeops); 483219089Spjd if (error) 484219089Spjd return (error); 485219089Spjd 486219089Spjd error = vn_make_ops(MNTTYPE_ZFS, zfs_evnodeops_template, 487219089Spjd &zfs_evnodeops); 488219089Spjd if (error) 489219089Spjd return (error); 490219089Spjd 491219089Spjd error = vn_make_ops(MNTTYPE_ZFS, zfs_sharevnodeops_template, 492219089Spjd &zfs_sharevnodeops); 493219089Spjd 494219089Spjd return (error); 495219089Spjd} 496219089Spjd#endif /* sun */ 497219089Spjd 498219089Spjdint 499209962Smmzfs_create_share_dir(zfsvfs_t *zfsvfs, dmu_tx_t *tx) 500168404Spjd{ 501209962Smm zfs_acl_ids_t acl_ids; 502209962Smm vattr_t vattr; 503209962Smm znode_t *sharezp; 504209962Smm vnode_t *vp, vnode; 505209962Smm znode_t *zp; 506209962Smm int error; 507168404Spjd 508209962Smm vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE; 509209962Smm vattr.va_type = VDIR; 510209962Smm vattr.va_mode = S_IFDIR|0555; 511209962Smm vattr.va_uid = crgetuid(kcred); 512209962Smm vattr.va_gid = crgetgid(kcred); 513168404Spjd 514209962Smm sharezp = kmem_cache_alloc(znode_cache, KM_SLEEP); 515209962Smm zfs_znode_cache_constructor(sharezp, zfsvfs->z_parent->z_vfs, 0); 516219089Spjd ASSERT(!POINTER_IS_VALID(sharezp->z_zfsvfs)); 517219089Spjd sharezp->z_moved = 0; 518209962Smm sharezp->z_unlinked = 0; 519209962Smm sharezp->z_atime_dirty = 0; 520209962Smm sharezp->z_zfsvfs = zfsvfs; 521219089Spjd sharezp->z_is_sa = zfsvfs->z_use_sa; 522168404Spjd 523209962Smm sharezp->z_vnode = &vnode; 524209962Smm vnode.v_data = sharezp; 525185029Spjd 526209962Smm vp = ZTOV(sharezp); 527209962Smm vp->v_type = VDIR; 528168404Spjd 529209962Smm VERIFY(0 == zfs_acl_ids_create(sharezp, IS_ROOT_NODE, &vattr, 530209962Smm kcred, NULL, &acl_ids)); 531219089Spjd zfs_mknode(sharezp, &vattr, tx, kcred, IS_ROOT_NODE, &zp, &acl_ids); 532209962Smm ASSERT3P(zp, ==, sharezp); 533209962Smm POINTER_INVALIDATE(&sharezp->z_zfsvfs); 534209962Smm error = zap_add(zfsvfs->z_os, MASTER_NODE_OBJ, 535209962Smm ZFS_SHARES_DIR, 8, 1, &sharezp->z_id, tx); 536209962Smm zfsvfs->z_shares_dir = sharezp->z_id; 537168404Spjd 538209962Smm zfs_acl_ids_free(&acl_ids); 539209962Smm ZTOV(sharezp)->v_data = NULL; 540209962Smm ZTOV(sharezp)->v_count = 0; 541209962Smm ZTOV(sharezp)->v_holdcnt = 0; 542209962Smm zp->z_vnode = NULL; 543219089Spjd sa_handle_destroy(sharezp->z_sa_hdl); 544209962Smm sharezp->z_vnode = NULL; 545209962Smm kmem_cache_free(znode_cache, sharezp); 546168404Spjd 547209962Smm return (error); 548168404Spjd} 549168404Spjd 550168404Spjd/* 551168404Spjd * define a couple of values we need available 552168404Spjd * for both 64 and 32 bit environments. 553168404Spjd */ 554168404Spjd#ifndef NBITSMINOR64 555168404Spjd#define NBITSMINOR64 32 556168404Spjd#endif 557168404Spjd#ifndef MAXMAJ64 558168404Spjd#define MAXMAJ64 0xffffffffUL 559168404Spjd#endif 560168404Spjd#ifndef MAXMIN64 561168404Spjd#define MAXMIN64 0xffffffffUL 562168404Spjd#endif 563168404Spjd 564168404Spjd/* 565168404Spjd * Create special expldev for ZFS private use. 566168404Spjd * Can't use standard expldev since it doesn't do 567168404Spjd * what we want. The standard expldev() takes a 568168404Spjd * dev32_t in LP64 and expands it to a long dev_t. 569168404Spjd * We need an interface that takes a dev32_t in ILP32 570168404Spjd * and expands it to a long dev_t. 571168404Spjd */ 572168404Spjdstatic uint64_t 573168404Spjdzfs_expldev(dev_t dev) 574168404Spjd{ 575187830Sed return (((uint64_t)major(dev) << NBITSMINOR64) | minor(dev)); 576168404Spjd} 577168404Spjd/* 578168404Spjd * Special cmpldev for ZFS private use. 579168404Spjd * Can't use standard cmpldev since it takes 580168404Spjd * a long dev_t and compresses it to dev32_t in 581168404Spjd * LP64. We need to do a compaction of a long dev_t 582168404Spjd * to a dev32_t in ILP32. 583168404Spjd */ 584168404Spjddev_t 585168404Spjdzfs_cmpldev(uint64_t dev) 586168404Spjd{ 587168958Spjd return (makedev((dev >> NBITSMINOR64), (dev & MAXMIN64))); 588168404Spjd} 589168404Spjd 590185029Spjdstatic void 591219089Spjdzfs_znode_sa_init(zfsvfs_t *zfsvfs, znode_t *zp, 592219089Spjd dmu_buf_t *db, dmu_object_type_t obj_type, sa_handle_t *sa_hdl) 593185029Spjd{ 594185029Spjd ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs) || (zfsvfs == zp->z_zfsvfs)); 595185029Spjd ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id))); 596185029Spjd 597185029Spjd mutex_enter(&zp->z_lock); 598185029Spjd 599219089Spjd ASSERT(zp->z_sa_hdl == NULL); 600211932Smm ASSERT(zp->z_acl_cached == NULL); 601219089Spjd if (sa_hdl == NULL) { 602219089Spjd VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, zp, 603219089Spjd SA_HDL_SHARED, &zp->z_sa_hdl)); 604219089Spjd } else { 605219089Spjd zp->z_sa_hdl = sa_hdl; 606219089Spjd sa_set_userp(sa_hdl, zp); 607219089Spjd } 608185029Spjd 609219089Spjd zp->z_is_sa = (obj_type == DMU_OT_SA) ? B_TRUE : B_FALSE; 610185029Spjd 611185029Spjd /* 612185029Spjd * Slap on VROOT if we are the root znode 613185029Spjd */ 614185029Spjd if (zp->z_id == zfsvfs->z_root) 615185029Spjd ZTOV(zp)->v_flag |= VROOT; 616185029Spjd 617185029Spjd mutex_exit(&zp->z_lock); 618185029Spjd vn_exists(ZTOV(zp)); 619185029Spjd} 620185029Spjd 621185029Spjdvoid 622185029Spjdzfs_znode_dmu_fini(znode_t *zp) 623185029Spjd{ 624185029Spjd ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zp->z_zfsvfs, zp->z_id)) || 625185029Spjd zp->z_unlinked || 626185029Spjd RW_WRITE_HELD(&zp->z_zfsvfs->z_teardown_inactive_lock)); 627219089Spjd 628219089Spjd sa_handle_destroy(zp->z_sa_hdl); 629219089Spjd zp->z_sa_hdl = NULL; 630185029Spjd} 631185029Spjd 632225153Spjdstatic void 633225153Spjdzfs_vnode_forget(vnode_t *vp) 634225153Spjd{ 635225153Spjd 636242571Savg /* copied from insmntque_stddtr */ 637225153Spjd vp->v_data = NULL; 638242571Savg vp->v_op = &dead_vnodeops; 639242571Savg vgone(vp); 640242571Savg vput(vp); 641225153Spjd} 642225153Spjd 643168404Spjd/* 644168404Spjd * Construct a new znode/vnode and intialize. 645168404Spjd * 646168404Spjd * This does not do a call to dmu_set_user() that is 647168404Spjd * up to the caller to do, in case you don't want to 648168404Spjd * return the znode 649168404Spjd */ 650168404Spjdstatic znode_t * 651219089Spjdzfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz, 652219089Spjd dmu_object_type_t obj_type, sa_handle_t *hdl) 653168404Spjd{ 654168404Spjd znode_t *zp; 655168404Spjd vnode_t *vp; 656219089Spjd uint64_t mode; 657219089Spjd uint64_t parent; 658219089Spjd sa_bulk_attr_t bulk[9]; 659219089Spjd int count = 0; 660168404Spjd 661168404Spjd zp = kmem_cache_alloc(znode_cache, KM_SLEEP); 662185029Spjd zfs_znode_cache_constructor(zp, zfsvfs->z_parent->z_vfs, 0); 663168404Spjd 664168404Spjd ASSERT(zp->z_dirlocks == NULL); 665185029Spjd ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs)); 666219089Spjd zp->z_moved = 0; 667168404Spjd 668185029Spjd /* 669185029Spjd * Defer setting z_zfsvfs until the znode is ready to be a candidate for 670185029Spjd * the zfs_znode_move() callback. 671185029Spjd */ 672219089Spjd zp->z_sa_hdl = NULL; 673168404Spjd zp->z_unlinked = 0; 674168404Spjd zp->z_atime_dirty = 0; 675168404Spjd zp->z_mapcnt = 0; 676185029Spjd zp->z_id = db->db_object; 677168404Spjd zp->z_blksz = blksz; 678168404Spjd zp->z_seq = 0x7A4653; 679168404Spjd zp->z_sync_cnt = 0; 680168404Spjd 681185029Spjd vp = ZTOV(zp); 682168404Spjd 683219089Spjd zfs_znode_sa_init(zfsvfs, zp, db, obj_type, hdl); 684185029Spjd 685219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8); 686219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, &zp->z_gen, 8); 687219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, 688219089Spjd &zp->z_size, 8); 689219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL, 690219089Spjd &zp->z_links, 8); 691219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, 692219089Spjd &zp->z_pflags, 8); 693219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8); 694219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, 695219089Spjd &zp->z_atime, 16); 696219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, 697219089Spjd &zp->z_uid, 8); 698219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, 699219089Spjd &zp->z_gid, 8); 700185029Spjd 701219089Spjd if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0) { 702219089Spjd if (hdl == NULL) 703219089Spjd sa_handle_destroy(zp->z_sa_hdl); 704225153Spjd zfs_vnode_forget(vp); 705225153Spjd zp->z_vnode = NULL; 706219089Spjd kmem_cache_free(znode_cache, zp); 707219089Spjd return (NULL); 708219089Spjd } 709168404Spjd 710219089Spjd zp->z_mode = mode; 711219089Spjd 712219089Spjd vp->v_type = IFTOVT((mode_t)mode); 713219089Spjd 714168404Spjd switch (vp->v_type) { 715168404Spjd case VDIR: 716168404Spjd zp->z_zn_prefetch = B_TRUE; /* z_prefetch default is enabled */ 717168404Spjd break; 718224251Sdelphij#ifdef sun 719224231Sdelphij case VBLK: 720224231Sdelphij case VCHR: 721224231Sdelphij { 722224231Sdelphij uint64_t rdev; 723224231Sdelphij VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_RDEV(zfsvfs), 724224231Sdelphij &rdev, sizeof (rdev)) == 0); 725224231Sdelphij 726224251Sdelphij vp->v_rdev = zfs_cmpldev(rdev); 727224231Sdelphij } 728224231Sdelphij break; 729224251Sdelphij#endif /* sun */ 730168404Spjd case VFIFO: 731224252Sdelphij#ifdef sun 732224252Sdelphij case VSOCK: 733224252Sdelphij case VDOOR: 734224252Sdelphij#endif /* sun */ 735168404Spjd vp->v_op = &zfs_fifoops; 736168404Spjd break; 737219089Spjd case VREG: 738219089Spjd if (parent == zfsvfs->z_shares_dir) { 739219089Spjd ASSERT(zp->z_uid == 0 && zp->z_gid == 0); 740209962Smm vp->v_op = &zfs_shareops; 741209962Smm } 742209962Smm break; 743224252Sdelphij#ifdef sun 744224252Sdelphij case VLNK: 745224252Sdelphij vn_setops(vp, zfs_symvnodeops); 746224252Sdelphij break; 747224252Sdelphij default: 748224252Sdelphij vn_setops(vp, zfs_evnodeops); 749224252Sdelphij break; 750224252Sdelphij#endif /* sun */ 751168404Spjd } 752210172Sjhb if (vp->v_type != VFIFO) 753189696Sjhb VN_LOCK_ASHARE(vp); 754168404Spjd 755185029Spjd mutex_enter(&zfsvfs->z_znodes_lock); 756185029Spjd list_insert_tail(&zfsvfs->z_all_znodes, zp); 757185029Spjd membar_producer(); 758168404Spjd /* 759185029Spjd * Everything else must be valid before assigning z_zfsvfs makes the 760185029Spjd * znode eligible for zfs_znode_move(). 761168404Spjd */ 762185029Spjd zp->z_zfsvfs = zfsvfs; 763185029Spjd mutex_exit(&zfsvfs->z_znodes_lock); 764168404Spjd 765168404Spjd VFS_HOLD(zfsvfs->z_vfs); 766185029Spjd return (zp); 767168404Spjd} 768168404Spjd 769219089Spjdstatic uint64_t empty_xattr; 770219089Spjdstatic uint64_t pad[4]; 771219089Spjdstatic zfs_acl_phys_t acl_phys; 772168404Spjd/* 773168404Spjd * Create a new DMU object to hold a zfs znode. 774168404Spjd * 775168404Spjd * IN: dzp - parent directory for new znode 776168404Spjd * vap - file attributes for new znode 777168404Spjd * tx - dmu transaction id for zap operations 778168404Spjd * cr - credentials of caller 779168404Spjd * flag - flags: 780168404Spjd * IS_ROOT_NODE - new object will be root 781168404Spjd * IS_XATTR - new object is an attribute 782185029Spjd * bonuslen - length of bonus buffer 783185029Spjd * setaclp - File/Dir initial ACL 784185029Spjd * fuidp - Tracks fuid allocation. 785168404Spjd * 786185029Spjd * OUT: zpp - allocated znode 787168404Spjd * 788168404Spjd */ 789168404Spjdvoid 790185029Spjdzfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr, 791219089Spjd uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids) 792168404Spjd{ 793219089Spjd uint64_t crtime[2], atime[2], mtime[2], ctime[2]; 794219089Spjd uint64_t mode, size, links, parent, pflags; 795219089Spjd uint64_t dzp_pflags = 0; 796219089Spjd uint64_t rdev = 0; 797219089Spjd zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 798185029Spjd dmu_buf_t *db; 799168404Spjd timestruc_t now; 800185029Spjd uint64_t gen, obj; 801168404Spjd int err; 802219089Spjd int bonuslen; 803219089Spjd sa_handle_t *sa_hdl; 804219089Spjd dmu_object_type_t obj_type; 805219089Spjd sa_bulk_attr_t sa_attrs[ZPL_END]; 806219089Spjd int cnt = 0; 807219089Spjd zfs_acl_locator_cb_t locate = { 0 }; 808168404Spjd 809168404Spjd ASSERT(vap && (vap->va_mask & (AT_TYPE|AT_MODE)) == (AT_TYPE|AT_MODE)); 810168404Spjd 811209962Smm if (zfsvfs->z_replay) { 812185029Spjd obj = vap->va_nodeid; 813168404Spjd now = vap->va_ctime; /* see zfs_replay_create() */ 814168404Spjd gen = vap->va_nblocks; /* ditto */ 815168404Spjd } else { 816185029Spjd obj = 0; 817168404Spjd gethrestime(&now); 818168404Spjd gen = dmu_tx_get_txg(tx); 819168404Spjd } 820168404Spjd 821219089Spjd obj_type = zfsvfs->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE; 822219089Spjd bonuslen = (obj_type == DMU_OT_SA) ? 823219089Spjd DN_MAX_BONUSLEN : ZFS_OLD_ZNODE_PHYS_SIZE; 824219089Spjd 825168404Spjd /* 826168404Spjd * Create a new DMU object. 827168404Spjd */ 828168404Spjd /* 829168404Spjd * There's currently no mechanism for pre-reading the blocks that will 830219089Spjd * be needed to allocate a new object, so we accept the small chance 831168404Spjd * that there will be an i/o error and we will fail one of the 832168404Spjd * assertions below. 833168404Spjd */ 834168404Spjd if (vap->va_type == VDIR) { 835209962Smm if (zfsvfs->z_replay) { 836185029Spjd err = zap_create_claim_norm(zfsvfs->z_os, obj, 837185029Spjd zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS, 838219089Spjd obj_type, bonuslen, tx); 839240415Smm ASSERT0(err); 840168404Spjd } else { 841185029Spjd obj = zap_create_norm(zfsvfs->z_os, 842185029Spjd zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS, 843219089Spjd obj_type, bonuslen, tx); 844168404Spjd } 845168404Spjd } else { 846209962Smm if (zfsvfs->z_replay) { 847185029Spjd err = dmu_object_claim(zfsvfs->z_os, obj, 848168404Spjd DMU_OT_PLAIN_FILE_CONTENTS, 0, 849219089Spjd obj_type, bonuslen, tx); 850240415Smm ASSERT0(err); 851168404Spjd } else { 852185029Spjd obj = dmu_object_alloc(zfsvfs->z_os, 853168404Spjd DMU_OT_PLAIN_FILE_CONTENTS, 0, 854219089Spjd obj_type, bonuslen, tx); 855168404Spjd } 856168404Spjd } 857207334Spjd 858241628Savg getnewvnode_reserve(1); 859207334Spjd ZFS_OBJ_HOLD_ENTER(zfsvfs, obj); 860219089Spjd VERIFY(0 == sa_buf_hold(zfsvfs->z_os, obj, NULL, &db)); 861168404Spjd 862168404Spjd /* 863168404Spjd * If this is the root, fix up the half-initialized parent pointer 864168404Spjd * to reference the just-allocated physical data area. 865168404Spjd */ 866168404Spjd if (flag & IS_ROOT_NODE) { 867185029Spjd dzp->z_id = obj; 868219089Spjd } else { 869219089Spjd dzp_pflags = dzp->z_pflags; 870168404Spjd } 871168404Spjd 872168404Spjd /* 873168404Spjd * If parent is an xattr, so am I. 874168404Spjd */ 875219089Spjd if (dzp_pflags & ZFS_XATTR) { 876168404Spjd flag |= IS_XATTR; 877168404Spjd } 878168404Spjd 879185029Spjd if (zfsvfs->z_use_fuids) 880219089Spjd pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED; 881219089Spjd else 882219089Spjd pflags = 0; 883185029Spjd 884168404Spjd if (vap->va_type == VDIR) { 885219089Spjd size = 2; /* contents ("." and "..") */ 886219089Spjd links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1; 887219089Spjd } else { 888219089Spjd size = links = 0; 889168404Spjd } 890168404Spjd 891219089Spjd if (vap->va_type == VBLK || vap->va_type == VCHR) { 892219089Spjd rdev = zfs_expldev(vap->va_rdev); 893219089Spjd } 894219089Spjd 895219089Spjd parent = dzp->z_id; 896219089Spjd mode = acl_ids->z_mode; 897168404Spjd if (flag & IS_XATTR) 898219089Spjd pflags |= ZFS_XATTR; 899168404Spjd 900219089Spjd /* 901219089Spjd * No execs denied will be deterimed when zfs_mode_compute() is called. 902219089Spjd */ 903219089Spjd pflags |= acl_ids->z_aclp->z_hints & 904219089Spjd (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT| 905219089Spjd ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED); 906168404Spjd 907219089Spjd ZFS_TIME_ENCODE(&now, crtime); 908219089Spjd ZFS_TIME_ENCODE(&now, ctime); 909168404Spjd 910168404Spjd if (vap->va_mask & AT_ATIME) { 911219089Spjd ZFS_TIME_ENCODE(&vap->va_atime, atime); 912168404Spjd } else { 913219089Spjd ZFS_TIME_ENCODE(&now, atime); 914168404Spjd } 915168404Spjd 916168404Spjd if (vap->va_mask & AT_MTIME) { 917219089Spjd ZFS_TIME_ENCODE(&vap->va_mtime, mtime); 918168404Spjd } else { 919219089Spjd ZFS_TIME_ENCODE(&now, mtime); 920168404Spjd } 921168404Spjd 922219089Spjd /* Now add in all of the "SA" attributes */ 923219089Spjd VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, NULL, SA_HDL_SHARED, 924219089Spjd &sa_hdl)); 925219089Spjd 926219089Spjd /* 927219089Spjd * Setup the array of attributes to be replaced/set on the new file 928219089Spjd * 929219089Spjd * order for DMU_OT_ZNODE is critical since it needs to be constructed 930219089Spjd * in the old znode_phys_t format. Don't change this ordering 931219089Spjd */ 932219089Spjd 933219089Spjd if (obj_type == DMU_OT_ZNODE) { 934219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs), 935219089Spjd NULL, &atime, 16); 936219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs), 937219089Spjd NULL, &mtime, 16); 938219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs), 939219089Spjd NULL, &ctime, 16); 940219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs), 941219089Spjd NULL, &crtime, 16); 942219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs), 943219089Spjd NULL, &gen, 8); 944219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs), 945219089Spjd NULL, &mode, 8); 946219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs), 947219089Spjd NULL, &size, 8); 948219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs), 949219089Spjd NULL, &parent, 8); 950219089Spjd } else { 951219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs), 952219089Spjd NULL, &mode, 8); 953219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs), 954219089Spjd NULL, &size, 8); 955219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs), 956219089Spjd NULL, &gen, 8); 957219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL, 958219089Spjd &acl_ids->z_fuid, 8); 959219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL, 960219089Spjd &acl_ids->z_fgid, 8); 961219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs), 962219089Spjd NULL, &parent, 8); 963219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs), 964219089Spjd NULL, &pflags, 8); 965219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs), 966219089Spjd NULL, &atime, 16); 967219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs), 968219089Spjd NULL, &mtime, 16); 969219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs), 970219089Spjd NULL, &ctime, 16); 971219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs), 972219089Spjd NULL, &crtime, 16); 973219089Spjd } 974219089Spjd 975219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8); 976219089Spjd 977219089Spjd if (obj_type == DMU_OT_ZNODE) { 978219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zfsvfs), NULL, 979219089Spjd &empty_xattr, 8); 980219089Spjd } 981219089Spjd if (obj_type == DMU_OT_ZNODE || 982219089Spjd (vap->va_type == VBLK || vap->va_type == VCHR)) { 983219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zfsvfs), 984219089Spjd NULL, &rdev, 8); 985219089Spjd 986219089Spjd } 987219089Spjd if (obj_type == DMU_OT_ZNODE) { 988219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs), 989219089Spjd NULL, &pflags, 8); 990219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL, 991219089Spjd &acl_ids->z_fuid, 8); 992219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL, 993219089Spjd &acl_ids->z_fgid, 8); 994219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zfsvfs), NULL, pad, 995219089Spjd sizeof (uint64_t) * 4); 996219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zfsvfs), NULL, 997219089Spjd &acl_phys, sizeof (zfs_acl_phys_t)); 998219089Spjd } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) { 999219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zfsvfs), NULL, 1000219089Spjd &acl_ids->z_aclp->z_acl_count, 8); 1001219089Spjd locate.cb_aclp = acl_ids->z_aclp; 1002219089Spjd SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zfsvfs), 1003219089Spjd zfs_acl_data_locator, &locate, 1004219089Spjd acl_ids->z_aclp->z_acl_bytes); 1005219089Spjd mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags, 1006219089Spjd acl_ids->z_fuid, acl_ids->z_fgid); 1007219089Spjd } 1008219089Spjd 1009219089Spjd VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0); 1010219089Spjd 1011185029Spjd if (!(flag & IS_ROOT_NODE)) { 1012219089Spjd *zpp = zfs_znode_alloc(zfsvfs, db, 0, obj_type, sa_hdl); 1013219089Spjd ASSERT(*zpp != NULL); 1014185029Spjd } else { 1015185029Spjd /* 1016185029Spjd * If we are creating the root node, the "parent" we 1017185029Spjd * passed in is the znode for the root. 1018185029Spjd */ 1019185029Spjd *zpp = dzp; 1020219089Spjd 1021219089Spjd (*zpp)->z_sa_hdl = sa_hdl; 1022185029Spjd } 1023219089Spjd 1024219089Spjd (*zpp)->z_pflags = pflags; 1025219089Spjd (*zpp)->z_mode = mode; 1026219089Spjd 1027209962Smm if (vap->va_mask & AT_XVATTR) 1028219089Spjd zfs_xvattr_set(*zpp, (xvattr_t *)vap, tx); 1029219089Spjd 1030219089Spjd if (obj_type == DMU_OT_ZNODE || 1031219089Spjd acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) { 1032219089Spjd err = zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx); 1033240415Smm ASSERT0(err); 1034219089Spjd } 1035185029Spjd if (!(flag & IS_ROOT_NODE)) { 1036185029Spjd vnode_t *vp; 1037168404Spjd 1038185029Spjd vp = ZTOV(*zpp); 1039185029Spjd vp->v_vflag |= VV_FORCEINSMQ; 1040185029Spjd err = insmntque(vp, zfsvfs->z_vfs); 1041185029Spjd vp->v_vflag &= ~VV_FORCEINSMQ; 1042185029Spjd KASSERT(err == 0, ("insmntque() failed: error %d", err)); 1043185029Spjd } 1044219089Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, obj); 1045241628Savg getnewvnode_drop_reserve(); 1046185029Spjd} 1047168404Spjd 1048219089Spjd/* 1049219089Spjd * zfs_xvattr_set only updates the in-core attributes 1050219089Spjd * it is assumed the caller will be doing an sa_bulk_update 1051219089Spjd * to push the changes out 1052219089Spjd */ 1053185029Spjdvoid 1054219089Spjdzfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx) 1055185029Spjd{ 1056185029Spjd xoptattr_t *xoap; 1057168404Spjd 1058185029Spjd xoap = xva_getxoptattr(xvap); 1059185029Spjd ASSERT(xoap); 1060168404Spjd 1061185029Spjd if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) { 1062219089Spjd uint64_t times[2]; 1063219089Spjd ZFS_TIME_ENCODE(&xoap->xoa_createtime, times); 1064219089Spjd (void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(zp->z_zfsvfs), 1065219089Spjd ×, sizeof (times), tx); 1066185029Spjd XVA_SET_RTN(xvap, XAT_CREATETIME); 1067168404Spjd } 1068185029Spjd if (XVA_ISSET_REQ(xvap, XAT_READONLY)) { 1069219089Spjd ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly, 1070219089Spjd zp->z_pflags, tx); 1071185029Spjd XVA_SET_RTN(xvap, XAT_READONLY); 1072185029Spjd } 1073185029Spjd if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) { 1074219089Spjd ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden, 1075219089Spjd zp->z_pflags, tx); 1076185029Spjd XVA_SET_RTN(xvap, XAT_HIDDEN); 1077185029Spjd } 1078185029Spjd if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) { 1079219089Spjd ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system, 1080219089Spjd zp->z_pflags, tx); 1081185029Spjd XVA_SET_RTN(xvap, XAT_SYSTEM); 1082185029Spjd } 1083185029Spjd if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) { 1084219089Spjd ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive, 1085219089Spjd zp->z_pflags, tx); 1086185029Spjd XVA_SET_RTN(xvap, XAT_ARCHIVE); 1087185029Spjd } 1088185029Spjd if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { 1089219089Spjd ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable, 1090219089Spjd zp->z_pflags, tx); 1091185029Spjd XVA_SET_RTN(xvap, XAT_IMMUTABLE); 1092185029Spjd } 1093185029Spjd if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { 1094219089Spjd ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink, 1095219089Spjd zp->z_pflags, tx); 1096185029Spjd XVA_SET_RTN(xvap, XAT_NOUNLINK); 1097185029Spjd } 1098185029Spjd if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { 1099219089Spjd ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly, 1100219089Spjd zp->z_pflags, tx); 1101185029Spjd XVA_SET_RTN(xvap, XAT_APPENDONLY); 1102185029Spjd } 1103185029Spjd if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { 1104219089Spjd ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump, 1105219089Spjd zp->z_pflags, tx); 1106185029Spjd XVA_SET_RTN(xvap, XAT_NODUMP); 1107185029Spjd } 1108185029Spjd if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) { 1109219089Spjd ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque, 1110219089Spjd zp->z_pflags, tx); 1111185029Spjd XVA_SET_RTN(xvap, XAT_OPAQUE); 1112185029Spjd } 1113185029Spjd if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { 1114185029Spjd ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED, 1115219089Spjd xoap->xoa_av_quarantined, zp->z_pflags, tx); 1116185029Spjd XVA_SET_RTN(xvap, XAT_AV_QUARANTINED); 1117185029Spjd } 1118185029Spjd if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { 1119219089Spjd ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified, 1120219089Spjd zp->z_pflags, tx); 1121185029Spjd XVA_SET_RTN(xvap, XAT_AV_MODIFIED); 1122185029Spjd } 1123185029Spjd if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) { 1124219089Spjd zfs_sa_set_scanstamp(zp, xvap, tx); 1125185029Spjd XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP); 1126185029Spjd } 1127219089Spjd if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { 1128219089Spjd ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse, 1129219089Spjd zp->z_pflags, tx); 1130219089Spjd XVA_SET_RTN(xvap, XAT_REPARSE); 1131219089Spjd } 1132219089Spjd if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) { 1133219089Spjd ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline, 1134219089Spjd zp->z_pflags, tx); 1135219089Spjd XVA_SET_RTN(xvap, XAT_OFFLINE); 1136219089Spjd } 1137219089Spjd if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) { 1138219089Spjd ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse, 1139219089Spjd zp->z_pflags, tx); 1140219089Spjd XVA_SET_RTN(xvap, XAT_SPARSE); 1141219089Spjd } 1142168404Spjd} 1143168404Spjd 1144168404Spjdint 1145168404Spjdzfs_zget(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp) 1146168404Spjd{ 1147168404Spjd dmu_object_info_t doi; 1148168404Spjd dmu_buf_t *db; 1149168404Spjd znode_t *zp; 1150243520Savg vnode_t *vp; 1151243520Savg sa_handle_t *hdl; 1152243520Savg struct thread *td; 1153243520Savg int locked; 1154219089Spjd int err; 1155168404Spjd 1156243520Savg td = curthread; 1157241628Savg getnewvnode_reserve(1); 1158185029Spjdagain: 1159243520Savg *zpp = NULL; 1160168404Spjd ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num); 1161168404Spjd 1162219089Spjd err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db); 1163168404Spjd if (err) { 1164168404Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1165241628Savg getnewvnode_drop_reserve(); 1166168404Spjd return (err); 1167168404Spjd } 1168168404Spjd 1169168404Spjd dmu_object_info_from_db(db, &doi); 1170219089Spjd if (doi.doi_bonus_type != DMU_OT_SA && 1171219089Spjd (doi.doi_bonus_type != DMU_OT_ZNODE || 1172219089Spjd (doi.doi_bonus_type == DMU_OT_ZNODE && 1173219089Spjd doi.doi_bonus_size < sizeof (znode_phys_t)))) { 1174219089Spjd sa_buf_rele(db, NULL); 1175168404Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1176241628Savg getnewvnode_drop_reserve(); 1177168404Spjd return (EINVAL); 1178168404Spjd } 1179168404Spjd 1180219089Spjd hdl = dmu_buf_get_user(db); 1181219089Spjd if (hdl != NULL) { 1182219089Spjd zp = sa_get_userdata(hdl); 1183168404Spjd 1184219089Spjd 1185185029Spjd /* 1186219089Spjd * Since "SA" does immediate eviction we 1187219089Spjd * should never find a sa handle that doesn't 1188219089Spjd * know about the znode. 1189185029Spjd */ 1190219089Spjd 1191219089Spjd ASSERT3P(zp, !=, NULL); 1192219089Spjd 1193219089Spjd mutex_enter(&zp->z_lock); 1194168404Spjd ASSERT3U(zp->z_id, ==, obj_num); 1195168404Spjd if (zp->z_unlinked) { 1196185029Spjd err = ENOENT; 1197168404Spjd } else { 1198197458Spjd vp = ZTOV(zp); 1199185029Spjd *zpp = zp; 1200185029Spjd err = 0; 1201168404Spjd } 1202219089Spjd sa_buf_rele(db, NULL); 1203243520Savg 1204243520Savg /* Don't let the vnode disappear after ZFS_OBJ_HOLD_EXIT. */ 1205243520Savg if (err == 0) 1206243520Savg VN_HOLD(vp); 1207243520Savg 1208168404Spjd mutex_exit(&zp->z_lock); 1209168404Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1210243520Savg 1211243520Savg if (err == 0) { 1212243520Savg locked = VOP_ISLOCKED(vp); 1213243520Savg VI_LOCK(vp); 1214243520Savg if ((vp->v_iflag & VI_DOOMED) != 0 && 1215243520Savg locked != LK_EXCLUSIVE) { 1216243520Savg /* 1217243520Savg * The vnode is doomed and this thread doesn't 1218243520Savg * hold the exclusive lock on it, so the vnode 1219243520Savg * must be being reclaimed by another thread. 1220243520Savg * Otherwise the doomed vnode is being reclaimed 1221243520Savg * by this thread and zfs_zget is called from 1222243520Savg * ZIL internals. 1223243520Savg */ 1224243520Savg VI_UNLOCK(vp); 1225243520Savg VN_RELE(vp); 1226243520Savg goto again; 1227243520Savg } 1228243520Savg VI_UNLOCK(vp); 1229243520Savg } 1230241628Savg getnewvnode_drop_reserve(); 1231185029Spjd return (err); 1232168404Spjd } 1233168404Spjd 1234168404Spjd /* 1235168404Spjd * Not found create new znode/vnode 1236207334Spjd * but only if file exists. 1237207334Spjd * 1238207334Spjd * There is a small window where zfs_vget() could 1239207334Spjd * find this object while a file create is still in 1240219089Spjd * progress. This is checked for in zfs_znode_alloc() 1241219089Spjd * 1242219089Spjd * if zfs_znode_alloc() fails it will drop the hold on the 1243219089Spjd * bonus buffer. 1244168404Spjd */ 1245219089Spjd zp = zfs_znode_alloc(zfsvfs, db, doi.doi_data_block_size, 1246219089Spjd doi.doi_bonus_type, NULL); 1247219089Spjd if (zp == NULL) { 1248219089Spjd err = ENOENT; 1249219089Spjd } else { 1250219089Spjd *zpp = zp; 1251219089Spjd } 1252219089Spjd if (err == 0) { 1253219089Spjd vnode_t *vp = ZTOV(zp); 1254185029Spjd 1255207334Spjd err = insmntque(vp, zfsvfs->z_vfs); 1256219089Spjd if (err == 0) 1257219089Spjd VOP_UNLOCK(vp, 0); 1258219089Spjd else { 1259219089Spjd zp->z_vnode = NULL; 1260219089Spjd zfs_znode_dmu_fini(zp); 1261219089Spjd zfs_znode_free(zp); 1262219089Spjd *zpp = NULL; 1263219089Spjd } 1264207334Spjd } 1265168404Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1266241628Savg getnewvnode_drop_reserve(); 1267207334Spjd return (err); 1268168404Spjd} 1269168404Spjd 1270185029Spjdint 1271185029Spjdzfs_rezget(znode_t *zp) 1272185029Spjd{ 1273185029Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1274185029Spjd dmu_object_info_t doi; 1275185029Spjd dmu_buf_t *db; 1276225166Smm vnode_t *vp; 1277185029Spjd uint64_t obj_num = zp->z_id; 1278219089Spjd uint64_t mode, size; 1279219089Spjd sa_bulk_attr_t bulk[8]; 1280185029Spjd int err; 1281219089Spjd int count = 0; 1282219089Spjd uint64_t gen; 1283185029Spjd 1284185029Spjd ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num); 1285185029Spjd 1286219089Spjd mutex_enter(&zp->z_acl_lock); 1287219089Spjd if (zp->z_acl_cached) { 1288219089Spjd zfs_acl_free(zp->z_acl_cached); 1289219089Spjd zp->z_acl_cached = NULL; 1290219089Spjd } 1291219089Spjd 1292219089Spjd mutex_exit(&zp->z_acl_lock); 1293219089Spjd ASSERT(zp->z_sa_hdl == NULL); 1294219089Spjd err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db); 1295185029Spjd if (err) { 1296185029Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1297185029Spjd return (err); 1298185029Spjd } 1299185029Spjd 1300185029Spjd dmu_object_info_from_db(db, &doi); 1301219089Spjd if (doi.doi_bonus_type != DMU_OT_SA && 1302219089Spjd (doi.doi_bonus_type != DMU_OT_ZNODE || 1303219089Spjd (doi.doi_bonus_type == DMU_OT_ZNODE && 1304219089Spjd doi.doi_bonus_size < sizeof (znode_phys_t)))) { 1305219089Spjd sa_buf_rele(db, NULL); 1306185029Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1307185029Spjd return (EINVAL); 1308185029Spjd } 1309185029Spjd 1310219089Spjd zfs_znode_sa_init(zfsvfs, zp, db, doi.doi_bonus_type, NULL); 1311219089Spjd size = zp->z_size; 1312219089Spjd 1313219089Spjd /* reload cached values */ 1314219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, 1315219089Spjd &gen, sizeof (gen)); 1316219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, 1317219089Spjd &zp->z_size, sizeof (zp->z_size)); 1318219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL, 1319219089Spjd &zp->z_links, sizeof (zp->z_links)); 1320219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, 1321219089Spjd &zp->z_pflags, sizeof (zp->z_pflags)); 1322219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, 1323219089Spjd &zp->z_atime, sizeof (zp->z_atime)); 1324219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, 1325219089Spjd &zp->z_uid, sizeof (zp->z_uid)); 1326219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, 1327219089Spjd &zp->z_gid, sizeof (zp->z_gid)); 1328219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, 1329219089Spjd &mode, sizeof (mode)); 1330219089Spjd 1331219089Spjd if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) { 1332219089Spjd zfs_znode_dmu_fini(zp); 1333185029Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1334185029Spjd return (EIO); 1335185029Spjd } 1336185029Spjd 1337219089Spjd zp->z_mode = mode; 1338219089Spjd 1339219089Spjd if (gen != zp->z_gen) { 1340219089Spjd zfs_znode_dmu_fini(zp); 1341219089Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1342219089Spjd return (EIO); 1343211932Smm } 1344211932Smm 1345219089Spjd /* 1346219089Spjd * XXXPJD: Not sure how is that possible, but under heavy 1347219089Spjd * zfs recv -F load it happens that z_gen is the same, but 1348219089Spjd * vnode type is different than znode type. This would mean 1349219089Spjd * that for example regular file was replaced with directory 1350219089Spjd * which has the same object number. 1351219089Spjd */ 1352225166Smm vp = ZTOV(zp); 1353225166Smm if (vp != NULL && 1354225166Smm vp->v_type != IFTOVT((mode_t)zp->z_mode)) { 1355219089Spjd zfs_znode_dmu_fini(zp); 1356219089Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1357219089Spjd return (EIO); 1358219089Spjd } 1359219089Spjd 1360219089Spjd zp->z_unlinked = (zp->z_links == 0); 1361185029Spjd zp->z_blksz = doi.doi_data_block_size; 1362225166Smm if (vp != NULL) { 1363225166Smm vn_pages_remove(vp, 0, 0); 1364225166Smm if (zp->z_size != size) 1365225166Smm vnode_pager_setsize(vp, zp->z_size); 1366225166Smm } 1367185029Spjd 1368185029Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1369185029Spjd 1370185029Spjd return (0); 1371185029Spjd} 1372185029Spjd 1373168404Spjdvoid 1374168404Spjdzfs_znode_delete(znode_t *zp, dmu_tx_t *tx) 1375168404Spjd{ 1376168404Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1377185029Spjd objset_t *os = zfsvfs->z_os; 1378185029Spjd uint64_t obj = zp->z_id; 1379219089Spjd uint64_t acl_obj = zfs_external_acl(zp); 1380168404Spjd 1381185029Spjd ZFS_OBJ_HOLD_ENTER(zfsvfs, obj); 1382219089Spjd if (acl_obj) { 1383219089Spjd VERIFY(!zp->z_is_sa); 1384185029Spjd VERIFY(0 == dmu_object_free(os, acl_obj, tx)); 1385219089Spjd } 1386185029Spjd VERIFY(0 == dmu_object_free(os, obj, tx)); 1387185029Spjd zfs_znode_dmu_fini(zp); 1388185029Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, obj); 1389185029Spjd zfs_znode_free(zp); 1390168404Spjd} 1391168404Spjd 1392168404Spjdvoid 1393168404Spjdzfs_zinactive(znode_t *zp) 1394168404Spjd{ 1395168404Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1396168404Spjd uint64_t z_id = zp->z_id; 1397168404Spjd 1398219089Spjd ASSERT(zp->z_sa_hdl); 1399168404Spjd 1400168404Spjd /* 1401168404Spjd * Don't allow a zfs_zget() while were trying to release this znode 1402168404Spjd */ 1403168404Spjd ZFS_OBJ_HOLD_ENTER(zfsvfs, z_id); 1404168404Spjd 1405168404Spjd mutex_enter(&zp->z_lock); 1406168404Spjd 1407168404Spjd /* 1408168404Spjd * If this was the last reference to a file with no links, 1409168404Spjd * remove the file from the file system. 1410168404Spjd */ 1411168404Spjd if (zp->z_unlinked) { 1412168404Spjd mutex_exit(&zp->z_lock); 1413168404Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id); 1414168404Spjd zfs_rmnode(zp); 1415168404Spjd return; 1416168404Spjd } 1417219089Spjd 1418168404Spjd mutex_exit(&zp->z_lock); 1419243520Savg zfs_znode_dmu_fini(zp); 1420168404Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id); 1421243520Savg zfs_znode_free(zp); 1422168404Spjd} 1423168404Spjd 1424168404Spjdvoid 1425168404Spjdzfs_znode_free(znode_t *zp) 1426168404Spjd{ 1427168404Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1428168404Spjd 1429219089Spjd ASSERT(zp->z_sa_hdl == NULL); 1430243520Savg zp->z_vnode = NULL; 1431168404Spjd mutex_enter(&zfsvfs->z_znodes_lock); 1432185029Spjd POINTER_INVALIDATE(&zp->z_zfsvfs); 1433168404Spjd list_remove(&zfsvfs->z_all_znodes, zp); 1434168404Spjd mutex_exit(&zfsvfs->z_znodes_lock); 1435168404Spjd 1436211932Smm if (zp->z_acl_cached) { 1437211932Smm zfs_acl_free(zp->z_acl_cached); 1438211932Smm zp->z_acl_cached = NULL; 1439211932Smm } 1440211932Smm 1441168404Spjd kmem_cache_free(znode_cache, zp); 1442185029Spjd 1443185029Spjd VFS_RELE(zfsvfs->z_vfs); 1444168404Spjd} 1445168404Spjd 1446168404Spjdvoid 1447219089Spjdzfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2], 1448219089Spjd uint64_t ctime[2], boolean_t have_tx) 1449168404Spjd{ 1450168404Spjd timestruc_t now; 1451168404Spjd 1452168404Spjd gethrestime(&now); 1453168404Spjd 1454219089Spjd if (have_tx) { /* will sa_bulk_update happen really soon? */ 1455168404Spjd zp->z_atime_dirty = 0; 1456168404Spjd zp->z_seq++; 1457168404Spjd } else { 1458168404Spjd zp->z_atime_dirty = 1; 1459168404Spjd } 1460168404Spjd 1461219089Spjd if (flag & AT_ATIME) { 1462219089Spjd ZFS_TIME_ENCODE(&now, zp->z_atime); 1463219089Spjd } 1464168404Spjd 1465185029Spjd if (flag & AT_MTIME) { 1466219089Spjd ZFS_TIME_ENCODE(&now, mtime); 1467219089Spjd if (zp->z_zfsvfs->z_use_fuids) { 1468219089Spjd zp->z_pflags |= (ZFS_ARCHIVE | 1469219089Spjd ZFS_AV_MODIFIED); 1470219089Spjd } 1471185029Spjd } 1472168404Spjd 1473185029Spjd if (flag & AT_CTIME) { 1474219089Spjd ZFS_TIME_ENCODE(&now, ctime); 1475185029Spjd if (zp->z_zfsvfs->z_use_fuids) 1476219089Spjd zp->z_pflags |= ZFS_ARCHIVE; 1477185029Spjd } 1478168404Spjd} 1479168404Spjd 1480168404Spjd/* 1481168404Spjd * Grow the block size for a file. 1482168404Spjd * 1483168404Spjd * IN: zp - znode of file to free data in. 1484168404Spjd * size - requested block size 1485168404Spjd * tx - open transaction. 1486168404Spjd * 1487168404Spjd * NOTE: this function assumes that the znode is write locked. 1488168404Spjd */ 1489168404Spjdvoid 1490168404Spjdzfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx) 1491168404Spjd{ 1492168404Spjd int error; 1493168404Spjd u_longlong_t dummy; 1494168404Spjd 1495168404Spjd if (size <= zp->z_blksz) 1496168404Spjd return; 1497168404Spjd /* 1498168404Spjd * If the file size is already greater than the current blocksize, 1499168404Spjd * we will not grow. If there is more than one block in a file, 1500168404Spjd * the blocksize cannot change. 1501168404Spjd */ 1502219089Spjd if (zp->z_blksz && zp->z_size > zp->z_blksz) 1503168404Spjd return; 1504168404Spjd 1505168404Spjd error = dmu_object_set_blocksize(zp->z_zfsvfs->z_os, zp->z_id, 1506168404Spjd size, 0, tx); 1507219089Spjd 1508168404Spjd if (error == ENOTSUP) 1509168404Spjd return; 1510240415Smm ASSERT0(error); 1511168404Spjd 1512168404Spjd /* What blocksize did we actually get? */ 1513219089Spjd dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy); 1514168404Spjd} 1515168404Spjd 1516219089Spjd#ifdef sun 1517168404Spjd/* 1518219089Spjd * This is a dummy interface used when pvn_vplist_dirty() should *not* 1519219089Spjd * be calling back into the fs for a putpage(). E.g.: when truncating 1520219089Spjd * a file, the pages being "thrown away* don't need to be written out. 1521219089Spjd */ 1522219089Spjd/* ARGSUSED */ 1523219089Spjdstatic int 1524219089Spjdzfs_no_putpage(vnode_t *vp, page_t *pp, u_offset_t *offp, size_t *lenp, 1525219089Spjd int flags, cred_t *cr) 1526219089Spjd{ 1527219089Spjd ASSERT(0); 1528219089Spjd return (0); 1529219089Spjd} 1530219089Spjd#endif /* sun */ 1531219089Spjd 1532219089Spjd/* 1533185029Spjd * Increase the file length 1534168404Spjd * 1535168404Spjd * IN: zp - znode of file to free data in. 1536185029Spjd * end - new end-of-file 1537168404Spjd * 1538168404Spjd * RETURN: 0 if success 1539168404Spjd * error code if failure 1540168404Spjd */ 1541185029Spjdstatic int 1542185029Spjdzfs_extend(znode_t *zp, uint64_t end) 1543168404Spjd{ 1544185029Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1545168404Spjd dmu_tx_t *tx; 1546168404Spjd rl_t *rl; 1547185029Spjd uint64_t newblksz; 1548168404Spjd int error; 1549168404Spjd 1550168404Spjd /* 1551185029Spjd * We will change zp_size, lock the whole file. 1552168404Spjd */ 1553185029Spjd rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER); 1554168404Spjd 1555168404Spjd /* 1556168404Spjd * Nothing to do if file already at desired length. 1557168404Spjd */ 1558219089Spjd if (end <= zp->z_size) { 1559168404Spjd zfs_range_unlock(rl); 1560168404Spjd return (0); 1561168404Spjd } 1562185029Spjdtop: 1563168404Spjd tx = dmu_tx_create(zfsvfs->z_os); 1564219089Spjd dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); 1565219089Spjd zfs_sa_upgrade_txholds(tx, zp); 1566185029Spjd if (end > zp->z_blksz && 1567168404Spjd (!ISP2(zp->z_blksz) || zp->z_blksz < zfsvfs->z_max_blksz)) { 1568168404Spjd /* 1569168404Spjd * We are growing the file past the current block size. 1570168404Spjd */ 1571168404Spjd if (zp->z_blksz > zp->z_zfsvfs->z_max_blksz) { 1572168404Spjd ASSERT(!ISP2(zp->z_blksz)); 1573185029Spjd newblksz = MIN(end, SPA_MAXBLOCKSIZE); 1574168404Spjd } else { 1575185029Spjd newblksz = MIN(end, zp->z_zfsvfs->z_max_blksz); 1576168404Spjd } 1577185029Spjd dmu_tx_hold_write(tx, zp->z_id, 0, newblksz); 1578185029Spjd } else { 1579185029Spjd newblksz = 0; 1580168404Spjd } 1581168404Spjd 1582209962Smm error = dmu_tx_assign(tx, TXG_NOWAIT); 1583168404Spjd if (error) { 1584209962Smm if (error == ERESTART) { 1585168404Spjd dmu_tx_wait(tx); 1586185029Spjd dmu_tx_abort(tx); 1587185029Spjd goto top; 1588185029Spjd } 1589168404Spjd dmu_tx_abort(tx); 1590168404Spjd zfs_range_unlock(rl); 1591168404Spjd return (error); 1592168404Spjd } 1593168404Spjd 1594185029Spjd if (newblksz) 1595185029Spjd zfs_grow_blocksize(zp, newblksz, tx); 1596168404Spjd 1597219089Spjd zp->z_size = end; 1598168404Spjd 1599219089Spjd VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zp->z_zfsvfs), 1600219089Spjd &zp->z_size, sizeof (zp->z_size), tx)); 1601219089Spjd 1602219089Spjd vnode_pager_setsize(ZTOV(zp), end); 1603219089Spjd 1604185029Spjd zfs_range_unlock(rl); 1605168404Spjd 1606185029Spjd dmu_tx_commit(tx); 1607185029Spjd 1608185029Spjd return (0); 1609185029Spjd} 1610185029Spjd 1611185029Spjd/* 1612185029Spjd * Free space in a file. 1613185029Spjd * 1614185029Spjd * IN: zp - znode of file to free data in. 1615185029Spjd * off - start of section to free. 1616185029Spjd * len - length of section to free. 1617185029Spjd * 1618185029Spjd * RETURN: 0 if success 1619185029Spjd * error code if failure 1620185029Spjd */ 1621185029Spjdstatic int 1622185029Spjdzfs_free_range(znode_t *zp, uint64_t off, uint64_t len) 1623185029Spjd{ 1624185029Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1625185029Spjd rl_t *rl; 1626185029Spjd int error; 1627185029Spjd 1628185029Spjd /* 1629185029Spjd * Lock the range being freed. 1630185029Spjd */ 1631185029Spjd rl = zfs_range_lock(zp, off, len, RL_WRITER); 1632185029Spjd 1633185029Spjd /* 1634185029Spjd * Nothing to do if file already at desired length. 1635185029Spjd */ 1636219089Spjd if (off >= zp->z_size) { 1637185029Spjd zfs_range_unlock(rl); 1638185029Spjd return (0); 1639168404Spjd } 1640168404Spjd 1641219089Spjd if (off + len > zp->z_size) 1642219089Spjd len = zp->z_size - off; 1643185029Spjd 1644185029Spjd error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, off, len); 1645185029Spjd 1646185029Spjd if (error == 0) { 1647185029Spjd /* 1648185029Spjd * In FreeBSD we cannot free block in the middle of a file, 1649219089Spjd * but only at the end of a file, so this code path should 1650219089Spjd * never happen. 1651185029Spjd */ 1652185029Spjd vnode_pager_setsize(ZTOV(zp), off); 1653168404Spjd } 1654168404Spjd 1655168404Spjd zfs_range_unlock(rl); 1656168404Spjd 1657185029Spjd return (error); 1658185029Spjd} 1659185029Spjd 1660185029Spjd/* 1661185029Spjd * Truncate a file 1662185029Spjd * 1663185029Spjd * IN: zp - znode of file to free data in. 1664185029Spjd * end - new end-of-file. 1665185029Spjd * 1666185029Spjd * RETURN: 0 if success 1667185029Spjd * error code if failure 1668185029Spjd */ 1669185029Spjdstatic int 1670185029Spjdzfs_trunc(znode_t *zp, uint64_t end) 1671185029Spjd{ 1672185029Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1673185029Spjd vnode_t *vp = ZTOV(zp); 1674185029Spjd dmu_tx_t *tx; 1675185029Spjd rl_t *rl; 1676185029Spjd int error; 1677219089Spjd sa_bulk_attr_t bulk[2]; 1678219089Spjd int count = 0; 1679185029Spjd 1680185029Spjd /* 1681185029Spjd * We will change zp_size, lock the whole file. 1682185029Spjd */ 1683185029Spjd rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER); 1684185029Spjd 1685185029Spjd /* 1686185029Spjd * Nothing to do if file already at desired length. 1687185029Spjd */ 1688219089Spjd if (end >= zp->z_size) { 1689185029Spjd zfs_range_unlock(rl); 1690185029Spjd return (0); 1691185029Spjd } 1692185029Spjd 1693185029Spjd error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, end, -1); 1694185029Spjd if (error) { 1695185029Spjd zfs_range_unlock(rl); 1696185029Spjd return (error); 1697185029Spjd } 1698185029Spjdtop: 1699185029Spjd tx = dmu_tx_create(zfsvfs->z_os); 1700219089Spjd dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); 1701219089Spjd zfs_sa_upgrade_txholds(tx, zp); 1702209962Smm error = dmu_tx_assign(tx, TXG_NOWAIT); 1703185029Spjd if (error) { 1704209962Smm if (error == ERESTART) { 1705185029Spjd dmu_tx_wait(tx); 1706185029Spjd dmu_tx_abort(tx); 1707185029Spjd goto top; 1708185029Spjd } 1709185029Spjd dmu_tx_abort(tx); 1710185029Spjd zfs_range_unlock(rl); 1711185029Spjd return (error); 1712185029Spjd } 1713185029Spjd 1714219089Spjd zp->z_size = end; 1715219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), 1716219089Spjd NULL, &zp->z_size, sizeof (zp->z_size)); 1717185029Spjd 1718219089Spjd if (end == 0) { 1719219089Spjd zp->z_pflags &= ~ZFS_SPARSE; 1720219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), 1721219089Spjd NULL, &zp->z_pflags, 8); 1722219089Spjd } 1723219089Spjd VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0); 1724219089Spjd 1725168404Spjd dmu_tx_commit(tx); 1726168404Spjd 1727168404Spjd /* 1728168404Spjd * Clear any mapped pages in the truncated region. This has to 1729168404Spjd * happen outside of the transaction to avoid the possibility of 1730168404Spjd * a deadlock with someone trying to push a page that we are 1731168404Spjd * about to invalidate. 1732168404Spjd */ 1733185029Spjd vnode_pager_setsize(vp, end); 1734168404Spjd 1735209962Smm zfs_range_unlock(rl); 1736209962Smm 1737168404Spjd return (0); 1738168404Spjd} 1739168404Spjd 1740185029Spjd/* 1741185029Spjd * Free space in a file 1742185029Spjd * 1743185029Spjd * IN: zp - znode of file to free data in. 1744185029Spjd * off - start of range 1745185029Spjd * len - end of range (0 => EOF) 1746185029Spjd * flag - current file open mode flags. 1747185029Spjd * log - TRUE if this action should be logged 1748185029Spjd * 1749185029Spjd * RETURN: 0 if success 1750185029Spjd * error code if failure 1751185029Spjd */ 1752185029Spjdint 1753185029Spjdzfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log) 1754185029Spjd{ 1755185029Spjd vnode_t *vp = ZTOV(zp); 1756185029Spjd dmu_tx_t *tx; 1757185029Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1758185029Spjd zilog_t *zilog = zfsvfs->z_log; 1759219089Spjd uint64_t mode; 1760219089Spjd uint64_t mtime[2], ctime[2]; 1761219089Spjd sa_bulk_attr_t bulk[3]; 1762219089Spjd int count = 0; 1763185029Spjd int error; 1764185029Spjd 1765219089Spjd if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), &mode, 1766219089Spjd sizeof (mode))) != 0) 1767219089Spjd return (error); 1768219089Spjd 1769219089Spjd if (off > zp->z_size) { 1770185029Spjd error = zfs_extend(zp, off+len); 1771185029Spjd if (error == 0 && log) 1772185029Spjd goto log; 1773185029Spjd else 1774185029Spjd return (error); 1775185029Spjd } 1776185029Spjd 1777219089Spjd /* 1778219089Spjd * Check for any locks in the region to be freed. 1779219089Spjd */ 1780219089Spjd 1781219089Spjd if (MANDLOCK(vp, (mode_t)mode)) { 1782219089Spjd uint64_t length = (len ? len : zp->z_size - off); 1783219089Spjd if (error = chklock(vp, FWRITE, off, length, flag, NULL)) 1784219089Spjd return (error); 1785219089Spjd } 1786219089Spjd 1787185029Spjd if (len == 0) { 1788185029Spjd error = zfs_trunc(zp, off); 1789185029Spjd } else { 1790185029Spjd if ((error = zfs_free_range(zp, off, len)) == 0 && 1791219089Spjd off + len > zp->z_size) 1792185029Spjd error = zfs_extend(zp, off+len); 1793185029Spjd } 1794185029Spjd if (error || !log) 1795185029Spjd return (error); 1796185029Spjdlog: 1797185029Spjd tx = dmu_tx_create(zfsvfs->z_os); 1798219089Spjd dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); 1799219089Spjd zfs_sa_upgrade_txholds(tx, zp); 1800209962Smm error = dmu_tx_assign(tx, TXG_NOWAIT); 1801185029Spjd if (error) { 1802209962Smm if (error == ERESTART) { 1803185029Spjd dmu_tx_wait(tx); 1804185029Spjd dmu_tx_abort(tx); 1805185029Spjd goto log; 1806185029Spjd } 1807185029Spjd dmu_tx_abort(tx); 1808185029Spjd return (error); 1809185029Spjd } 1810185029Spjd 1811219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, mtime, 16); 1812219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, ctime, 16); 1813219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), 1814219089Spjd NULL, &zp->z_pflags, 8); 1815219089Spjd zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, B_TRUE); 1816219089Spjd error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); 1817219089Spjd ASSERT(error == 0); 1818219089Spjd 1819185029Spjd zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len); 1820185029Spjd 1821185029Spjd dmu_tx_commit(tx); 1822185029Spjd return (0); 1823185029Spjd} 1824185029Spjd 1825168404Spjdvoid 1826185029Spjdzfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx) 1827168404Spjd{ 1828168404Spjd zfsvfs_t zfsvfs; 1829219089Spjd uint64_t moid, obj, sa_obj, version; 1830185029Spjd uint64_t sense = ZFS_CASE_SENSITIVE; 1831185029Spjd uint64_t norm = 0; 1832185029Spjd nvpair_t *elem; 1833168404Spjd int error; 1834207334Spjd int i; 1835168404Spjd znode_t *rootzp = NULL; 1836199156Spjd vnode_t vnode; 1837168404Spjd vattr_t vattr; 1838185029Spjd znode_t *zp; 1839209962Smm zfs_acl_ids_t acl_ids; 1840168404Spjd 1841168404Spjd /* 1842168404Spjd * First attempt to create master node. 1843168404Spjd */ 1844168404Spjd /* 1845168404Spjd * In an empty objset, there are no blocks to read and thus 1846168404Spjd * there can be no i/o errors (which we assert below). 1847168404Spjd */ 1848168404Spjd moid = MASTER_NODE_OBJ; 1849168404Spjd error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE, 1850168404Spjd DMU_OT_NONE, 0, tx); 1851168404Spjd ASSERT(error == 0); 1852168404Spjd 1853168404Spjd /* 1854168404Spjd * Set starting attributes. 1855168404Spjd */ 1856219089Spjd version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os))); 1857185029Spjd elem = NULL; 1858185029Spjd while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) { 1859185029Spjd /* For the moment we expect all zpl props to be uint64_ts */ 1860185029Spjd uint64_t val; 1861185029Spjd char *name; 1862168404Spjd 1863185029Spjd ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64); 1864185029Spjd VERIFY(nvpair_value_uint64(elem, &val) == 0); 1865185029Spjd name = nvpair_name(elem); 1866185029Spjd if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) { 1867209962Smm if (val < version) 1868209962Smm version = val; 1869185029Spjd } else { 1870185029Spjd error = zap_update(os, moid, name, 8, 1, &val, tx); 1871185029Spjd } 1872185029Spjd ASSERT(error == 0); 1873185029Spjd if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0) 1874185029Spjd norm = val; 1875185029Spjd else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0) 1876185029Spjd sense = val; 1877185029Spjd } 1878185029Spjd ASSERT(version != 0); 1879209962Smm error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx); 1880168404Spjd 1881168404Spjd /* 1882219089Spjd * Create zap object used for SA attribute registration 1883219089Spjd */ 1884219089Spjd 1885219089Spjd if (version >= ZPL_VERSION_SA) { 1886219089Spjd sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE, 1887219089Spjd DMU_OT_NONE, 0, tx); 1888219089Spjd error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx); 1889219089Spjd ASSERT(error == 0); 1890219089Spjd } else { 1891219089Spjd sa_obj = 0; 1892219089Spjd } 1893219089Spjd /* 1894168404Spjd * Create a delete queue. 1895168404Spjd */ 1896209962Smm obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx); 1897168404Spjd 1898209962Smm error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx); 1899168404Spjd ASSERT(error == 0); 1900168404Spjd 1901168404Spjd /* 1902168404Spjd * Create root znode. Create minimal znode/vnode/zfsvfs 1903168404Spjd * to allow zfs_mknode to work. 1904168404Spjd */ 1905185029Spjd VATTR_NULL(&vattr); 1906168404Spjd vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE; 1907168404Spjd vattr.va_type = VDIR; 1908168404Spjd vattr.va_mode = S_IFDIR|0755; 1909185029Spjd vattr.va_uid = crgetuid(cr); 1910185029Spjd vattr.va_gid = crgetgid(cr); 1911168404Spjd 1912219089Spjd bzero(&zfsvfs, sizeof (zfsvfs_t)); 1913219089Spjd 1914168404Spjd rootzp = kmem_cache_alloc(znode_cache, KM_SLEEP); 1915199156Spjd zfs_znode_cache_constructor(rootzp, NULL, 0); 1916219089Spjd ASSERT(!POINTER_IS_VALID(rootzp->z_zfsvfs)); 1917219089Spjd rootzp->z_moved = 0; 1918168404Spjd rootzp->z_unlinked = 0; 1919168404Spjd rootzp->z_atime_dirty = 0; 1920219089Spjd rootzp->z_is_sa = USE_SA(version, os); 1921168404Spjd 1922199156Spjd vnode.v_type = VDIR; 1923199156Spjd vnode.v_data = rootzp; 1924199156Spjd rootzp->z_vnode = &vnode; 1925185029Spjd 1926168404Spjd zfsvfs.z_os = os; 1927168404Spjd zfsvfs.z_parent = &zfsvfs; 1928185029Spjd zfsvfs.z_version = version; 1929185029Spjd zfsvfs.z_use_fuids = USE_FUIDS(version, os); 1930219089Spjd zfsvfs.z_use_sa = USE_SA(version, os); 1931185029Spjd zfsvfs.z_norm = norm; 1932219089Spjd 1933219089Spjd error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END, 1934219089Spjd &zfsvfs.z_attr_table); 1935219089Spjd 1936219089Spjd ASSERT(error == 0); 1937219089Spjd 1938185029Spjd /* 1939185029Spjd * Fold case on file systems that are always or sometimes case 1940185029Spjd * insensitive. 1941185029Spjd */ 1942185029Spjd if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED) 1943185029Spjd zfsvfs.z_norm |= U8_TEXTPREP_TOUPPER; 1944168404Spjd 1945168404Spjd mutex_init(&zfsvfs.z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); 1946168404Spjd list_create(&zfsvfs.z_all_znodes, sizeof (znode_t), 1947168404Spjd offsetof(znode_t, z_link_node)); 1948168404Spjd 1949207334Spjd for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) 1950207334Spjd mutex_init(&zfsvfs.z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL); 1951207334Spjd 1952185029Spjd rootzp->z_zfsvfs = &zfsvfs; 1953209962Smm VERIFY(0 == zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr, 1954209962Smm cr, NULL, &acl_ids)); 1955219089Spjd zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, &acl_ids); 1956185029Spjd ASSERT3P(zp, ==, rootzp); 1957185029Spjd error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx); 1958168404Spjd ASSERT(error == 0); 1959209962Smm zfs_acl_ids_free(&acl_ids); 1960185029Spjd POINTER_INVALIDATE(&rootzp->z_zfsvfs); 1961168404Spjd 1962219089Spjd sa_handle_destroy(rootzp->z_sa_hdl); 1963209962Smm rootzp->z_vnode = NULL; 1964209962Smm kmem_cache_free(znode_cache, rootzp); 1965209962Smm 1966209962Smm /* 1967209962Smm * Create shares directory 1968209962Smm */ 1969209962Smm 1970209962Smm error = zfs_create_share_dir(&zfsvfs, tx); 1971209962Smm 1972209962Smm ASSERT(error == 0); 1973209962Smm 1974207334Spjd for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) 1975207334Spjd mutex_destroy(&zfsvfs.z_hold_mtx[i]); 1976168404Spjd} 1977185029Spjd 1978168404Spjd#endif /* _KERNEL */ 1979219089Spjd 1980168404Spjdstatic int 1981219089Spjdzfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table) 1982168404Spjd{ 1983219089Spjd uint64_t sa_obj = 0; 1984219089Spjd int error; 1985219089Spjd 1986219089Spjd error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj); 1987219089Spjd if (error != 0 && error != ENOENT) 1988219089Spjd return (error); 1989219089Spjd 1990219089Spjd error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table); 1991219089Spjd return (error); 1992219089Spjd} 1993219089Spjd 1994219089Spjdstatic int 1995219089Spjdzfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp, 1996219089Spjd dmu_buf_t **db, void *tag) 1997219089Spjd{ 1998168404Spjd dmu_object_info_t doi; 1999168404Spjd int error; 2000168404Spjd 2001219089Spjd if ((error = sa_buf_hold(osp, obj, tag, db)) != 0) 2002168404Spjd return (error); 2003168404Spjd 2004219089Spjd dmu_object_info_from_db(*db, &doi); 2005219089Spjd if ((doi.doi_bonus_type != DMU_OT_SA && 2006219089Spjd doi.doi_bonus_type != DMU_OT_ZNODE) || 2007219089Spjd doi.doi_bonus_type == DMU_OT_ZNODE && 2008168404Spjd doi.doi_bonus_size < sizeof (znode_phys_t)) { 2009219089Spjd sa_buf_rele(*db, tag); 2010219089Spjd return (ENOTSUP); 2011168404Spjd } 2012168404Spjd 2013219089Spjd error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp); 2014219089Spjd if (error != 0) { 2015219089Spjd sa_buf_rele(*db, tag); 2016219089Spjd return (error); 2017219089Spjd } 2018168404Spjd 2019168404Spjd return (0); 2020168404Spjd} 2021168404Spjd 2022219089Spjdvoid 2023219089Spjdzfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db, void *tag) 2024168404Spjd{ 2025219089Spjd sa_handle_destroy(hdl); 2026219089Spjd sa_buf_rele(db, tag); 2027219089Spjd} 2028219089Spjd 2029219089Spjd/* 2030219089Spjd * Given an object number, return its parent object number and whether 2031219089Spjd * or not the object is an extended attribute directory. 2032219089Spjd */ 2033219089Spjdstatic int 2034240955Smmzfs_obj_to_pobj(objset_t *osp, sa_handle_t *hdl, sa_attr_type_t *sa_table, 2035240955Smm uint64_t *pobjp, int *is_xattrdir) 2036219089Spjd{ 2037219089Spjd uint64_t parent; 2038219089Spjd uint64_t pflags; 2039219089Spjd uint64_t mode; 2040240955Smm uint64_t parent_mode; 2041219089Spjd sa_bulk_attr_t bulk[3]; 2042240955Smm sa_handle_t *sa_hdl; 2043240955Smm dmu_buf_t *sa_db; 2044219089Spjd int count = 0; 2045219089Spjd int error; 2046219089Spjd 2047219089Spjd SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL, 2048219089Spjd &parent, sizeof (parent)); 2049219089Spjd SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL, 2050219089Spjd &pflags, sizeof (pflags)); 2051219089Spjd SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL, 2052219089Spjd &mode, sizeof (mode)); 2053219089Spjd 2054219089Spjd if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0) 2055219089Spjd return (error); 2056219089Spjd 2057240955Smm /* 2058240955Smm * When a link is removed its parent pointer is not changed and will 2059240955Smm * be invalid. There are two cases where a link is removed but the 2060240955Smm * file stays around, when it goes to the delete queue and when there 2061240955Smm * are additional links. 2062240955Smm */ 2063240955Smm error = zfs_grab_sa_handle(osp, parent, &sa_hdl, &sa_db, FTAG); 2064240955Smm if (error != 0) 2065240955Smm return (error); 2066240955Smm 2067240955Smm error = sa_lookup(sa_hdl, ZPL_MODE, &parent_mode, sizeof (parent_mode)); 2068240955Smm zfs_release_sa_handle(sa_hdl, sa_db, FTAG); 2069240955Smm if (error != 0) 2070240955Smm return (error); 2071240955Smm 2072219089Spjd *is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode); 2073219089Spjd 2074240955Smm /* 2075240955Smm * Extended attributes can be applied to files, directories, etc. 2076240955Smm * Otherwise the parent must be a directory. 2077240955Smm */ 2078240955Smm if (!*is_xattrdir && !S_ISDIR(parent_mode)) 2079240955Smm return (EINVAL); 2080240955Smm 2081240955Smm *pobjp = parent; 2082240955Smm 2083219089Spjd return (0); 2084219089Spjd} 2085219089Spjd 2086219089Spjd/* 2087219089Spjd * Given an object number, return some zpl level statistics 2088219089Spjd */ 2089219089Spjdstatic int 2090219089Spjdzfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table, 2091219089Spjd zfs_stat_t *sb) 2092219089Spjd{ 2093219089Spjd sa_bulk_attr_t bulk[4]; 2094219089Spjd int count = 0; 2095219089Spjd 2096219089Spjd SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL, 2097219089Spjd &sb->zs_mode, sizeof (sb->zs_mode)); 2098219089Spjd SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL, 2099219089Spjd &sb->zs_gen, sizeof (sb->zs_gen)); 2100219089Spjd SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL, 2101219089Spjd &sb->zs_links, sizeof (sb->zs_links)); 2102219089Spjd SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL, 2103219089Spjd &sb->zs_ctime, sizeof (sb->zs_ctime)); 2104219089Spjd 2105219089Spjd return (sa_bulk_lookup(hdl, bulk, count)); 2106219089Spjd} 2107219089Spjd 2108219089Spjdstatic int 2109219089Spjdzfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl, 2110219089Spjd sa_attr_type_t *sa_table, char *buf, int len) 2111219089Spjd{ 2112219089Spjd sa_handle_t *sa_hdl; 2113219089Spjd sa_handle_t *prevhdl = NULL; 2114219089Spjd dmu_buf_t *prevdb = NULL; 2115219089Spjd dmu_buf_t *sa_db = NULL; 2116168404Spjd char *path = buf + len - 1; 2117168404Spjd int error; 2118168404Spjd 2119168404Spjd *path = '\0'; 2120219089Spjd sa_hdl = hdl; 2121168404Spjd 2122168404Spjd for (;;) { 2123168404Spjd uint64_t pobj; 2124168404Spjd char component[MAXNAMELEN + 2]; 2125168404Spjd size_t complen; 2126168404Spjd int is_xattrdir; 2127168404Spjd 2128219089Spjd if (prevdb) 2129219089Spjd zfs_release_sa_handle(prevhdl, prevdb, FTAG); 2130219089Spjd 2131240955Smm if ((error = zfs_obj_to_pobj(osp, sa_hdl, sa_table, &pobj, 2132168404Spjd &is_xattrdir)) != 0) 2133168404Spjd break; 2134168404Spjd 2135168404Spjd if (pobj == obj) { 2136168404Spjd if (path[0] != '/') 2137168404Spjd *--path = '/'; 2138168404Spjd break; 2139168404Spjd } 2140168404Spjd 2141168404Spjd component[0] = '/'; 2142168404Spjd if (is_xattrdir) { 2143168404Spjd (void) sprintf(component + 1, "<xattrdir>"); 2144168404Spjd } else { 2145185029Spjd error = zap_value_search(osp, pobj, obj, 2146185029Spjd ZFS_DIRENT_OBJ(-1ULL), component + 1); 2147168404Spjd if (error != 0) 2148168404Spjd break; 2149168404Spjd } 2150168404Spjd 2151168404Spjd complen = strlen(component); 2152168404Spjd path -= complen; 2153168404Spjd ASSERT(path >= buf); 2154168404Spjd bcopy(component, path, complen); 2155168404Spjd obj = pobj; 2156219089Spjd 2157219089Spjd if (sa_hdl != hdl) { 2158219089Spjd prevhdl = sa_hdl; 2159219089Spjd prevdb = sa_db; 2160219089Spjd } 2161219089Spjd error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db, FTAG); 2162219089Spjd if (error != 0) { 2163219089Spjd sa_hdl = prevhdl; 2164219089Spjd sa_db = prevdb; 2165219089Spjd break; 2166219089Spjd } 2167168404Spjd } 2168168404Spjd 2169219089Spjd if (sa_hdl != NULL && sa_hdl != hdl) { 2170219089Spjd ASSERT(sa_db != NULL); 2171219089Spjd zfs_release_sa_handle(sa_hdl, sa_db, FTAG); 2172219089Spjd } 2173219089Spjd 2174168404Spjd if (error == 0) 2175168404Spjd (void) memmove(buf, path, buf + len - path); 2176219089Spjd 2177168404Spjd return (error); 2178168404Spjd} 2179219089Spjd 2180219089Spjdint 2181219089Spjdzfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len) 2182219089Spjd{ 2183219089Spjd sa_attr_type_t *sa_table; 2184219089Spjd sa_handle_t *hdl; 2185219089Spjd dmu_buf_t *db; 2186219089Spjd int error; 2187219089Spjd 2188219089Spjd error = zfs_sa_setup(osp, &sa_table); 2189219089Spjd if (error != 0) 2190219089Spjd return (error); 2191219089Spjd 2192219089Spjd error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG); 2193219089Spjd if (error != 0) 2194219089Spjd return (error); 2195219089Spjd 2196219089Spjd error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len); 2197219089Spjd 2198219089Spjd zfs_release_sa_handle(hdl, db, FTAG); 2199219089Spjd return (error); 2200219089Spjd} 2201219089Spjd 2202219089Spjdint 2203219089Spjdzfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb, 2204219089Spjd char *buf, int len) 2205219089Spjd{ 2206219089Spjd char *path = buf + len - 1; 2207219089Spjd sa_attr_type_t *sa_table; 2208219089Spjd sa_handle_t *hdl; 2209219089Spjd dmu_buf_t *db; 2210219089Spjd int error; 2211219089Spjd 2212219089Spjd *path = '\0'; 2213219089Spjd 2214219089Spjd error = zfs_sa_setup(osp, &sa_table); 2215219089Spjd if (error != 0) 2216219089Spjd return (error); 2217219089Spjd 2218219089Spjd error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG); 2219219089Spjd if (error != 0) 2220219089Spjd return (error); 2221219089Spjd 2222219089Spjd error = zfs_obj_to_stats_impl(hdl, sa_table, sb); 2223219089Spjd if (error != 0) { 2224219089Spjd zfs_release_sa_handle(hdl, db, FTAG); 2225219089Spjd return (error); 2226219089Spjd } 2227219089Spjd 2228219089Spjd error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len); 2229219089Spjd 2230219089Spjd zfs_release_sa_handle(hdl, db, FTAG); 2231219089Spjd return (error); 2232219089Spjd} 2233