zfs_znode.c revision 243520
11553Srgrimes/* 21553Srgrimes * CDDL HEADER START 31553Srgrimes * 41553Srgrimes * The contents of this file are subject to the terms of the 51553Srgrimes * Common Development and Distribution License (the "License"). 61553Srgrimes * You may not use this file except in compliance with the License. 71553Srgrimes * 81553Srgrimes * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 91553Srgrimes * or http://www.opensolaris.org/os/licensing. 101553Srgrimes * See the License for the specific language governing permissions 111553Srgrimes * and limitations under the License. 121553Srgrimes * 131553Srgrimes * When distributing Covered Code, include this CDDL HEADER in each 141553Srgrimes * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 151553Srgrimes * If applicable, add the following below this CDDL HEADER, with the 161553Srgrimes * fields enclosed by brackets "[]" replaced with your own identifying 171553Srgrimes * information: Portions Copyright [yyyy] [name of copyright owner] 181553Srgrimes * 191553Srgrimes * CDDL HEADER END 201553Srgrimes */ 211553Srgrimes/* 221553Srgrimes * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 231553Srgrimes * Copyright (c) 2012 by Delphix. All rights reserved. 241553Srgrimes */ 251553Srgrimes 261553Srgrimes/* Portions Copyright 2007 Jeremy Teo */ 271553Srgrimes/* Portions Copyright 2011 Martin Matuska <mm@FreeBSD.org> */ 281553Srgrimes 2950479Speter#ifdef _KERNEL 301553Srgrimes#include <sys/types.h> 31240409Seadler#include <sys/param.h> 321553Srgrimes#include <sys/time.h> 331553Srgrimes#include <sys/systm.h> 341553Srgrimes#include <sys/sysmacros.h> 351553Srgrimes#include <sys/resource.h> 361553Srgrimes#include <sys/mntent.h> 371553Srgrimes#include <sys/u8_textprep.h> 381553Srgrimes#include <sys/dsl_dataset.h> 391553Srgrimes#include <sys/vfs.h> 401553Srgrimes#include <sys/vnode.h> 4125276Sjmg#include <sys/file.h> 421553Srgrimes#include <sys/kmem.h> 431553Srgrimes#include <sys/errno.h> 441553Srgrimes#include <sys/unistd.h> 455367Sjkh#include <sys/atomic.h> 465367Sjkh#include <sys/zfs_dir.h> 475367Sjkh#include <sys/zfs_acl.h> 4863795Sdwmalone#include <sys/zfs_ioctl.h> 4963795Sdwmalone#include <sys/zfs_rlock.h> 50155954Sjulian#include <sys/zfs_fuid.h> 515367Sjkh#include <sys/dnode.h> 521553Srgrimes#include <sys/fs/zfs.h> 531553Srgrimes#include <sys/kidmap.h> 541553Srgrimes#endif /* _KERNEL */ 551553Srgrimes 561553Srgrimes#include <sys/dmu.h> 5756485Scharnier#include <sys/refcount.h> 581553Srgrimes#include <sys/stat.h> 591553Srgrimes#include <sys/zap.h> 601553Srgrimes#include <sys/zfs_znode.h> 611553Srgrimes#include <sys/sa.h> 621553Srgrimes#include <sys/zfs_sa.h> 6375169Sru#include <sys/zfs_stat.h> 641553Srgrimes#include <sys/refcount.h> 6579755Sdd 6682470Scjc#include "zfs_prop.h" 6741498Sjkh#include "zfs_comutil.h" 68108247Strhodes 69131500Sru/* Used by fstat(1). */ 7082470ScjcSYSCTL_INT(_debug_sizeof, OID_AUTO, znode, CTLFLAG_RD, 0, sizeof(znode_t), 7157673Ssheldonh "sizeof(znode_t)"); 7279755Sdd 73108247Strhodes/* 7482470Scjc * Define ZNODE_STATS to turn on statistic gathering. By default, it is only 75131500Sru * turned on when DEBUG is also defined. 7641498Sjkh */ 771553Srgrimes#ifdef DEBUG 7882470Scjc#define ZNODE_STATS 791553Srgrimes#endif /* DEBUG */ 801553Srgrimes 8125276Sjmg#ifdef ZNODE_STATS 8225276Sjmg#define ZNODE_STAT_ADD(stat) ((stat)++) 8337820Sphk#else 84107312Sru#define ZNODE_STAT_ADD(stat) /* nothing */ 8525276Sjmg#endif /* ZNODE_STATS */ 861553Srgrimes 871553Srgrimes/* 881553Srgrimes * Functions needed for userland (ie: libzpool) are not put under 891553Srgrimes * #ifdef_KERNEL; the rest of the functions have dependencies 901553Srgrimes * (such as VFS logic) that will not compile easily in userland. 911553Srgrimes */ 921553Srgrimes#ifdef _KERNEL 931553Srgrimes/* 941553Srgrimes * Needed to close a small window in zfs_znode_move() that allows the zfsvfs to 951553Srgrimes * be freed before it can be safely accessed. 961553Srgrimes */ 97163255Strhodeskrwlock_t zfsvfs_lock; 98163255Strhodes 99163255Strhodesstatic kmem_cache_t *znode_cache = NULL; 100163255Strhodes 101163255Strhodes/*ARGSUSED*/ 102163255Strhodesstatic void 103163255Strhodesznode_evict_error(dmu_buf_t *dbuf, void *user_ptr) 104108247Strhodes{ 1051553Srgrimes /* 1061553Srgrimes * We should never drop all dbuf refs without first clearing 1071553Srgrimes * the eviction callback. 1081553Srgrimes */ 1091553Srgrimes panic("evicting znode %p\n", user_ptr); 1101553Srgrimes} 1111553Srgrimes 1121553Srgrimesextern struct vop_vector zfs_vnodeops; 1131553Srgrimesextern struct vop_vector zfs_fifoops; 11437820Sphkextern struct vop_vector zfs_shareops; 11537820Sphk 116108247Strhodes/* 11737820Sphk * XXX: We cannot use this function as a cache constructor, because 11837820Sphk * there is one global cache for all file systems and we need 11971898Sru * to pass vfsp here, which is not possible, because argument 12037820Sphk * 'cdrarg' is defined at kmem_cache_create() time. 12137820Sphk */ 122108247Strhodes/*ARGSUSED*/ 123108247Strhodesstatic int 1241553Srgrimeszfs_znode_cache_constructor(void *buf, void *arg, int kmflags) 12537820Sphk{ 126102940Sdwmalone znode_t *zp = buf; 127107312Sru vnode_t *vp; 128102940Sdwmalone vfs_t *vfsp = arg; 129102940Sdwmalone int error; 130102940Sdwmalone 131102940Sdwmalone POINTER_INVALIDATE(&zp->z_zfsvfs); 132102940Sdwmalone ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs)); 133102940Sdwmalone 134102940Sdwmalone if (vfsp != NULL) { 13537820Sphk error = getnewvnode("zfs", vfsp, &zfs_vnodeops, &vp); 13637820Sphk if (error != 0 && (kmflags & KM_NOSLEEP)) 13737820Sphk return (-1); 1381553Srgrimes ASSERT(error == 0); 139163255Strhodes vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 140186440Strhodes zp->z_vnode = vp; 141163255Strhodes vp->v_data = (caddr_t)zp; 142163255Strhodes VN_LOCK_AREC(vp); 143108247Strhodes VN_LOCK_ASHARE(vp); 1441553Srgrimes } else { 14525276Sjmg zp->z_vnode = NULL; 1461553Srgrimes } 14756485Scharnier 1481553Srgrimes list_link_init(&zp->z_link_node); 1491553Srgrimes 15079755Sdd mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL); 15163795Sdwmalone rw_init(&zp->z_parent_lock, NULL, RW_DEFAULT, NULL); 1525367Sjkh rw_init(&zp->z_name_lock, NULL, RW_DEFAULT, NULL); 15363795Sdwmalone mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL); 15463795Sdwmalone 15563795Sdwmalone mutex_init(&zp->z_range_lock, NULL, MUTEX_DEFAULT, NULL); 15663795Sdwmalone avl_create(&zp->z_range_avl, zfs_range_compare, 15763795Sdwmalone sizeof (rl_t), offsetof(rl_t, r_node)); 15863795Sdwmalone 15963795Sdwmalone zp->z_dirlocks = NULL; 160108247Strhodes zp->z_acl_cached = NULL; 16163795Sdwmalone zp->z_moved = 0; 16263795Sdwmalone return (0); 16363795Sdwmalone} 16463795Sdwmalone 16563795Sdwmalone/*ARGSUSED*/ 16663795Sdwmalonestatic void 167108247Strhodeszfs_znode_cache_destructor(void *buf, void *arg) 16863795Sdwmalone{ 16963795Sdwmalone znode_t *zp = buf; 17063795Sdwmalone 17163795Sdwmalone ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs)); 17263795Sdwmalone ASSERT(ZTOV(zp) == NULL); 17363795Sdwmalone vn_free(ZTOV(zp)); 17463795Sdwmalone ASSERT(!list_link_active(&zp->z_link_node)); 17563795Sdwmalone mutex_destroy(&zp->z_lock); 17663795Sdwmalone rw_destroy(&zp->z_parent_lock); 17763795Sdwmalone rw_destroy(&zp->z_name_lock); 1785367Sjkh mutex_destroy(&zp->z_acl_lock); 17982470Scjc avl_destroy(&zp->z_range_avl); 18056485Scharnier mutex_destroy(&zp->z_range_lock); 18163795Sdwmalone 18263795Sdwmalone ASSERT(zp->z_dirlocks == NULL); 18363795Sdwmalone ASSERT(zp->z_acl_cached == NULL); 18463795Sdwmalone} 18563795Sdwmalone 18663795Sdwmalone#ifdef ZNODE_STATS 18763795Sdwmalonestatic struct { 18863795Sdwmalone uint64_t zms_zfsvfs_invalid; 18963795Sdwmalone uint64_t zms_zfsvfs_recheck1; 190103855Sdcs uint64_t zms_zfsvfs_unmounted; 191103855Sdcs uint64_t zms_zfsvfs_recheck2; 192103855Sdcs uint64_t zms_obj_held; 193103855Sdcs uint64_t zms_vnode_locked; 194103855Sdcs uint64_t zms_not_only_dnlc; 195103855Sdcs} znode_move_stats; 196103855Sdcs#endif /* ZNODE_STATS */ 197103855Sdcs 198103855Sdcs#ifdef sun 199103855Sdcsstatic void 200103855Sdcszfs_znode_move_impl(znode_t *ozp, znode_t *nzp) 201110776Sthomas{ 202110776Sthomas vnode_t *vp; 203110776Sthomas 204110776Sthomas /* Copy fields. */ 205110776Sthomas nzp->z_zfsvfs = ozp->z_zfsvfs; 206110776Sthomas 207110776Sthomas /* Swap vnodes. */ 208110776Sthomas vp = nzp->z_vnode; 209110776Sthomas nzp->z_vnode = ozp->z_vnode; 21063795Sdwmalone ozp->z_vnode = vp; /* let destructor free the overwritten vnode */ 21163795Sdwmalone ZTOV(ozp)->v_data = ozp; 21263795Sdwmalone ZTOV(nzp)->v_data = nzp; 21363795Sdwmalone 21463795Sdwmalone nzp->z_id = ozp->z_id; 21563795Sdwmalone ASSERT(ozp->z_dirlocks == NULL); /* znode not in use */ 216108247Strhodes ASSERT(avl_numnodes(&ozp->z_range_avl) == 0); 21763795Sdwmalone nzp->z_unlinked = ozp->z_unlinked; 218108247Strhodes nzp->z_atime_dirty = ozp->z_atime_dirty; 21963795Sdwmalone nzp->z_zn_prefetch = ozp->z_zn_prefetch; 22063795Sdwmalone nzp->z_blksz = ozp->z_blksz; 22163795Sdwmalone nzp->z_seq = ozp->z_seq; 22263795Sdwmalone nzp->z_mapcnt = ozp->z_mapcnt; 22363795Sdwmalone nzp->z_gen = ozp->z_gen; 22463795Sdwmalone nzp->z_sync_cnt = ozp->z_sync_cnt; 22563795Sdwmalone nzp->z_is_sa = ozp->z_is_sa; 22663795Sdwmalone nzp->z_sa_hdl = ozp->z_sa_hdl; 22763795Sdwmalone bcopy(ozp->z_atime, nzp->z_atime, sizeof (uint64_t) * 2); 22863795Sdwmalone nzp->z_links = ozp->z_links; 229148580Skeramida nzp->z_size = ozp->z_size; 230110776Sthomas nzp->z_pflags = ozp->z_pflags; 231110776Sthomas nzp->z_uid = ozp->z_uid; 23263795Sdwmalone nzp->z_gid = ozp->z_gid; 23363795Sdwmalone nzp->z_mode = ozp->z_mode; 23463795Sdwmalone 23563795Sdwmalone /* 23663795Sdwmalone * Since this is just an idle znode and kmem is already dealing with 23763795Sdwmalone * memory pressure, release any cached ACL. 2385367Sjkh */ 2391553Srgrimes if (ozp->z_acl_cached) { 2401553Srgrimes zfs_acl_free(ozp->z_acl_cached); 241108247Strhodes ozp->z_acl_cached = NULL; 2421553Srgrimes } 2431553Srgrimes 2441553Srgrimes sa_set_userp(nzp->z_sa_hdl, nzp); 24557673Ssheldonh 246141851Sru /* 2475367Sjkh * Invalidate the original znode by clearing fields that provide a 2485367Sjkh * pointer back to the znode. Set the low bit of the vfs pointer to 2495367Sjkh * ensure that zfs_znode_move() recognizes the znode as invalid in any 25057673Ssheldonh * subsequent callback. 25157673Ssheldonh */ 2525367Sjkh ozp->z_sa_hdl = NULL; 2535367Sjkh POINTER_INVALIDATE(&ozp->z_zfsvfs); 254141846Sru 2551553Srgrimes /* 2561553Srgrimes * Mark the znode. 2571553Srgrimes */ 2581553Srgrimes nzp->z_moved = 1; 2591553Srgrimes ozp->z_moved = (uint8_t)-1; 2601553Srgrimes} 2615367Sjkh 2625367Sjkh/*ARGSUSED*/ 2631553Srgrimesstatic kmem_cbrc_t 2641553Srgrimeszfs_znode_move(void *buf, void *newbuf, size_t size, void *arg) 2651553Srgrimes{ 2661553Srgrimes znode_t *ozp = buf, *nzp = newbuf; 2671553Srgrimes zfsvfs_t *zfsvfs; 2681553Srgrimes vnode_t *vp; 2691553Srgrimes 2701553Srgrimes /* 27125276Sjmg * The znode is on the file system's list of known znodes if the vfs 27225276Sjmg * pointer is valid. We set the low bit of the vfs pointer when freeing 27325276Sjmg * the znode to invalidate it, and the memory patterns written by kmem 2741553Srgrimes * (baddcafe and deadbeef) set at least one of the two low bits. A newly 2751553Srgrimes * created znode sets the vfs pointer last of all to indicate that the 2761553Srgrimes * znode is known and in a valid state to be moved by this function. 2771553Srgrimes */ 2781553Srgrimes zfsvfs = ozp->z_zfsvfs; 2791553Srgrimes if (!POINTER_IS_VALID(zfsvfs)) { 2801553Srgrimes ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_invalid); 2811553Srgrimes return (KMEM_CBRC_DONT_KNOW); 28225276Sjmg } 28325276Sjmg 28425276Sjmg /* 2851553Srgrimes * Close a small window in which it's possible that the filesystem could 28625276Sjmg * be unmounted and freed, and zfsvfs, though valid in the previous 28725276Sjmg * statement, could point to unrelated memory by the time we try to 28825276Sjmg * prevent the filesystem from being unmounted. 289108247Strhodes */ 2905367Sjkh rw_enter(&zfsvfs_lock, RW_WRITER); 291108247Strhodes if (zfsvfs != ozp->z_zfsvfs) { 2921553Srgrimes rw_exit(&zfsvfs_lock); 2935367Sjkh ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_recheck1); 2941553Srgrimes return (KMEM_CBRC_DONT_KNOW); 2951553Srgrimes } 2961553Srgrimes 29725276Sjmg /* 29825276Sjmg * If the znode is still valid, then so is the file system. We know that 29925276Sjmg * no valid file system can be freed while we hold zfsvfs_lock, so we 30025276Sjmg * can safely ensure that the filesystem is not and will not be 3011553Srgrimes * unmounted. The next statement is equivalent to ZFS_ENTER(). 3021553Srgrimes */ 3031553Srgrimes rrw_enter(&zfsvfs->z_teardown_lock, RW_READER, FTAG); 3041553Srgrimes if (zfsvfs->z_unmounted) { 3051553Srgrimes ZFS_EXIT(zfsvfs); 3061553Srgrimes rw_exit(&zfsvfs_lock); 3071553Srgrimes ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_unmounted); 3081553Srgrimes return (KMEM_CBRC_DONT_KNOW); 30925276Sjmg } 31025276Sjmg rw_exit(&zfsvfs_lock); 3111553Srgrimes 3121553Srgrimes mutex_enter(&zfsvfs->z_znodes_lock); 3131553Srgrimes /* 3141553Srgrimes * Recheck the vfs pointer in case the znode was removed just before 3151553Srgrimes * acquiring the lock. 3161553Srgrimes */ 3171553Srgrimes if (zfsvfs != ozp->z_zfsvfs) { 31822984Sjoerg mutex_exit(&zfsvfs->z_znodes_lock); 3191553Srgrimes ZFS_EXIT(zfsvfs); 3201553Srgrimes ZNODE_STAT_ADD(znode_move_stats.zms_zfsvfs_recheck2); 3211553Srgrimes return (KMEM_CBRC_DONT_KNOW); 3221553Srgrimes } 323129865Sdwmalone 324129865Sdwmalone /* 325129865Sdwmalone * At this point we know that as long as we hold z_znodes_lock, the 326129865Sdwmalone * znode cannot be freed and fields within the znode can be safely 327129865Sdwmalone * accessed. Now, prevent a race with zfs_zget(). 328129865Sdwmalone */ 329129865Sdwmalone if (ZFS_OBJ_HOLD_TRYENTER(zfsvfs, ozp->z_id) == 0) { 330129865Sdwmalone mutex_exit(&zfsvfs->z_znodes_lock); 331129865Sdwmalone ZFS_EXIT(zfsvfs); 332129865Sdwmalone ZNODE_STAT_ADD(znode_move_stats.zms_obj_held); 333129865Sdwmalone return (KMEM_CBRC_LATER); 334129865Sdwmalone } 335129865Sdwmalone 336129865Sdwmalone vp = ZTOV(ozp); 337129865Sdwmalone if (mutex_tryenter(&vp->v_lock) == 0) { 338129865Sdwmalone ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id); 3391553Srgrimes mutex_exit(&zfsvfs->z_znodes_lock); 34025276Sjmg ZFS_EXIT(zfsvfs); 34125276Sjmg ZNODE_STAT_ADD(znode_move_stats.zms_vnode_locked); 34225276Sjmg return (KMEM_CBRC_LATER); 3431553Srgrimes } 34456485Scharnier 3451553Srgrimes /* Only move znodes that are referenced _only_ by the DNLC. */ 346157821Sjulian if (vp->v_count != 1 || !vn_in_dnlc(vp)) { 347162806Sru mutex_exit(&vp->v_lock); 348162806Sru ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id); 349162806Sru mutex_exit(&zfsvfs->z_znodes_lock); 350240409Seadler ZFS_EXIT(zfsvfs); 351240409Seadler ZNODE_STAT_ADD(znode_move_stats.zms_not_only_dnlc); 352240409Seadler return (KMEM_CBRC_LATER); 353240409Seadler } 354240409Seadler 355240409Seadler /* 356240409Seadler * The znode is known and in a valid state to move. We're holding the 357240409Seadler * locks needed to execute the critical section. 3581553Srgrimes */ 3591553Srgrimes zfs_znode_move_impl(ozp, nzp); 3601553Srgrimes mutex_exit(&vp->v_lock); 3611553Srgrimes ZFS_OBJ_HOLD_EXIT(zfsvfs, ozp->z_id); 3621553Srgrimes 3631553Srgrimes list_link_replace(&ozp->z_link_node, &nzp->z_link_node); 3641553Srgrimes mutex_exit(&zfsvfs->z_znodes_lock); 36522984Sjoerg ZFS_EXIT(zfsvfs); 36625276Sjmg 36725276Sjmg return (KMEM_CBRC_YES); 36825276Sjmg} 369131500Sru#endif /* sun */ 370131500Sru 37182470Scjcvoid 37222984Sjoergzfs_znode_init(void) 373131500Sru{ 374131500Sru /* 37522984Sjoerg * Initialize zcache 37622984Sjoerg */ 37722984Sjoerg rw_init(&zfsvfs_lock, NULL, RW_DEFAULT, NULL); 378131531Sru ASSERT(znode_cache == NULL); 379131531Sru znode_cache = kmem_cache_create("zfs_znode_cache", 38022984Sjoerg sizeof (znode_t), 0, /* zfs_znode_cache_constructor */ NULL, 38122984Sjoerg zfs_znode_cache_destructor, NULL, NULL, NULL, 0); 38222984Sjoerg kmem_cache_set_move(znode_cache, zfs_znode_move); 38322984Sjoerg} 38422984Sjoerg 38522984Sjoergvoid 38622984Sjoergzfs_znode_fini(void) 38782470Scjc{ 388131500Sru#ifdef sun 389141846Sru /* 39023004Smpp * Cleanup vfs & vnode ops 39122984Sjoerg */ 39229376Sjoerg zfs_remove_op_tables(); 39322984Sjoerg#endif /* sun */ 39423004Smpp 395131500Sru /* 396131500Sru * Cleanup zcache 397131500Sru */ 39829376Sjoerg if (znode_cache) 39929376Sjoerg kmem_cache_destroy(znode_cache); 40029376Sjoerg znode_cache = NULL; 401131500Sru rw_destroy(&zfsvfs_lock); 402131500Sru} 40322984Sjoerg 404141851Sru#ifdef sun 40522984Sjoergstruct vnodeops *zfs_dvnodeops; 40622984Sjoergstruct vnodeops *zfs_fvnodeops; 40722984Sjoergstruct vnodeops *zfs_symvnodeops; 408131500Srustruct vnodeops *zfs_xdvnodeops; 409131500Srustruct vnodeops *zfs_evnodeops; 41022984Sjoergstruct vnodeops *zfs_sharevnodeops; 41123004Smpp 4121553Srgrimesvoid 4131553Srgrimeszfs_remove_op_tables() 41425276Sjmg{ 41525276Sjmg /* 4161553Srgrimes * Remove vfs ops 417180380Ssobomax */ 418180380Ssobomax ASSERT(zfsfstype); 419180380Ssobomax (void) vfs_freevfsops_by_type(zfsfstype); 420180380Ssobomax zfsfstype = 0; 421180380Ssobomax 422180380Ssobomax /* 423180380Ssobomax * Remove vnode ops 424180380Ssobomax */ 425180380Ssobomax if (zfs_dvnodeops) 426180380Ssobomax vn_freevnodeops(zfs_dvnodeops); 427180380Ssobomax if (zfs_fvnodeops) 428180380Ssobomax vn_freevnodeops(zfs_fvnodeops); 429180380Ssobomax if (zfs_symvnodeops) 430180380Ssobomax vn_freevnodeops(zfs_symvnodeops); 431140442Sru if (zfs_xdvnodeops) 432140442Sru vn_freevnodeops(zfs_xdvnodeops); 433140442Sru if (zfs_evnodeops) 434140442Sru vn_freevnodeops(zfs_evnodeops); 435140442Sru if (zfs_sharevnodeops) 436140442Sru vn_freevnodeops(zfs_sharevnodeops); 437140442Sru 438140442Sru zfs_dvnodeops = NULL; 439140442Sru zfs_fvnodeops = NULL; 440140442Sru zfs_symvnodeops = NULL; 441140442Sru zfs_xdvnodeops = NULL; 442140442Sru zfs_evnodeops = NULL; 443140442Sru zfs_sharevnodeops = NULL; 444140442Sru} 445140442Sru 446140442Sruextern const fs_operation_def_t zfs_dvnodeops_template[]; 447140442Sruextern const fs_operation_def_t zfs_fvnodeops_template[]; 448140442Sruextern const fs_operation_def_t zfs_xdvnodeops_template[]; 449140442Sruextern const fs_operation_def_t zfs_symvnodeops_template[]; 4501553Srgrimesextern const fs_operation_def_t zfs_evnodeops_template[]; 4511553Srgrimesextern const fs_operation_def_t zfs_sharevnodeops_template[]; 4521553Srgrimes 4531553Srgrimesint 454108247Strhodeszfs_create_op_tables() 4551553Srgrimes{ 4561553Srgrimes int error; 457186440Strhodes 4581553Srgrimes /* 4591553Srgrimes * zfs_dvnodeops can be set if mod_remove() calls mod_installfs() 4601553Srgrimes * due to a failure to remove the the 2nd modlinkage (zfs_modldrv). 4611553Srgrimes * In this case we just return as the ops vectors are already set up. 4621553Srgrimes */ 46337820Sphk if (zfs_dvnodeops) 46437820Sphk return (0); 46537820Sphk 4661553Srgrimes error = vn_make_ops(MNTTYPE_ZFS, zfs_dvnodeops_template, 4671553Srgrimes &zfs_dvnodeops); 4681553Srgrimes if (error) 4691553Srgrimes return (error); 4701553Srgrimes 4711553Srgrimes error = vn_make_ops(MNTTYPE_ZFS, zfs_fvnodeops_template, 4721553Srgrimes &zfs_fvnodeops); 4731553Srgrimes if (error) 4741553Srgrimes return (error); 4751553Srgrimes 4761553Srgrimes error = vn_make_ops(MNTTYPE_ZFS, zfs_symvnodeops_template, 4771553Srgrimes &zfs_symvnodeops); 4781553Srgrimes if (error) 4791553Srgrimes return (error); 4801553Srgrimes 4811553Srgrimes error = vn_make_ops(MNTTYPE_ZFS, zfs_xdvnodeops_template, 4821553Srgrimes &zfs_xdvnodeops); 4835367Sjkh if (error) 48422984Sjoerg return (error); 48522984Sjoerg 48622984Sjoerg error = vn_make_ops(MNTTYPE_ZFS, zfs_evnodeops_template, 48750126Sgreen &zfs_evnodeops); 48850126Sgreen if (error) 48974875Scwt return (error); 49074875Scwt 49174875Scwt error = vn_make_ops(MNTTYPE_ZFS, zfs_sharevnodeops_template, 492129865Sdwmalone &zfs_sharevnodeops); 493194037Savg 494194037Savg return (error); 495194037Savg} 496194037Savg#endif /* sun */ 497129865Sdwmalone 498129865Sdwmaloneint 499129865Sdwmalonezfs_create_share_dir(zfsvfs_t *zfsvfs, dmu_tx_t *tx) 5001553Srgrimes{ 501140442Sru zfs_acl_ids_t acl_ids; 502140442Sru vattr_t vattr; 50367249Sdwmalone znode_t *sharezp; 5041553Srgrimes vnode_t *vp, vnode; 50556485Scharnier znode_t *zp; 50656485Scharnier int error; 50756485Scharnier 50825276Sjmg vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE; 50925276Sjmg vattr.va_type = VDIR; 51025276Sjmg vattr.va_mode = S_IFDIR|0555; 51125276Sjmg vattr.va_uid = crgetuid(kcred); 51225276Sjmg vattr.va_gid = crgetgid(kcred); 51325276Sjmg 51425276Sjmg sharezp = kmem_cache_alloc(znode_cache, KM_SLEEP); 51525276Sjmg zfs_znode_cache_constructor(sharezp, zfsvfs->z_parent->z_vfs, 0); 51625276Sjmg ASSERT(!POINTER_IS_VALID(sharezp->z_zfsvfs)); 51730184Sjoerg sharezp->z_moved = 0; 51830184Sjoerg sharezp->z_unlinked = 0; 519131500Sru sharezp->z_atime_dirty = 0; 520131500Sru sharezp->z_zfsvfs = zfsvfs; 52130184Sjoerg sharezp->z_is_sa = zfsvfs->z_use_sa; 522131500Sru 523131500Sru sharezp->z_vnode = &vnode; 52430184Sjoerg vnode.v_data = sharezp; 525107788Sru 526 vp = ZTOV(sharezp); 527 vp->v_type = VDIR; 528 529 VERIFY(0 == zfs_acl_ids_create(sharezp, IS_ROOT_NODE, &vattr, 530 kcred, NULL, &acl_ids)); 531 zfs_mknode(sharezp, &vattr, tx, kcred, IS_ROOT_NODE, &zp, &acl_ids); 532 ASSERT3P(zp, ==, sharezp); 533 POINTER_INVALIDATE(&sharezp->z_zfsvfs); 534 error = zap_add(zfsvfs->z_os, MASTER_NODE_OBJ, 535 ZFS_SHARES_DIR, 8, 1, &sharezp->z_id, tx); 536 zfsvfs->z_shares_dir = sharezp->z_id; 537 538 zfs_acl_ids_free(&acl_ids); 539 ZTOV(sharezp)->v_data = NULL; 540 ZTOV(sharezp)->v_count = 0; 541 ZTOV(sharezp)->v_holdcnt = 0; 542 zp->z_vnode = NULL; 543 sa_handle_destroy(sharezp->z_sa_hdl); 544 sharezp->z_vnode = NULL; 545 kmem_cache_free(znode_cache, sharezp); 546 547 return (error); 548} 549 550/* 551 * define a couple of values we need available 552 * for both 64 and 32 bit environments. 553 */ 554#ifndef NBITSMINOR64 555#define NBITSMINOR64 32 556#endif 557#ifndef MAXMAJ64 558#define MAXMAJ64 0xffffffffUL 559#endif 560#ifndef MAXMIN64 561#define MAXMIN64 0xffffffffUL 562#endif 563 564/* 565 * Create special expldev for ZFS private use. 566 * Can't use standard expldev since it doesn't do 567 * what we want. The standard expldev() takes a 568 * dev32_t in LP64 and expands it to a long dev_t. 569 * We need an interface that takes a dev32_t in ILP32 570 * and expands it to a long dev_t. 571 */ 572static uint64_t 573zfs_expldev(dev_t dev) 574{ 575 return (((uint64_t)major(dev) << NBITSMINOR64) | minor(dev)); 576} 577/* 578 * Special cmpldev for ZFS private use. 579 * Can't use standard cmpldev since it takes 580 * a long dev_t and compresses it to dev32_t in 581 * LP64. We need to do a compaction of a long dev_t 582 * to a dev32_t in ILP32. 583 */ 584dev_t 585zfs_cmpldev(uint64_t dev) 586{ 587 return (makedev((dev >> NBITSMINOR64), (dev & MAXMIN64))); 588} 589 590static void 591zfs_znode_sa_init(zfsvfs_t *zfsvfs, znode_t *zp, 592 dmu_buf_t *db, dmu_object_type_t obj_type, sa_handle_t *sa_hdl) 593{ 594 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs) || (zfsvfs == zp->z_zfsvfs)); 595 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id))); 596 597 mutex_enter(&zp->z_lock); 598 599 ASSERT(zp->z_sa_hdl == NULL); 600 ASSERT(zp->z_acl_cached == NULL); 601 if (sa_hdl == NULL) { 602 VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, zp, 603 SA_HDL_SHARED, &zp->z_sa_hdl)); 604 } else { 605 zp->z_sa_hdl = sa_hdl; 606 sa_set_userp(sa_hdl, zp); 607 } 608 609 zp->z_is_sa = (obj_type == DMU_OT_SA) ? B_TRUE : B_FALSE; 610 611 /* 612 * Slap on VROOT if we are the root znode 613 */ 614 if (zp->z_id == zfsvfs->z_root) 615 ZTOV(zp)->v_flag |= VROOT; 616 617 mutex_exit(&zp->z_lock); 618 vn_exists(ZTOV(zp)); 619} 620 621void 622zfs_znode_dmu_fini(znode_t *zp) 623{ 624 ASSERT(MUTEX_HELD(ZFS_OBJ_MUTEX(zp->z_zfsvfs, zp->z_id)) || 625 zp->z_unlinked || 626 RW_WRITE_HELD(&zp->z_zfsvfs->z_teardown_inactive_lock)); 627 628 sa_handle_destroy(zp->z_sa_hdl); 629 zp->z_sa_hdl = NULL; 630} 631 632static void 633zfs_vnode_forget(vnode_t *vp) 634{ 635 636 /* copied from insmntque_stddtr */ 637 vp->v_data = NULL; 638 vp->v_op = &dead_vnodeops; 639 vgone(vp); 640 vput(vp); 641} 642 643/* 644 * Construct a new znode/vnode and intialize. 645 * 646 * This does not do a call to dmu_set_user() that is 647 * up to the caller to do, in case you don't want to 648 * return the znode 649 */ 650static znode_t * 651zfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, int blksz, 652 dmu_object_type_t obj_type, sa_handle_t *hdl) 653{ 654 znode_t *zp; 655 vnode_t *vp; 656 uint64_t mode; 657 uint64_t parent; 658 sa_bulk_attr_t bulk[9]; 659 int count = 0; 660 661 zp = kmem_cache_alloc(znode_cache, KM_SLEEP); 662 zfs_znode_cache_constructor(zp, zfsvfs->z_parent->z_vfs, 0); 663 664 ASSERT(zp->z_dirlocks == NULL); 665 ASSERT(!POINTER_IS_VALID(zp->z_zfsvfs)); 666 zp->z_moved = 0; 667 668 /* 669 * Defer setting z_zfsvfs until the znode is ready to be a candidate for 670 * the zfs_znode_move() callback. 671 */ 672 zp->z_sa_hdl = NULL; 673 zp->z_unlinked = 0; 674 zp->z_atime_dirty = 0; 675 zp->z_mapcnt = 0; 676 zp->z_id = db->db_object; 677 zp->z_blksz = blksz; 678 zp->z_seq = 0x7A4653; 679 zp->z_sync_cnt = 0; 680 681 vp = ZTOV(zp); 682 683 zfs_znode_sa_init(zfsvfs, zp, db, obj_type, hdl); 684 685 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, &mode, 8); 686 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, &zp->z_gen, 8); 687 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, 688 &zp->z_size, 8); 689 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL, 690 &zp->z_links, 8); 691 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, 692 &zp->z_pflags, 8); 693 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, &parent, 8); 694 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, 695 &zp->z_atime, 16); 696 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, 697 &zp->z_uid, 8); 698 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, 699 &zp->z_gid, 8); 700 701 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count) != 0 || zp->z_gen == 0) { 702 if (hdl == NULL) 703 sa_handle_destroy(zp->z_sa_hdl); 704 zfs_vnode_forget(vp); 705 zp->z_vnode = NULL; 706 kmem_cache_free(znode_cache, zp); 707 return (NULL); 708 } 709 710 zp->z_mode = mode; 711 712 vp->v_type = IFTOVT((mode_t)mode); 713 714 switch (vp->v_type) { 715 case VDIR: 716 zp->z_zn_prefetch = B_TRUE; /* z_prefetch default is enabled */ 717 break; 718#ifdef sun 719 case VBLK: 720 case VCHR: 721 { 722 uint64_t rdev; 723 VERIFY(sa_lookup(zp->z_sa_hdl, SA_ZPL_RDEV(zfsvfs), 724 &rdev, sizeof (rdev)) == 0); 725 726 vp->v_rdev = zfs_cmpldev(rdev); 727 } 728 break; 729#endif /* sun */ 730 case VFIFO: 731#ifdef sun 732 case VSOCK: 733 case VDOOR: 734#endif /* sun */ 735 vp->v_op = &zfs_fifoops; 736 break; 737 case VREG: 738 if (parent == zfsvfs->z_shares_dir) { 739 ASSERT(zp->z_uid == 0 && zp->z_gid == 0); 740 vp->v_op = &zfs_shareops; 741 } 742 break; 743#ifdef sun 744 case VLNK: 745 vn_setops(vp, zfs_symvnodeops); 746 break; 747 default: 748 vn_setops(vp, zfs_evnodeops); 749 break; 750#endif /* sun */ 751 } 752 if (vp->v_type != VFIFO) 753 VN_LOCK_ASHARE(vp); 754 755 mutex_enter(&zfsvfs->z_znodes_lock); 756 list_insert_tail(&zfsvfs->z_all_znodes, zp); 757 membar_producer(); 758 /* 759 * Everything else must be valid before assigning z_zfsvfs makes the 760 * znode eligible for zfs_znode_move(). 761 */ 762 zp->z_zfsvfs = zfsvfs; 763 mutex_exit(&zfsvfs->z_znodes_lock); 764 765 VFS_HOLD(zfsvfs->z_vfs); 766 return (zp); 767} 768 769static uint64_t empty_xattr; 770static uint64_t pad[4]; 771static zfs_acl_phys_t acl_phys; 772/* 773 * Create a new DMU object to hold a zfs znode. 774 * 775 * IN: dzp - parent directory for new znode 776 * vap - file attributes for new znode 777 * tx - dmu transaction id for zap operations 778 * cr - credentials of caller 779 * flag - flags: 780 * IS_ROOT_NODE - new object will be root 781 * IS_XATTR - new object is an attribute 782 * bonuslen - length of bonus buffer 783 * setaclp - File/Dir initial ACL 784 * fuidp - Tracks fuid allocation. 785 * 786 * OUT: zpp - allocated znode 787 * 788 */ 789void 790zfs_mknode(znode_t *dzp, vattr_t *vap, dmu_tx_t *tx, cred_t *cr, 791 uint_t flag, znode_t **zpp, zfs_acl_ids_t *acl_ids) 792{ 793 uint64_t crtime[2], atime[2], mtime[2], ctime[2]; 794 uint64_t mode, size, links, parent, pflags; 795 uint64_t dzp_pflags = 0; 796 uint64_t rdev = 0; 797 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 798 dmu_buf_t *db; 799 timestruc_t now; 800 uint64_t gen, obj; 801 int err; 802 int bonuslen; 803 sa_handle_t *sa_hdl; 804 dmu_object_type_t obj_type; 805 sa_bulk_attr_t sa_attrs[ZPL_END]; 806 int cnt = 0; 807 zfs_acl_locator_cb_t locate = { 0 }; 808 809 ASSERT(vap && (vap->va_mask & (AT_TYPE|AT_MODE)) == (AT_TYPE|AT_MODE)); 810 811 if (zfsvfs->z_replay) { 812 obj = vap->va_nodeid; 813 now = vap->va_ctime; /* see zfs_replay_create() */ 814 gen = vap->va_nblocks; /* ditto */ 815 } else { 816 obj = 0; 817 gethrestime(&now); 818 gen = dmu_tx_get_txg(tx); 819 } 820 821 obj_type = zfsvfs->z_use_sa ? DMU_OT_SA : DMU_OT_ZNODE; 822 bonuslen = (obj_type == DMU_OT_SA) ? 823 DN_MAX_BONUSLEN : ZFS_OLD_ZNODE_PHYS_SIZE; 824 825 /* 826 * Create a new DMU object. 827 */ 828 /* 829 * There's currently no mechanism for pre-reading the blocks that will 830 * be needed to allocate a new object, so we accept the small chance 831 * that there will be an i/o error and we will fail one of the 832 * assertions below. 833 */ 834 if (vap->va_type == VDIR) { 835 if (zfsvfs->z_replay) { 836 err = zap_create_claim_norm(zfsvfs->z_os, obj, 837 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS, 838 obj_type, bonuslen, tx); 839 ASSERT0(err); 840 } else { 841 obj = zap_create_norm(zfsvfs->z_os, 842 zfsvfs->z_norm, DMU_OT_DIRECTORY_CONTENTS, 843 obj_type, bonuslen, tx); 844 } 845 } else { 846 if (zfsvfs->z_replay) { 847 err = dmu_object_claim(zfsvfs->z_os, obj, 848 DMU_OT_PLAIN_FILE_CONTENTS, 0, 849 obj_type, bonuslen, tx); 850 ASSERT0(err); 851 } else { 852 obj = dmu_object_alloc(zfsvfs->z_os, 853 DMU_OT_PLAIN_FILE_CONTENTS, 0, 854 obj_type, bonuslen, tx); 855 } 856 } 857 858 getnewvnode_reserve(1); 859 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj); 860 VERIFY(0 == sa_buf_hold(zfsvfs->z_os, obj, NULL, &db)); 861 862 /* 863 * If this is the root, fix up the half-initialized parent pointer 864 * to reference the just-allocated physical data area. 865 */ 866 if (flag & IS_ROOT_NODE) { 867 dzp->z_id = obj; 868 } else { 869 dzp_pflags = dzp->z_pflags; 870 } 871 872 /* 873 * If parent is an xattr, so am I. 874 */ 875 if (dzp_pflags & ZFS_XATTR) { 876 flag |= IS_XATTR; 877 } 878 879 if (zfsvfs->z_use_fuids) 880 pflags = ZFS_ARCHIVE | ZFS_AV_MODIFIED; 881 else 882 pflags = 0; 883 884 if (vap->va_type == VDIR) { 885 size = 2; /* contents ("." and "..") */ 886 links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1; 887 } else { 888 size = links = 0; 889 } 890 891 if (vap->va_type == VBLK || vap->va_type == VCHR) { 892 rdev = zfs_expldev(vap->va_rdev); 893 } 894 895 parent = dzp->z_id; 896 mode = acl_ids->z_mode; 897 if (flag & IS_XATTR) 898 pflags |= ZFS_XATTR; 899 900 /* 901 * No execs denied will be deterimed when zfs_mode_compute() is called. 902 */ 903 pflags |= acl_ids->z_aclp->z_hints & 904 (ZFS_ACL_TRIVIAL|ZFS_INHERIT_ACE|ZFS_ACL_AUTO_INHERIT| 905 ZFS_ACL_DEFAULTED|ZFS_ACL_PROTECTED); 906 907 ZFS_TIME_ENCODE(&now, crtime); 908 ZFS_TIME_ENCODE(&now, ctime); 909 910 if (vap->va_mask & AT_ATIME) { 911 ZFS_TIME_ENCODE(&vap->va_atime, atime); 912 } else { 913 ZFS_TIME_ENCODE(&now, atime); 914 } 915 916 if (vap->va_mask & AT_MTIME) { 917 ZFS_TIME_ENCODE(&vap->va_mtime, mtime); 918 } else { 919 ZFS_TIME_ENCODE(&now, mtime); 920 } 921 922 /* Now add in all of the "SA" attributes */ 923 VERIFY(0 == sa_handle_get_from_db(zfsvfs->z_os, db, NULL, SA_HDL_SHARED, 924 &sa_hdl)); 925 926 /* 927 * Setup the array of attributes to be replaced/set on the new file 928 * 929 * order for DMU_OT_ZNODE is critical since it needs to be constructed 930 * in the old znode_phys_t format. Don't change this ordering 931 */ 932 933 if (obj_type == DMU_OT_ZNODE) { 934 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs), 935 NULL, &atime, 16); 936 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs), 937 NULL, &mtime, 16); 938 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs), 939 NULL, &ctime, 16); 940 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs), 941 NULL, &crtime, 16); 942 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs), 943 NULL, &gen, 8); 944 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs), 945 NULL, &mode, 8); 946 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs), 947 NULL, &size, 8); 948 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs), 949 NULL, &parent, 8); 950 } else { 951 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MODE(zfsvfs), 952 NULL, &mode, 8); 953 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_SIZE(zfsvfs), 954 NULL, &size, 8); 955 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GEN(zfsvfs), 956 NULL, &gen, 8); 957 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL, 958 &acl_ids->z_fuid, 8); 959 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL, 960 &acl_ids->z_fgid, 8); 961 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PARENT(zfsvfs), 962 NULL, &parent, 8); 963 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs), 964 NULL, &pflags, 8); 965 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ATIME(zfsvfs), 966 NULL, &atime, 16); 967 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_MTIME(zfsvfs), 968 NULL, &mtime, 16); 969 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CTIME(zfsvfs), 970 NULL, &ctime, 16); 971 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_CRTIME(zfsvfs), 972 NULL, &crtime, 16); 973 } 974 975 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_LINKS(zfsvfs), NULL, &links, 8); 976 977 if (obj_type == DMU_OT_ZNODE) { 978 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_XATTR(zfsvfs), NULL, 979 &empty_xattr, 8); 980 } 981 if (obj_type == DMU_OT_ZNODE || 982 (vap->va_type == VBLK || vap->va_type == VCHR)) { 983 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_RDEV(zfsvfs), 984 NULL, &rdev, 8); 985 986 } 987 if (obj_type == DMU_OT_ZNODE) { 988 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_FLAGS(zfsvfs), 989 NULL, &pflags, 8); 990 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_UID(zfsvfs), NULL, 991 &acl_ids->z_fuid, 8); 992 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_GID(zfsvfs), NULL, 993 &acl_ids->z_fgid, 8); 994 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_PAD(zfsvfs), NULL, pad, 995 sizeof (uint64_t) * 4); 996 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_ZNODE_ACL(zfsvfs), NULL, 997 &acl_phys, sizeof (zfs_acl_phys_t)); 998 } else if (acl_ids->z_aclp->z_version >= ZFS_ACL_VERSION_FUID) { 999 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_COUNT(zfsvfs), NULL, 1000 &acl_ids->z_aclp->z_acl_count, 8); 1001 locate.cb_aclp = acl_ids->z_aclp; 1002 SA_ADD_BULK_ATTR(sa_attrs, cnt, SA_ZPL_DACL_ACES(zfsvfs), 1003 zfs_acl_data_locator, &locate, 1004 acl_ids->z_aclp->z_acl_bytes); 1005 mode = zfs_mode_compute(mode, acl_ids->z_aclp, &pflags, 1006 acl_ids->z_fuid, acl_ids->z_fgid); 1007 } 1008 1009 VERIFY(sa_replace_all_by_template(sa_hdl, sa_attrs, cnt, tx) == 0); 1010 1011 if (!(flag & IS_ROOT_NODE)) { 1012 *zpp = zfs_znode_alloc(zfsvfs, db, 0, obj_type, sa_hdl); 1013 ASSERT(*zpp != NULL); 1014 } else { 1015 /* 1016 * If we are creating the root node, the "parent" we 1017 * passed in is the znode for the root. 1018 */ 1019 *zpp = dzp; 1020 1021 (*zpp)->z_sa_hdl = sa_hdl; 1022 } 1023 1024 (*zpp)->z_pflags = pflags; 1025 (*zpp)->z_mode = mode; 1026 1027 if (vap->va_mask & AT_XVATTR) 1028 zfs_xvattr_set(*zpp, (xvattr_t *)vap, tx); 1029 1030 if (obj_type == DMU_OT_ZNODE || 1031 acl_ids->z_aclp->z_version < ZFS_ACL_VERSION_FUID) { 1032 err = zfs_aclset_common(*zpp, acl_ids->z_aclp, cr, tx); 1033 ASSERT0(err); 1034 } 1035 if (!(flag & IS_ROOT_NODE)) { 1036 vnode_t *vp; 1037 1038 vp = ZTOV(*zpp); 1039 vp->v_vflag |= VV_FORCEINSMQ; 1040 err = insmntque(vp, zfsvfs->z_vfs); 1041 vp->v_vflag &= ~VV_FORCEINSMQ; 1042 KASSERT(err == 0, ("insmntque() failed: error %d", err)); 1043 } 1044 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj); 1045 getnewvnode_drop_reserve(); 1046} 1047 1048/* 1049 * zfs_xvattr_set only updates the in-core attributes 1050 * it is assumed the caller will be doing an sa_bulk_update 1051 * to push the changes out 1052 */ 1053void 1054zfs_xvattr_set(znode_t *zp, xvattr_t *xvap, dmu_tx_t *tx) 1055{ 1056 xoptattr_t *xoap; 1057 1058 xoap = xva_getxoptattr(xvap); 1059 ASSERT(xoap); 1060 1061 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) { 1062 uint64_t times[2]; 1063 ZFS_TIME_ENCODE(&xoap->xoa_createtime, times); 1064 (void) sa_update(zp->z_sa_hdl, SA_ZPL_CRTIME(zp->z_zfsvfs), 1065 ×, sizeof (times), tx); 1066 XVA_SET_RTN(xvap, XAT_CREATETIME); 1067 } 1068 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) { 1069 ZFS_ATTR_SET(zp, ZFS_READONLY, xoap->xoa_readonly, 1070 zp->z_pflags, tx); 1071 XVA_SET_RTN(xvap, XAT_READONLY); 1072 } 1073 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) { 1074 ZFS_ATTR_SET(zp, ZFS_HIDDEN, xoap->xoa_hidden, 1075 zp->z_pflags, tx); 1076 XVA_SET_RTN(xvap, XAT_HIDDEN); 1077 } 1078 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) { 1079 ZFS_ATTR_SET(zp, ZFS_SYSTEM, xoap->xoa_system, 1080 zp->z_pflags, tx); 1081 XVA_SET_RTN(xvap, XAT_SYSTEM); 1082 } 1083 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) { 1084 ZFS_ATTR_SET(zp, ZFS_ARCHIVE, xoap->xoa_archive, 1085 zp->z_pflags, tx); 1086 XVA_SET_RTN(xvap, XAT_ARCHIVE); 1087 } 1088 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { 1089 ZFS_ATTR_SET(zp, ZFS_IMMUTABLE, xoap->xoa_immutable, 1090 zp->z_pflags, tx); 1091 XVA_SET_RTN(xvap, XAT_IMMUTABLE); 1092 } 1093 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { 1094 ZFS_ATTR_SET(zp, ZFS_NOUNLINK, xoap->xoa_nounlink, 1095 zp->z_pflags, tx); 1096 XVA_SET_RTN(xvap, XAT_NOUNLINK); 1097 } 1098 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { 1099 ZFS_ATTR_SET(zp, ZFS_APPENDONLY, xoap->xoa_appendonly, 1100 zp->z_pflags, tx); 1101 XVA_SET_RTN(xvap, XAT_APPENDONLY); 1102 } 1103 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { 1104 ZFS_ATTR_SET(zp, ZFS_NODUMP, xoap->xoa_nodump, 1105 zp->z_pflags, tx); 1106 XVA_SET_RTN(xvap, XAT_NODUMP); 1107 } 1108 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) { 1109 ZFS_ATTR_SET(zp, ZFS_OPAQUE, xoap->xoa_opaque, 1110 zp->z_pflags, tx); 1111 XVA_SET_RTN(xvap, XAT_OPAQUE); 1112 } 1113 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { 1114 ZFS_ATTR_SET(zp, ZFS_AV_QUARANTINED, 1115 xoap->xoa_av_quarantined, zp->z_pflags, tx); 1116 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED); 1117 } 1118 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { 1119 ZFS_ATTR_SET(zp, ZFS_AV_MODIFIED, xoap->xoa_av_modified, 1120 zp->z_pflags, tx); 1121 XVA_SET_RTN(xvap, XAT_AV_MODIFIED); 1122 } 1123 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) { 1124 zfs_sa_set_scanstamp(zp, xvap, tx); 1125 XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP); 1126 } 1127 if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { 1128 ZFS_ATTR_SET(zp, ZFS_REPARSE, xoap->xoa_reparse, 1129 zp->z_pflags, tx); 1130 XVA_SET_RTN(xvap, XAT_REPARSE); 1131 } 1132 if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) { 1133 ZFS_ATTR_SET(zp, ZFS_OFFLINE, xoap->xoa_offline, 1134 zp->z_pflags, tx); 1135 XVA_SET_RTN(xvap, XAT_OFFLINE); 1136 } 1137 if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) { 1138 ZFS_ATTR_SET(zp, ZFS_SPARSE, xoap->xoa_sparse, 1139 zp->z_pflags, tx); 1140 XVA_SET_RTN(xvap, XAT_SPARSE); 1141 } 1142} 1143 1144int 1145zfs_zget(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp) 1146{ 1147 dmu_object_info_t doi; 1148 dmu_buf_t *db; 1149 znode_t *zp; 1150 vnode_t *vp; 1151 sa_handle_t *hdl; 1152 struct thread *td; 1153 int locked; 1154 int err; 1155 1156 td = curthread; 1157 getnewvnode_reserve(1); 1158again: 1159 *zpp = NULL; 1160 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num); 1161 1162 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db); 1163 if (err) { 1164 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1165 getnewvnode_drop_reserve(); 1166 return (err); 1167 } 1168 1169 dmu_object_info_from_db(db, &doi); 1170 if (doi.doi_bonus_type != DMU_OT_SA && 1171 (doi.doi_bonus_type != DMU_OT_ZNODE || 1172 (doi.doi_bonus_type == DMU_OT_ZNODE && 1173 doi.doi_bonus_size < sizeof (znode_phys_t)))) { 1174 sa_buf_rele(db, NULL); 1175 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1176 getnewvnode_drop_reserve(); 1177 return (EINVAL); 1178 } 1179 1180 hdl = dmu_buf_get_user(db); 1181 if (hdl != NULL) { 1182 zp = sa_get_userdata(hdl); 1183 1184 1185 /* 1186 * Since "SA" does immediate eviction we 1187 * should never find a sa handle that doesn't 1188 * know about the znode. 1189 */ 1190 1191 ASSERT3P(zp, !=, NULL); 1192 1193 mutex_enter(&zp->z_lock); 1194 ASSERT3U(zp->z_id, ==, obj_num); 1195 if (zp->z_unlinked) { 1196 err = ENOENT; 1197 } else { 1198 vp = ZTOV(zp); 1199 *zpp = zp; 1200 err = 0; 1201 } 1202 sa_buf_rele(db, NULL); 1203 1204 /* Don't let the vnode disappear after ZFS_OBJ_HOLD_EXIT. */ 1205 if (err == 0) 1206 VN_HOLD(vp); 1207 1208 mutex_exit(&zp->z_lock); 1209 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1210 1211 if (err == 0) { 1212 locked = VOP_ISLOCKED(vp); 1213 VI_LOCK(vp); 1214 if ((vp->v_iflag & VI_DOOMED) != 0 && 1215 locked != LK_EXCLUSIVE) { 1216 /* 1217 * The vnode is doomed and this thread doesn't 1218 * hold the exclusive lock on it, so the vnode 1219 * must be being reclaimed by another thread. 1220 * Otherwise the doomed vnode is being reclaimed 1221 * by this thread and zfs_zget is called from 1222 * ZIL internals. 1223 */ 1224 VI_UNLOCK(vp); 1225 VN_RELE(vp); 1226 goto again; 1227 } 1228 VI_UNLOCK(vp); 1229 } 1230 getnewvnode_drop_reserve(); 1231 return (err); 1232 } 1233 1234 /* 1235 * Not found create new znode/vnode 1236 * but only if file exists. 1237 * 1238 * There is a small window where zfs_vget() could 1239 * find this object while a file create is still in 1240 * progress. This is checked for in zfs_znode_alloc() 1241 * 1242 * if zfs_znode_alloc() fails it will drop the hold on the 1243 * bonus buffer. 1244 */ 1245 zp = zfs_znode_alloc(zfsvfs, db, doi.doi_data_block_size, 1246 doi.doi_bonus_type, NULL); 1247 if (zp == NULL) { 1248 err = ENOENT; 1249 } else { 1250 *zpp = zp; 1251 } 1252 if (err == 0) { 1253 vnode_t *vp = ZTOV(zp); 1254 1255 err = insmntque(vp, zfsvfs->z_vfs); 1256 if (err == 0) 1257 VOP_UNLOCK(vp, 0); 1258 else { 1259 zp->z_vnode = NULL; 1260 zfs_znode_dmu_fini(zp); 1261 zfs_znode_free(zp); 1262 *zpp = NULL; 1263 } 1264 } 1265 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1266 getnewvnode_drop_reserve(); 1267 return (err); 1268} 1269 1270int 1271zfs_rezget(znode_t *zp) 1272{ 1273 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1274 dmu_object_info_t doi; 1275 dmu_buf_t *db; 1276 vnode_t *vp; 1277 uint64_t obj_num = zp->z_id; 1278 uint64_t mode, size; 1279 sa_bulk_attr_t bulk[8]; 1280 int err; 1281 int count = 0; 1282 uint64_t gen; 1283 1284 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num); 1285 1286 mutex_enter(&zp->z_acl_lock); 1287 if (zp->z_acl_cached) { 1288 zfs_acl_free(zp->z_acl_cached); 1289 zp->z_acl_cached = NULL; 1290 } 1291 1292 mutex_exit(&zp->z_acl_lock); 1293 ASSERT(zp->z_sa_hdl == NULL); 1294 err = sa_buf_hold(zfsvfs->z_os, obj_num, NULL, &db); 1295 if (err) { 1296 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1297 return (err); 1298 } 1299 1300 dmu_object_info_from_db(db, &doi); 1301 if (doi.doi_bonus_type != DMU_OT_SA && 1302 (doi.doi_bonus_type != DMU_OT_ZNODE || 1303 (doi.doi_bonus_type == DMU_OT_ZNODE && 1304 doi.doi_bonus_size < sizeof (znode_phys_t)))) { 1305 sa_buf_rele(db, NULL); 1306 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1307 return (EINVAL); 1308 } 1309 1310 zfs_znode_sa_init(zfsvfs, zp, db, doi.doi_bonus_type, NULL); 1311 size = zp->z_size; 1312 1313 /* reload cached values */ 1314 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GEN(zfsvfs), NULL, 1315 &gen, sizeof (gen)); 1316 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, 1317 &zp->z_size, sizeof (zp->z_size)); 1318 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_LINKS(zfsvfs), NULL, 1319 &zp->z_links, sizeof (zp->z_links)); 1320 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, 1321 &zp->z_pflags, sizeof (zp->z_pflags)); 1322 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, 1323 &zp->z_atime, sizeof (zp->z_atime)); 1324 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, 1325 &zp->z_uid, sizeof (zp->z_uid)); 1326 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), NULL, 1327 &zp->z_gid, sizeof (zp->z_gid)); 1328 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, 1329 &mode, sizeof (mode)); 1330 1331 if (sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) { 1332 zfs_znode_dmu_fini(zp); 1333 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1334 return (EIO); 1335 } 1336 1337 zp->z_mode = mode; 1338 1339 if (gen != zp->z_gen) { 1340 zfs_znode_dmu_fini(zp); 1341 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1342 return (EIO); 1343 } 1344 1345 /* 1346 * XXXPJD: Not sure how is that possible, but under heavy 1347 * zfs recv -F load it happens that z_gen is the same, but 1348 * vnode type is different than znode type. This would mean 1349 * that for example regular file was replaced with directory 1350 * which has the same object number. 1351 */ 1352 vp = ZTOV(zp); 1353 if (vp != NULL && 1354 vp->v_type != IFTOVT((mode_t)zp->z_mode)) { 1355 zfs_znode_dmu_fini(zp); 1356 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1357 return (EIO); 1358 } 1359 1360 zp->z_unlinked = (zp->z_links == 0); 1361 zp->z_blksz = doi.doi_data_block_size; 1362 if (vp != NULL) { 1363 vn_pages_remove(vp, 0, 0); 1364 if (zp->z_size != size) 1365 vnode_pager_setsize(vp, zp->z_size); 1366 } 1367 1368 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 1369 1370 return (0); 1371} 1372 1373void 1374zfs_znode_delete(znode_t *zp, dmu_tx_t *tx) 1375{ 1376 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1377 objset_t *os = zfsvfs->z_os; 1378 uint64_t obj = zp->z_id; 1379 uint64_t acl_obj = zfs_external_acl(zp); 1380 1381 ZFS_OBJ_HOLD_ENTER(zfsvfs, obj); 1382 if (acl_obj) { 1383 VERIFY(!zp->z_is_sa); 1384 VERIFY(0 == dmu_object_free(os, acl_obj, tx)); 1385 } 1386 VERIFY(0 == dmu_object_free(os, obj, tx)); 1387 zfs_znode_dmu_fini(zp); 1388 ZFS_OBJ_HOLD_EXIT(zfsvfs, obj); 1389 zfs_znode_free(zp); 1390} 1391 1392void 1393zfs_zinactive(znode_t *zp) 1394{ 1395 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1396 uint64_t z_id = zp->z_id; 1397 1398 ASSERT(zp->z_sa_hdl); 1399 1400 /* 1401 * Don't allow a zfs_zget() while were trying to release this znode 1402 */ 1403 ZFS_OBJ_HOLD_ENTER(zfsvfs, z_id); 1404 1405 mutex_enter(&zp->z_lock); 1406 1407 /* 1408 * If this was the last reference to a file with no links, 1409 * remove the file from the file system. 1410 */ 1411 if (zp->z_unlinked) { 1412 mutex_exit(&zp->z_lock); 1413 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id); 1414 zfs_rmnode(zp); 1415 return; 1416 } 1417 1418 mutex_exit(&zp->z_lock); 1419 zfs_znode_dmu_fini(zp); 1420 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id); 1421 zfs_znode_free(zp); 1422} 1423 1424void 1425zfs_znode_free(znode_t *zp) 1426{ 1427 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1428 1429 ASSERT(zp->z_sa_hdl == NULL); 1430 zp->z_vnode = NULL; 1431 mutex_enter(&zfsvfs->z_znodes_lock); 1432 POINTER_INVALIDATE(&zp->z_zfsvfs); 1433 list_remove(&zfsvfs->z_all_znodes, zp); 1434 mutex_exit(&zfsvfs->z_znodes_lock); 1435 1436 if (zp->z_acl_cached) { 1437 zfs_acl_free(zp->z_acl_cached); 1438 zp->z_acl_cached = NULL; 1439 } 1440 1441 kmem_cache_free(znode_cache, zp); 1442 1443 VFS_RELE(zfsvfs->z_vfs); 1444} 1445 1446void 1447zfs_tstamp_update_setup(znode_t *zp, uint_t flag, uint64_t mtime[2], 1448 uint64_t ctime[2], boolean_t have_tx) 1449{ 1450 timestruc_t now; 1451 1452 gethrestime(&now); 1453 1454 if (have_tx) { /* will sa_bulk_update happen really soon? */ 1455 zp->z_atime_dirty = 0; 1456 zp->z_seq++; 1457 } else { 1458 zp->z_atime_dirty = 1; 1459 } 1460 1461 if (flag & AT_ATIME) { 1462 ZFS_TIME_ENCODE(&now, zp->z_atime); 1463 } 1464 1465 if (flag & AT_MTIME) { 1466 ZFS_TIME_ENCODE(&now, mtime); 1467 if (zp->z_zfsvfs->z_use_fuids) { 1468 zp->z_pflags |= (ZFS_ARCHIVE | 1469 ZFS_AV_MODIFIED); 1470 } 1471 } 1472 1473 if (flag & AT_CTIME) { 1474 ZFS_TIME_ENCODE(&now, ctime); 1475 if (zp->z_zfsvfs->z_use_fuids) 1476 zp->z_pflags |= ZFS_ARCHIVE; 1477 } 1478} 1479 1480/* 1481 * Grow the block size for a file. 1482 * 1483 * IN: zp - znode of file to free data in. 1484 * size - requested block size 1485 * tx - open transaction. 1486 * 1487 * NOTE: this function assumes that the znode is write locked. 1488 */ 1489void 1490zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx) 1491{ 1492 int error; 1493 u_longlong_t dummy; 1494 1495 if (size <= zp->z_blksz) 1496 return; 1497 /* 1498 * If the file size is already greater than the current blocksize, 1499 * we will not grow. If there is more than one block in a file, 1500 * the blocksize cannot change. 1501 */ 1502 if (zp->z_blksz && zp->z_size > zp->z_blksz) 1503 return; 1504 1505 error = dmu_object_set_blocksize(zp->z_zfsvfs->z_os, zp->z_id, 1506 size, 0, tx); 1507 1508 if (error == ENOTSUP) 1509 return; 1510 ASSERT0(error); 1511 1512 /* What blocksize did we actually get? */ 1513 dmu_object_size_from_db(sa_get_db(zp->z_sa_hdl), &zp->z_blksz, &dummy); 1514} 1515 1516#ifdef sun 1517/* 1518 * This is a dummy interface used when pvn_vplist_dirty() should *not* 1519 * be calling back into the fs for a putpage(). E.g.: when truncating 1520 * a file, the pages being "thrown away* don't need to be written out. 1521 */ 1522/* ARGSUSED */ 1523static int 1524zfs_no_putpage(vnode_t *vp, page_t *pp, u_offset_t *offp, size_t *lenp, 1525 int flags, cred_t *cr) 1526{ 1527 ASSERT(0); 1528 return (0); 1529} 1530#endif /* sun */ 1531 1532/* 1533 * Increase the file length 1534 * 1535 * IN: zp - znode of file to free data in. 1536 * end - new end-of-file 1537 * 1538 * RETURN: 0 if success 1539 * error code if failure 1540 */ 1541static int 1542zfs_extend(znode_t *zp, uint64_t end) 1543{ 1544 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1545 dmu_tx_t *tx; 1546 rl_t *rl; 1547 uint64_t newblksz; 1548 int error; 1549 1550 /* 1551 * We will change zp_size, lock the whole file. 1552 */ 1553 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER); 1554 1555 /* 1556 * Nothing to do if file already at desired length. 1557 */ 1558 if (end <= zp->z_size) { 1559 zfs_range_unlock(rl); 1560 return (0); 1561 } 1562top: 1563 tx = dmu_tx_create(zfsvfs->z_os); 1564 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); 1565 zfs_sa_upgrade_txholds(tx, zp); 1566 if (end > zp->z_blksz && 1567 (!ISP2(zp->z_blksz) || zp->z_blksz < zfsvfs->z_max_blksz)) { 1568 /* 1569 * We are growing the file past the current block size. 1570 */ 1571 if (zp->z_blksz > zp->z_zfsvfs->z_max_blksz) { 1572 ASSERT(!ISP2(zp->z_blksz)); 1573 newblksz = MIN(end, SPA_MAXBLOCKSIZE); 1574 } else { 1575 newblksz = MIN(end, zp->z_zfsvfs->z_max_blksz); 1576 } 1577 dmu_tx_hold_write(tx, zp->z_id, 0, newblksz); 1578 } else { 1579 newblksz = 0; 1580 } 1581 1582 error = dmu_tx_assign(tx, TXG_NOWAIT); 1583 if (error) { 1584 if (error == ERESTART) { 1585 dmu_tx_wait(tx); 1586 dmu_tx_abort(tx); 1587 goto top; 1588 } 1589 dmu_tx_abort(tx); 1590 zfs_range_unlock(rl); 1591 return (error); 1592 } 1593 1594 if (newblksz) 1595 zfs_grow_blocksize(zp, newblksz, tx); 1596 1597 zp->z_size = end; 1598 1599 VERIFY(0 == sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zp->z_zfsvfs), 1600 &zp->z_size, sizeof (zp->z_size), tx)); 1601 1602 vnode_pager_setsize(ZTOV(zp), end); 1603 1604 zfs_range_unlock(rl); 1605 1606 dmu_tx_commit(tx); 1607 1608 return (0); 1609} 1610 1611/* 1612 * Free space in a file. 1613 * 1614 * IN: zp - znode of file to free data in. 1615 * off - start of section to free. 1616 * len - length of section to free. 1617 * 1618 * RETURN: 0 if success 1619 * error code if failure 1620 */ 1621static int 1622zfs_free_range(znode_t *zp, uint64_t off, uint64_t len) 1623{ 1624 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1625 rl_t *rl; 1626 int error; 1627 1628 /* 1629 * Lock the range being freed. 1630 */ 1631 rl = zfs_range_lock(zp, off, len, RL_WRITER); 1632 1633 /* 1634 * Nothing to do if file already at desired length. 1635 */ 1636 if (off >= zp->z_size) { 1637 zfs_range_unlock(rl); 1638 return (0); 1639 } 1640 1641 if (off + len > zp->z_size) 1642 len = zp->z_size - off; 1643 1644 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, off, len); 1645 1646 if (error == 0) { 1647 /* 1648 * In FreeBSD we cannot free block in the middle of a file, 1649 * but only at the end of a file, so this code path should 1650 * never happen. 1651 */ 1652 vnode_pager_setsize(ZTOV(zp), off); 1653 } 1654 1655 zfs_range_unlock(rl); 1656 1657 return (error); 1658} 1659 1660/* 1661 * Truncate a file 1662 * 1663 * IN: zp - znode of file to free data in. 1664 * end - new end-of-file. 1665 * 1666 * RETURN: 0 if success 1667 * error code if failure 1668 */ 1669static int 1670zfs_trunc(znode_t *zp, uint64_t end) 1671{ 1672 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1673 vnode_t *vp = ZTOV(zp); 1674 dmu_tx_t *tx; 1675 rl_t *rl; 1676 int error; 1677 sa_bulk_attr_t bulk[2]; 1678 int count = 0; 1679 1680 /* 1681 * We will change zp_size, lock the whole file. 1682 */ 1683 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER); 1684 1685 /* 1686 * Nothing to do if file already at desired length. 1687 */ 1688 if (end >= zp->z_size) { 1689 zfs_range_unlock(rl); 1690 return (0); 1691 } 1692 1693 error = dmu_free_long_range(zfsvfs->z_os, zp->z_id, end, -1); 1694 if (error) { 1695 zfs_range_unlock(rl); 1696 return (error); 1697 } 1698top: 1699 tx = dmu_tx_create(zfsvfs->z_os); 1700 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); 1701 zfs_sa_upgrade_txholds(tx, zp); 1702 error = dmu_tx_assign(tx, TXG_NOWAIT); 1703 if (error) { 1704 if (error == ERESTART) { 1705 dmu_tx_wait(tx); 1706 dmu_tx_abort(tx); 1707 goto top; 1708 } 1709 dmu_tx_abort(tx); 1710 zfs_range_unlock(rl); 1711 return (error); 1712 } 1713 1714 zp->z_size = end; 1715 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), 1716 NULL, &zp->z_size, sizeof (zp->z_size)); 1717 1718 if (end == 0) { 1719 zp->z_pflags &= ~ZFS_SPARSE; 1720 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), 1721 NULL, &zp->z_pflags, 8); 1722 } 1723 VERIFY(sa_bulk_update(zp->z_sa_hdl, bulk, count, tx) == 0); 1724 1725 dmu_tx_commit(tx); 1726 1727 /* 1728 * Clear any mapped pages in the truncated region. This has to 1729 * happen outside of the transaction to avoid the possibility of 1730 * a deadlock with someone trying to push a page that we are 1731 * about to invalidate. 1732 */ 1733 vnode_pager_setsize(vp, end); 1734 1735 zfs_range_unlock(rl); 1736 1737 return (0); 1738} 1739 1740/* 1741 * Free space in a file 1742 * 1743 * IN: zp - znode of file to free data in. 1744 * off - start of range 1745 * len - end of range (0 => EOF) 1746 * flag - current file open mode flags. 1747 * log - TRUE if this action should be logged 1748 * 1749 * RETURN: 0 if success 1750 * error code if failure 1751 */ 1752int 1753zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log) 1754{ 1755 vnode_t *vp = ZTOV(zp); 1756 dmu_tx_t *tx; 1757 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1758 zilog_t *zilog = zfsvfs->z_log; 1759 uint64_t mode; 1760 uint64_t mtime[2], ctime[2]; 1761 sa_bulk_attr_t bulk[3]; 1762 int count = 0; 1763 int error; 1764 1765 if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), &mode, 1766 sizeof (mode))) != 0) 1767 return (error); 1768 1769 if (off > zp->z_size) { 1770 error = zfs_extend(zp, off+len); 1771 if (error == 0 && log) 1772 goto log; 1773 else 1774 return (error); 1775 } 1776 1777 /* 1778 * Check for any locks in the region to be freed. 1779 */ 1780 1781 if (MANDLOCK(vp, (mode_t)mode)) { 1782 uint64_t length = (len ? len : zp->z_size - off); 1783 if (error = chklock(vp, FWRITE, off, length, flag, NULL)) 1784 return (error); 1785 } 1786 1787 if (len == 0) { 1788 error = zfs_trunc(zp, off); 1789 } else { 1790 if ((error = zfs_free_range(zp, off, len)) == 0 && 1791 off + len > zp->z_size) 1792 error = zfs_extend(zp, off+len); 1793 } 1794 if (error || !log) 1795 return (error); 1796log: 1797 tx = dmu_tx_create(zfsvfs->z_os); 1798 dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); 1799 zfs_sa_upgrade_txholds(tx, zp); 1800 error = dmu_tx_assign(tx, TXG_NOWAIT); 1801 if (error) { 1802 if (error == ERESTART) { 1803 dmu_tx_wait(tx); 1804 dmu_tx_abort(tx); 1805 goto log; 1806 } 1807 dmu_tx_abort(tx); 1808 return (error); 1809 } 1810 1811 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, mtime, 16); 1812 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, ctime, 16); 1813 SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), 1814 NULL, &zp->z_pflags, 8); 1815 zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, B_TRUE); 1816 error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); 1817 ASSERT(error == 0); 1818 1819 zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len); 1820 1821 dmu_tx_commit(tx); 1822 return (0); 1823} 1824 1825void 1826zfs_create_fs(objset_t *os, cred_t *cr, nvlist_t *zplprops, dmu_tx_t *tx) 1827{ 1828 zfsvfs_t zfsvfs; 1829 uint64_t moid, obj, sa_obj, version; 1830 uint64_t sense = ZFS_CASE_SENSITIVE; 1831 uint64_t norm = 0; 1832 nvpair_t *elem; 1833 int error; 1834 int i; 1835 znode_t *rootzp = NULL; 1836 vnode_t vnode; 1837 vattr_t vattr; 1838 znode_t *zp; 1839 zfs_acl_ids_t acl_ids; 1840 1841 /* 1842 * First attempt to create master node. 1843 */ 1844 /* 1845 * In an empty objset, there are no blocks to read and thus 1846 * there can be no i/o errors (which we assert below). 1847 */ 1848 moid = MASTER_NODE_OBJ; 1849 error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE, 1850 DMU_OT_NONE, 0, tx); 1851 ASSERT(error == 0); 1852 1853 /* 1854 * Set starting attributes. 1855 */ 1856 version = zfs_zpl_version_map(spa_version(dmu_objset_spa(os))); 1857 elem = NULL; 1858 while ((elem = nvlist_next_nvpair(zplprops, elem)) != NULL) { 1859 /* For the moment we expect all zpl props to be uint64_ts */ 1860 uint64_t val; 1861 char *name; 1862 1863 ASSERT(nvpair_type(elem) == DATA_TYPE_UINT64); 1864 VERIFY(nvpair_value_uint64(elem, &val) == 0); 1865 name = nvpair_name(elem); 1866 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_VERSION)) == 0) { 1867 if (val < version) 1868 version = val; 1869 } else { 1870 error = zap_update(os, moid, name, 8, 1, &val, tx); 1871 } 1872 ASSERT(error == 0); 1873 if (strcmp(name, zfs_prop_to_name(ZFS_PROP_NORMALIZE)) == 0) 1874 norm = val; 1875 else if (strcmp(name, zfs_prop_to_name(ZFS_PROP_CASE)) == 0) 1876 sense = val; 1877 } 1878 ASSERT(version != 0); 1879 error = zap_update(os, moid, ZPL_VERSION_STR, 8, 1, &version, tx); 1880 1881 /* 1882 * Create zap object used for SA attribute registration 1883 */ 1884 1885 if (version >= ZPL_VERSION_SA) { 1886 sa_obj = zap_create(os, DMU_OT_SA_MASTER_NODE, 1887 DMU_OT_NONE, 0, tx); 1888 error = zap_add(os, moid, ZFS_SA_ATTRS, 8, 1, &sa_obj, tx); 1889 ASSERT(error == 0); 1890 } else { 1891 sa_obj = 0; 1892 } 1893 /* 1894 * Create a delete queue. 1895 */ 1896 obj = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx); 1897 1898 error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &obj, tx); 1899 ASSERT(error == 0); 1900 1901 /* 1902 * Create root znode. Create minimal znode/vnode/zfsvfs 1903 * to allow zfs_mknode to work. 1904 */ 1905 VATTR_NULL(&vattr); 1906 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE; 1907 vattr.va_type = VDIR; 1908 vattr.va_mode = S_IFDIR|0755; 1909 vattr.va_uid = crgetuid(cr); 1910 vattr.va_gid = crgetgid(cr); 1911 1912 bzero(&zfsvfs, sizeof (zfsvfs_t)); 1913 1914 rootzp = kmem_cache_alloc(znode_cache, KM_SLEEP); 1915 zfs_znode_cache_constructor(rootzp, NULL, 0); 1916 ASSERT(!POINTER_IS_VALID(rootzp->z_zfsvfs)); 1917 rootzp->z_moved = 0; 1918 rootzp->z_unlinked = 0; 1919 rootzp->z_atime_dirty = 0; 1920 rootzp->z_is_sa = USE_SA(version, os); 1921 1922 vnode.v_type = VDIR; 1923 vnode.v_data = rootzp; 1924 rootzp->z_vnode = &vnode; 1925 1926 zfsvfs.z_os = os; 1927 zfsvfs.z_parent = &zfsvfs; 1928 zfsvfs.z_version = version; 1929 zfsvfs.z_use_fuids = USE_FUIDS(version, os); 1930 zfsvfs.z_use_sa = USE_SA(version, os); 1931 zfsvfs.z_norm = norm; 1932 1933 error = sa_setup(os, sa_obj, zfs_attr_table, ZPL_END, 1934 &zfsvfs.z_attr_table); 1935 1936 ASSERT(error == 0); 1937 1938 /* 1939 * Fold case on file systems that are always or sometimes case 1940 * insensitive. 1941 */ 1942 if (sense == ZFS_CASE_INSENSITIVE || sense == ZFS_CASE_MIXED) 1943 zfsvfs.z_norm |= U8_TEXTPREP_TOUPPER; 1944 1945 mutex_init(&zfsvfs.z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); 1946 list_create(&zfsvfs.z_all_znodes, sizeof (znode_t), 1947 offsetof(znode_t, z_link_node)); 1948 1949 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) 1950 mutex_init(&zfsvfs.z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL); 1951 1952 rootzp->z_zfsvfs = &zfsvfs; 1953 VERIFY(0 == zfs_acl_ids_create(rootzp, IS_ROOT_NODE, &vattr, 1954 cr, NULL, &acl_ids)); 1955 zfs_mknode(rootzp, &vattr, tx, cr, IS_ROOT_NODE, &zp, &acl_ids); 1956 ASSERT3P(zp, ==, rootzp); 1957 error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &rootzp->z_id, tx); 1958 ASSERT(error == 0); 1959 zfs_acl_ids_free(&acl_ids); 1960 POINTER_INVALIDATE(&rootzp->z_zfsvfs); 1961 1962 sa_handle_destroy(rootzp->z_sa_hdl); 1963 rootzp->z_vnode = NULL; 1964 kmem_cache_free(znode_cache, rootzp); 1965 1966 /* 1967 * Create shares directory 1968 */ 1969 1970 error = zfs_create_share_dir(&zfsvfs, tx); 1971 1972 ASSERT(error == 0); 1973 1974 for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) 1975 mutex_destroy(&zfsvfs.z_hold_mtx[i]); 1976} 1977 1978#endif /* _KERNEL */ 1979 1980static int 1981zfs_sa_setup(objset_t *osp, sa_attr_type_t **sa_table) 1982{ 1983 uint64_t sa_obj = 0; 1984 int error; 1985 1986 error = zap_lookup(osp, MASTER_NODE_OBJ, ZFS_SA_ATTRS, 8, 1, &sa_obj); 1987 if (error != 0 && error != ENOENT) 1988 return (error); 1989 1990 error = sa_setup(osp, sa_obj, zfs_attr_table, ZPL_END, sa_table); 1991 return (error); 1992} 1993 1994static int 1995zfs_grab_sa_handle(objset_t *osp, uint64_t obj, sa_handle_t **hdlp, 1996 dmu_buf_t **db, void *tag) 1997{ 1998 dmu_object_info_t doi; 1999 int error; 2000 2001 if ((error = sa_buf_hold(osp, obj, tag, db)) != 0) 2002 return (error); 2003 2004 dmu_object_info_from_db(*db, &doi); 2005 if ((doi.doi_bonus_type != DMU_OT_SA && 2006 doi.doi_bonus_type != DMU_OT_ZNODE) || 2007 doi.doi_bonus_type == DMU_OT_ZNODE && 2008 doi.doi_bonus_size < sizeof (znode_phys_t)) { 2009 sa_buf_rele(*db, tag); 2010 return (ENOTSUP); 2011 } 2012 2013 error = sa_handle_get(osp, obj, NULL, SA_HDL_PRIVATE, hdlp); 2014 if (error != 0) { 2015 sa_buf_rele(*db, tag); 2016 return (error); 2017 } 2018 2019 return (0); 2020} 2021 2022void 2023zfs_release_sa_handle(sa_handle_t *hdl, dmu_buf_t *db, void *tag) 2024{ 2025 sa_handle_destroy(hdl); 2026 sa_buf_rele(db, tag); 2027} 2028 2029/* 2030 * Given an object number, return its parent object number and whether 2031 * or not the object is an extended attribute directory. 2032 */ 2033static int 2034zfs_obj_to_pobj(objset_t *osp, sa_handle_t *hdl, sa_attr_type_t *sa_table, 2035 uint64_t *pobjp, int *is_xattrdir) 2036{ 2037 uint64_t parent; 2038 uint64_t pflags; 2039 uint64_t mode; 2040 uint64_t parent_mode; 2041 sa_bulk_attr_t bulk[3]; 2042 sa_handle_t *sa_hdl; 2043 dmu_buf_t *sa_db; 2044 int count = 0; 2045 int error; 2046 2047 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_PARENT], NULL, 2048 &parent, sizeof (parent)); 2049 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_FLAGS], NULL, 2050 &pflags, sizeof (pflags)); 2051 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL, 2052 &mode, sizeof (mode)); 2053 2054 if ((error = sa_bulk_lookup(hdl, bulk, count)) != 0) 2055 return (error); 2056 2057 /* 2058 * When a link is removed its parent pointer is not changed and will 2059 * be invalid. There are two cases where a link is removed but the 2060 * file stays around, when it goes to the delete queue and when there 2061 * are additional links. 2062 */ 2063 error = zfs_grab_sa_handle(osp, parent, &sa_hdl, &sa_db, FTAG); 2064 if (error != 0) 2065 return (error); 2066 2067 error = sa_lookup(sa_hdl, ZPL_MODE, &parent_mode, sizeof (parent_mode)); 2068 zfs_release_sa_handle(sa_hdl, sa_db, FTAG); 2069 if (error != 0) 2070 return (error); 2071 2072 *is_xattrdir = ((pflags & ZFS_XATTR) != 0) && S_ISDIR(mode); 2073 2074 /* 2075 * Extended attributes can be applied to files, directories, etc. 2076 * Otherwise the parent must be a directory. 2077 */ 2078 if (!*is_xattrdir && !S_ISDIR(parent_mode)) 2079 return (EINVAL); 2080 2081 *pobjp = parent; 2082 2083 return (0); 2084} 2085 2086/* 2087 * Given an object number, return some zpl level statistics 2088 */ 2089static int 2090zfs_obj_to_stats_impl(sa_handle_t *hdl, sa_attr_type_t *sa_table, 2091 zfs_stat_t *sb) 2092{ 2093 sa_bulk_attr_t bulk[4]; 2094 int count = 0; 2095 2096 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_MODE], NULL, 2097 &sb->zs_mode, sizeof (sb->zs_mode)); 2098 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_GEN], NULL, 2099 &sb->zs_gen, sizeof (sb->zs_gen)); 2100 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_LINKS], NULL, 2101 &sb->zs_links, sizeof (sb->zs_links)); 2102 SA_ADD_BULK_ATTR(bulk, count, sa_table[ZPL_CTIME], NULL, 2103 &sb->zs_ctime, sizeof (sb->zs_ctime)); 2104 2105 return (sa_bulk_lookup(hdl, bulk, count)); 2106} 2107 2108static int 2109zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl, 2110 sa_attr_type_t *sa_table, char *buf, int len) 2111{ 2112 sa_handle_t *sa_hdl; 2113 sa_handle_t *prevhdl = NULL; 2114 dmu_buf_t *prevdb = NULL; 2115 dmu_buf_t *sa_db = NULL; 2116 char *path = buf + len - 1; 2117 int error; 2118 2119 *path = '\0'; 2120 sa_hdl = hdl; 2121 2122 for (;;) { 2123 uint64_t pobj; 2124 char component[MAXNAMELEN + 2]; 2125 size_t complen; 2126 int is_xattrdir; 2127 2128 if (prevdb) 2129 zfs_release_sa_handle(prevhdl, prevdb, FTAG); 2130 2131 if ((error = zfs_obj_to_pobj(osp, sa_hdl, sa_table, &pobj, 2132 &is_xattrdir)) != 0) 2133 break; 2134 2135 if (pobj == obj) { 2136 if (path[0] != '/') 2137 *--path = '/'; 2138 break; 2139 } 2140 2141 component[0] = '/'; 2142 if (is_xattrdir) { 2143 (void) sprintf(component + 1, "<xattrdir>"); 2144 } else { 2145 error = zap_value_search(osp, pobj, obj, 2146 ZFS_DIRENT_OBJ(-1ULL), component + 1); 2147 if (error != 0) 2148 break; 2149 } 2150 2151 complen = strlen(component); 2152 path -= complen; 2153 ASSERT(path >= buf); 2154 bcopy(component, path, complen); 2155 obj = pobj; 2156 2157 if (sa_hdl != hdl) { 2158 prevhdl = sa_hdl; 2159 prevdb = sa_db; 2160 } 2161 error = zfs_grab_sa_handle(osp, obj, &sa_hdl, &sa_db, FTAG); 2162 if (error != 0) { 2163 sa_hdl = prevhdl; 2164 sa_db = prevdb; 2165 break; 2166 } 2167 } 2168 2169 if (sa_hdl != NULL && sa_hdl != hdl) { 2170 ASSERT(sa_db != NULL); 2171 zfs_release_sa_handle(sa_hdl, sa_db, FTAG); 2172 } 2173 2174 if (error == 0) 2175 (void) memmove(buf, path, buf + len - path); 2176 2177 return (error); 2178} 2179 2180int 2181zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len) 2182{ 2183 sa_attr_type_t *sa_table; 2184 sa_handle_t *hdl; 2185 dmu_buf_t *db; 2186 int error; 2187 2188 error = zfs_sa_setup(osp, &sa_table); 2189 if (error != 0) 2190 return (error); 2191 2192 error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG); 2193 if (error != 0) 2194 return (error); 2195 2196 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len); 2197 2198 zfs_release_sa_handle(hdl, db, FTAG); 2199 return (error); 2200} 2201 2202int 2203zfs_obj_to_stats(objset_t *osp, uint64_t obj, zfs_stat_t *sb, 2204 char *buf, int len) 2205{ 2206 char *path = buf + len - 1; 2207 sa_attr_type_t *sa_table; 2208 sa_handle_t *hdl; 2209 dmu_buf_t *db; 2210 int error; 2211 2212 *path = '\0'; 2213 2214 error = zfs_sa_setup(osp, &sa_table); 2215 if (error != 0) 2216 return (error); 2217 2218 error = zfs_grab_sa_handle(osp, obj, &hdl, &db, FTAG); 2219 if (error != 0) 2220 return (error); 2221 2222 error = zfs_obj_to_stats_impl(hdl, sa_table, sb); 2223 if (error != 0) { 2224 zfs_release_sa_handle(hdl, db, FTAG); 2225 return (error); 2226 } 2227 2228 error = zfs_obj_to_path_impl(osp, obj, hdl, sa_table, buf, len); 2229 2230 zfs_release_sa_handle(hdl, db, FTAG); 2231 return (error); 2232} 2233