zfs_znode.c revision 182824
167754Smsmith/* 267754Smsmith * CDDL HEADER START 369450Smsmith * 467754Smsmith * The contents of this file are subject to the terms of the 567754Smsmith * Common Development and Distribution License (the "License"). 667754Smsmith * You may not use this file except in compliance with the License. 7217365Sjkim * 8298714Sjkim * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 970243Smsmith * or http://www.opensolaris.org/os/licensing. 1067754Smsmith * See the License for the specific language governing permissions 11217365Sjkim * and limitations under the License. 12217365Sjkim * 13217365Sjkim * When distributing Covered Code, include this CDDL HEADER in each 14217365Sjkim * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15217365Sjkim * If applicable, add the following below this CDDL HEADER, with the 16217365Sjkim * fields enclosed by brackets "[]" replaced with your own identifying 17217365Sjkim * information: Portions Copyright [yyyy] [name of copyright owner] 18217365Sjkim * 19217365Sjkim * CDDL HEADER END 20217365Sjkim */ 21217365Sjkim/* 22217365Sjkim * Copyright 2007 Sun Microsystems, Inc. All rights reserved. 23217365Sjkim * Use is subject to license terms. 24217365Sjkim */ 2567754Smsmith 26217365Sjkim/* Portions Copyright 2007 Jeremy Teo */ 27217365Sjkim 28217365Sjkim#pragma ident "%Z%%M% %I% %E% SMI" 2967754Smsmith 30217365Sjkim#ifdef _KERNEL 31217365Sjkim#include <sys/types.h> 32217365Sjkim#include <sys/param.h> 33217365Sjkim#include <sys/time.h> 34217365Sjkim#include <sys/systm.h> 35217365Sjkim#include <sys/sysmacros.h> 36217365Sjkim#include <sys/resource.h> 37217365Sjkim#include <sys/mntent.h> 38217365Sjkim#include <sys/vfs.h> 39217365Sjkim#include <sys/vnode.h> 40217365Sjkim#include <sys/file.h> 41217365Sjkim#include <sys/kmem.h> 42217365Sjkim#include <sys/cmn_err.h> 4367754Smsmith#include <sys/errno.h> 44193341Sjkim#include <sys/unistd.h> 45193341Sjkim#include <sys/atomic.h> 46193341Sjkim#include <sys/zfs_dir.h> 47193341Sjkim#include <sys/zfs_acl.h> 4867754Smsmith#include <sys/zfs_ioctl.h> 4977424Smsmith#include <sys/zfs_rlock.h> 5091116Smsmith#include <sys/fs/zfs.h> 5167754Smsmith#endif /* _KERNEL */ 52167802Sjkim 5367754Smsmith#include <sys/dmu.h> 54167802Sjkim#include <sys/refcount.h> 55167802Sjkim#include <sys/stat.h> 56167802Sjkim#include <sys/zap.h> 57167802Sjkim#include <sys/zfs_znode.h> 58167802Sjkim#include <sys/refcount.h> 5977424Smsmith 6067754Smsmith/* Used by fstat(1). */ 6167754SmsmithSYSCTL_INT(_debug_sizeof, OID_AUTO, znode, CTLFLAG_RD, 0, sizeof(znode_t), 6267754Smsmith "sizeof(znode_t)"); 63138287Smarks 6499679Siwasaki/* 6567754Smsmith * Functions needed for userland (ie: libzpool) are not put under 6667754Smsmith * #ifdef_KERNEL; the rest of the functions have dependencies 6767754Smsmith * (such as VFS logic) that will not compile easily in userland. 6867754Smsmith */ 6967754Smsmith#ifdef _KERNEL 70151937Sjkimstruct kmem_cache *znode_cache = NULL; 7167754Smsmith 7277424Smsmith/*ARGSUSED*/ 7367754Smsmithstatic void 7467754Smsmithznode_pageout_func(dmu_buf_t *dbuf, void *user_ptr) 7567754Smsmith{ 7667754Smsmith znode_t *zp = user_ptr; 7767754Smsmith vnode_t *vp; 7867754Smsmith 7967754Smsmith mutex_enter(&zp->z_lock); 8067754Smsmith vp = ZTOV(zp); 8192388Smsmith if (vp == NULL) { 8292388Smsmith mutex_exit(&zp->z_lock); 8392388Smsmith zfs_znode_free(zp); 8499679Siwasaki } else if (vp->v_count == 0) { 85167802Sjkim ZTOV(zp) = NULL; 8667754Smsmith vhold(vp); 8767754Smsmith mutex_exit(&zp->z_lock); 8867754Smsmith vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 8967754Smsmith vrecycle(vp, curthread); 9067754Smsmith VOP_UNLOCK(vp, 0); 9167754Smsmith vdrop(vp); 92167802Sjkim zfs_znode_free(zp); 93167802Sjkim } else { 94167802Sjkim /* signal force unmount that this znode can be freed */ 95167802Sjkim zp->z_dbuf = NULL; 96167802Sjkim mutex_exit(&zp->z_lock); 97167802Sjkim } 98167802Sjkim} 99167802Sjkim 100167802Sjkimextern struct vop_vector zfs_vnodeops; 101167802Sjkimextern struct vop_vector zfs_fifoops; 10267754Smsmith 10367754Smsmith/* 10467754Smsmith * XXX: We cannot use this function as a cache constructor, because 10567754Smsmith * there is one global cache for all file systems and we need 10667754Smsmith * to pass vfsp here, which is not possible, because argument 10799679Siwasaki * 'cdrarg' is defined at kmem_cache_create() time. 10867754Smsmith */ 109167802Sjkimstatic int 11092388Smsmithzfs_znode_cache_constructor(void *buf, void *cdrarg, int kmflags) 11167754Smsmith{ 11267754Smsmith znode_t *zp = buf; 11367754Smsmith vnode_t *vp; 11467754Smsmith vfs_t *vfsp = cdrarg; 11592388Smsmith int error; 11692388Smsmith 11792388Smsmith if (cdrarg != NULL) { 11892388Smsmith error = getnewvnode("zfs", vfsp, &zfs_vnodeops, &vp); 11992388Smsmith ASSERT(error == 0); 12092388Smsmith vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 12167754Smsmith zp->z_vnode = vp; 12267754Smsmith vp->v_data = (caddr_t)zp; 12367754Smsmith VN_LOCK_AREC(vp); 12467754Smsmith VN_LOCK_ASHARE(vp); 12577424Smsmith } else { 12667754Smsmith zp->z_vnode = NULL; 12767754Smsmith } 12867754Smsmith mutex_init(&zp->z_lock, NULL, MUTEX_DEFAULT, NULL); 129138287Smarks rw_init(&zp->z_map_lock, NULL, RW_DEFAULT, NULL); 13099679Siwasaki rw_init(&zp->z_parent_lock, NULL, RW_DEFAULT, NULL); 13167754Smsmith rw_init(&zp->z_name_lock, NULL, RW_DEFAULT, NULL); 13267754Smsmith mutex_init(&zp->z_acl_lock, NULL, MUTEX_DEFAULT, NULL); 13367754Smsmith 13467754Smsmith mutex_init(&zp->z_range_lock, NULL, MUTEX_DEFAULT, NULL); 13567754Smsmith avl_create(&zp->z_range_avl, zfs_range_compare, 136151937Sjkim sizeof (rl_t), offsetof(rl_t, r_node)); 13767754Smsmith 13877424Smsmith zp->z_dbuf_held = 0; 13967754Smsmith zp->z_dirlocks = 0; 14067754Smsmith return (0); 14167754Smsmith} 14267754Smsmith 14367754Smsmith/*ARGSUSED*/ 14467754Smsmithstatic void 14567754Smsmithzfs_znode_cache_destructor(void *buf, void *cdarg) 14667754Smsmith{ 147167802Sjkim znode_t *zp = buf; 14867754Smsmith 14977424Smsmith ASSERT(zp->z_dirlocks == 0); 15067754Smsmith mutex_destroy(&zp->z_lock); 15167754Smsmith rw_destroy(&zp->z_map_lock); 15267754Smsmith rw_destroy(&zp->z_parent_lock); 15367754Smsmith rw_destroy(&zp->z_name_lock); 15467754Smsmith mutex_destroy(&zp->z_acl_lock); 15567754Smsmith mutex_destroy(&zp->z_range_lock); 15667754Smsmith avl_destroy(&zp->z_range_avl); 15767754Smsmith 15867754Smsmith ASSERT(zp->z_dbuf_held == 0); 15967754Smsmith} 16067754Smsmith 16167754Smsmithvoid 16267754Smsmithzfs_znode_init(void) 16377424Smsmith{ 16467754Smsmith /* 16567754Smsmith * Initialize zcache 16667754Smsmith */ 167151937Sjkim ASSERT(znode_cache == NULL); 16899679Siwasaki znode_cache = kmem_cache_create("zfs_znode_cache", 16967754Smsmith sizeof (znode_t), 0, /* zfs_znode_cache_constructor */ NULL, 17067754Smsmith zfs_znode_cache_destructor, NULL, NULL, NULL, 0); 17167754Smsmith} 17267754Smsmith 17367754Smsmithvoid 174151937Sjkimzfs_znode_fini(void) 17567754Smsmith{ 17667754Smsmith /* 17767754Smsmith * Cleanup zcache 17877424Smsmith */ 17967754Smsmith if (znode_cache) 18067754Smsmith kmem_cache_destroy(znode_cache); 18167754Smsmith znode_cache = NULL; 18267754Smsmith} 18367754Smsmith 18467754Smsmith/* 18567754Smsmith * zfs_init_fs - Initialize the zfsvfs struct and the file system 18667754Smsmith * incore "master" object. Verify version compatibility. 18767754Smsmith */ 188202771Sjkimint 18980062Smsmithzfs_init_fs(zfsvfs_t *zfsvfs, znode_t **zpp, cred_t *cr) 19067754Smsmith{ 191117521Snjl objset_t *os = zfsvfs->z_os; 192117521Snjl uint64_t version = ZPL_VERSION; 193167802Sjkim int i, error; 19467754Smsmith dmu_object_info_t doi; 19567754Smsmith uint64_t fsid_guid; 19677424Smsmith 197167802Sjkim *zpp = NULL; 19867754Smsmith 19977424Smsmith /* 200123315Snjl * XXX - hack to auto-create the pool root filesystem at 20167754Smsmith * the first attempted mount. 20267754Smsmith */ 20367754Smsmith if (dmu_object_info(os, MASTER_NODE_OBJ, &doi) == ENOENT) { 20499679Siwasaki dmu_tx_t *tx = dmu_tx_create(os); 20599679Siwasaki 20667754Smsmith dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, TRUE, NULL); /* master */ 20782367Smsmith dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, TRUE, NULL); /* del queue */ 20885756Smsmith dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT); /* root node */ 20977424Smsmith error = dmu_tx_assign(tx, TXG_WAIT); 21067754Smsmith ASSERT3U(error, ==, 0); 21167754Smsmith zfs_create_fs(os, cr, tx); 212117521Snjl dmu_tx_commit(tx); 21367754Smsmith } 21467754Smsmith 21580062Smsmith error = zap_lookup(os, MASTER_NODE_OBJ, ZPL_VERSION_OBJ, 8, 1, 21667754Smsmith &version); 217167802Sjkim if (error) { 21867754Smsmith return (error); 21967754Smsmith } else if (version != ZPL_VERSION) { 22067754Smsmith (void) printf("Mismatched versions: File system " 22167754Smsmith "is version %lld on-disk format, which is " 222209746Sjkim "incompatible with this software version %lld!", 22367754Smsmith (u_longlong_t)version, ZPL_VERSION); 22467754Smsmith return (ENOTSUP); 22599679Siwasaki } 22699679Siwasaki 22767754Smsmith /* 22899679Siwasaki * The fsid is 64 bits, composed of an 8-bit fs type, which 22999679Siwasaki * separates our fsid from any other filesystem types, and a 23067754Smsmith * 56-bit objset unique ID. The objset unique ID is unique to 23167754Smsmith * all objsets open on this system, provided by unique_create(). 23267754Smsmith * The 8-bit fs type must be put in the low bits of fsid[1] 233117521Snjl * because that's where other Solaris filesystems put it. 23499679Siwasaki */ 23599679Siwasaki fsid_guid = dmu_objset_fsid_guid(os); 23667754Smsmith ASSERT((fsid_guid & ~((1ULL<<56)-1)) == 0); 237117521Snjl zfsvfs->z_vfs->vfs_fsid.val[0] = fsid_guid; 23867754Smsmith zfsvfs->z_vfs->vfs_fsid.val[1] = ((fsid_guid>>32) << 8) | 239117521Snjl zfsvfs->z_vfs->mnt_vfc->vfc_typenum & 0xFF; 240117521Snjl 241117521Snjl error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_ROOT_OBJ, 8, 1, 242117521Snjl &zfsvfs->z_root); 24367754Smsmith if (error) 244167802Sjkim return (error); 245167802Sjkim ASSERT(zfsvfs->z_root != 0); 246167802Sjkim 24767754Smsmith /* 248167802Sjkim * Create the per mount vop tables. 249167802Sjkim */ 250167802Sjkim 251298714Sjkim /* 252298714Sjkim * Initialize zget mutex's 253298714Sjkim */ 254167802Sjkim for (i = 0; i != ZFS_OBJ_MTX_SZ; i++) 25567754Smsmith mutex_init(&zfsvfs->z_hold_mtx[i], NULL, MUTEX_DEFAULT, NULL); 256167802Sjkim 25799679Siwasaki error = zfs_zget(zfsvfs, zfsvfs->z_root, zpp); 258167802Sjkim if (error) 259193267Sjkim return (error); 260193267Sjkim ASSERT3U((*zpp)->z_id, ==, zfsvfs->z_root); 261193267Sjkim 262167802Sjkim error = zap_lookup(os, MASTER_NODE_OBJ, ZFS_UNLINKED_SET, 8, 1, 263167802Sjkim &zfsvfs->z_unlinkedobj); 26499679Siwasaki if (error) 265167802Sjkim return (error); 266167802Sjkim 267167802Sjkim return (0); 268193267Sjkim} 269193267Sjkim 270167802Sjkim/* 271167802Sjkim * define a couple of values we need available 27267754Smsmith * for both 64 and 32 bit environments. 273167802Sjkim */ 27467754Smsmith#ifndef NBITSMINOR64 27567754Smsmith#define NBITSMINOR64 32 276209746Sjkim#endif 27767754Smsmith#ifndef MAXMAJ64 278117521Snjl#define MAXMAJ64 0xffffffffUL 279117521Snjl#endif 28067754Smsmith#ifndef MAXMIN64 28167754Smsmith#define MAXMIN64 0xffffffffUL 28267754Smsmith#endif 283117521Snjl 28467754Smsmith/* 28569450Smsmith * Create special expldev for ZFS private use. 28680062Smsmith * Can't use standard expldev since it doesn't do 287117521Snjl * what we want. The standard expldev() takes a 288117521Snjl * dev32_t in LP64 and expands it to a long dev_t. 28980062Smsmith * We need an interface that takes a dev32_t in ILP32 290117521Snjl * and expands it to a long dev_t. 29167754Smsmith */ 292117521Snjlstatic uint64_t 29367754Smsmithzfs_expldev(dev_t dev) 29467754Smsmith{ 295117521Snjl return (((uint64_t)umajor(dev) << NBITSMINOR64) | uminor(dev)); 296117521Snjl} 297167802Sjkim/* 298117521Snjl * Special cmpldev for ZFS private use. 299117521Snjl * Can't use standard cmpldev since it takes 300117521Snjl * a long dev_t and compresses it to dev32_t in 301117521Snjl * LP64. We need to do a compaction of a long dev_t 302117521Snjl * to a dev32_t in ILP32. 30380062Smsmith */ 304193267Sjkimdev_t 305193267Sjkimzfs_cmpldev(uint64_t dev) 306167802Sjkim{ 307167802Sjkim return (makedev((dev >> NBITSMINOR64), (dev & MAXMIN64))); 308167802Sjkim} 30980062Smsmith 310167802Sjkim/* 311167802Sjkim * Construct a new znode/vnode and intialize. 312167802Sjkim * 313209746Sjkim * This does not do a call to dmu_set_user() that is 314167802Sjkim * up to the caller to do, in case you don't want to 315117521Snjl * return the znode 316167802Sjkim */ 317167802Sjkimstatic znode_t * 318193267Sjkimzfs_znode_alloc(zfsvfs_t *zfsvfs, dmu_buf_t *db, uint64_t obj_num, int blksz) 319167802Sjkim{ 320167802Sjkim znode_t *zp; 321167802Sjkim vnode_t *vp; 322117521Snjl int error; 323212761Sjkim 324212761Sjkim zp = kmem_cache_alloc(znode_cache, KM_SLEEP); 325117521Snjl zfs_znode_cache_constructor(zp, zfsvfs->z_vfs, 0); 326193267Sjkim 327298714Sjkim ASSERT(zp->z_dirlocks == NULL); 328117521Snjl 329117521Snjl zp->z_phys = db->db_data; 330193267Sjkim zp->z_zfsvfs = zfsvfs; 331193267Sjkim zp->z_unlinked = 0; 332117521Snjl zp->z_atime_dirty = 0; 33367754Smsmith zp->z_dbuf_held = 0; 33467754Smsmith zp->z_mapcnt = 0; 335117521Snjl zp->z_last_itx = 0; 336117521Snjl zp->z_dbuf = db; 33767754Smsmith zp->z_id = obj_num; 33867754Smsmith zp->z_blksz = blksz; 339117521Snjl zp->z_seq = 0x7A4653; 340114237Snjl zp->z_sync_cnt = 0; 341193267Sjkim 342298714Sjkim mutex_enter(&zfsvfs->z_znodes_lock); 343117521Snjl list_insert_tail(&zfsvfs->z_all_znodes, zp); 344117521Snjl mutex_exit(&zfsvfs->z_znodes_lock); 345117521Snjl 346117521Snjl vp = ZTOV(zp); 347117521Snjl if (vp == NULL) 348117521Snjl return (zp); 349117521Snjl 350193267Sjkim vp->v_vflag |= VV_FORCEINSMQ; 351298714Sjkim error = insmntque(vp, zfsvfs->z_vfs); 352117521Snjl vp->v_vflag &= ~VV_FORCEINSMQ; 353117521Snjl KASSERT(error == 0, ("insmntque() failed: error %d", error)); 354117521Snjl 355117521Snjl vp->v_type = IFTOVT((mode_t)zp->z_phys->zp_mode); 356117521Snjl switch (vp->v_type) { 357212761Sjkim case VDIR: 358117521Snjl zp->z_zn_prefetch = B_TRUE; /* z_prefetch default is enabled */ 359212761Sjkim break; 360212761Sjkim case VFIFO: 361212761Sjkim vp->v_op = &zfs_fifoops; 362212761Sjkim break; 363212761Sjkim } 364212761Sjkim 365117521Snjl return (zp); 36680062Smsmith} 36767754Smsmith 36867754Smsmithstatic void 36967754Smsmithzfs_znode_dmu_init(znode_t *zp) 37067754Smsmith{ 37177424Smsmith znode_t *nzp; 37267754Smsmith zfsvfs_t *zfsvfs = zp->z_zfsvfs; 373167802Sjkim dmu_buf_t *db = zp->z_dbuf; 374167802Sjkim 375167802Sjkim mutex_enter(&zp->z_lock); 376167802Sjkim 377167802Sjkim nzp = dmu_buf_set_user_ie(db, zp, &zp->z_phys, znode_pageout_func); 378167802Sjkim 379167802Sjkim /* 380167802Sjkim * there should be no 381167802Sjkim * concurrent zgets on this object. 382167802Sjkim */ 383167802Sjkim ASSERT3P(nzp, ==, NULL); 384167802Sjkim 385167802Sjkim /* 386167802Sjkim * Slap on VROOT if we are the root znode 387167802Sjkim */ 388167802Sjkim if (zp->z_id == zfsvfs->z_root) { 389241973Sjkim ZTOV(zp)->v_flag |= VROOT; 390241973Sjkim } 391193267Sjkim 392197104Sjkim ASSERT(zp->z_dbuf_held == 0); 393167802Sjkim zp->z_dbuf_held = 1; 394167802Sjkim VFS_HOLD(zfsvfs->z_vfs); 395193267Sjkim mutex_exit(&zp->z_lock); 396193267Sjkim} 397167802Sjkim 398167802Sjkim/* 399167802Sjkim * Create a new DMU object to hold a zfs znode. 400167802Sjkim * 401167802Sjkim * IN: dzp - parent directory for new znode 402167802Sjkim * vap - file attributes for new znode 403197104Sjkim * tx - dmu transaction id for zap operations 404197104Sjkim * cr - credentials of caller 405197104Sjkim * flag - flags: 406197104Sjkim * IS_ROOT_NODE - new object will be root 407167802Sjkim * IS_XATTR - new object is an attribute 408167802Sjkim * IS_REPLAY - intent log replay 409167802Sjkim * 410167802Sjkim * OUT: oid - ID of created object 411193267Sjkim * 412193267Sjkim */ 413167802Sjkimvoid 414167802Sjkimzfs_mknode(znode_t *dzp, vattr_t *vap, uint64_t *oid, dmu_tx_t *tx, cred_t *cr, 415167802Sjkim uint_t flag, znode_t **zpp, int bonuslen) 416167802Sjkim{ 417167802Sjkim dmu_buf_t *dbp; 418167802Sjkim znode_phys_t *pzp; 419167802Sjkim znode_t *zp; 420167802Sjkim zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 421167802Sjkim timestruc_t now; 422167802Sjkim uint64_t gen; 423197104Sjkim int err; 424167802Sjkim 425167802Sjkim ASSERT(vap && (vap->va_mask & (AT_TYPE|AT_MODE)) == (AT_TYPE|AT_MODE)); 426167802Sjkim 427167802Sjkim if (zfsvfs->z_assign >= TXG_INITIAL) { /* ZIL replay */ 428167802Sjkim *oid = vap->va_nodeid; 429167802Sjkim flag |= IS_REPLAY; 430167802Sjkim now = vap->va_ctime; /* see zfs_replay_create() */ 431167802Sjkim gen = vap->va_nblocks; /* ditto */ 432167802Sjkim } else { 433167802Sjkim *oid = 0; 434167802Sjkim gethrestime(&now); 435167802Sjkim gen = dmu_tx_get_txg(tx); 436167802Sjkim } 43784491Smsmith 43884491Smsmith /* 439138287Smarks * Create a new DMU object. 44099679Siwasaki */ 44184491Smsmith /* 44284491Smsmith * There's currently no mechanism for pre-reading the blocks that will 44384491Smsmith * be to needed allocate a new object, so we accept the small chance 44484491Smsmith * that there will be an i/o error and we will fail one of the 44584491Smsmith * assertions below. 446151937Sjkim */ 44784491Smsmith if (vap->va_type == VDIR) { 44884491Smsmith if (flag & IS_REPLAY) { 44984491Smsmith err = zap_create_claim(zfsvfs->z_os, *oid, 45084491Smsmith DMU_OT_DIRECTORY_CONTENTS, 45184491Smsmith DMU_OT_ZNODE, sizeof (znode_phys_t) + bonuslen, tx); 45284491Smsmith ASSERT3U(err, ==, 0); 45384491Smsmith } else { 45484491Smsmith *oid = zap_create(zfsvfs->z_os, 45584491Smsmith DMU_OT_DIRECTORY_CONTENTS, 45684491Smsmith DMU_OT_ZNODE, sizeof (znode_phys_t) + bonuslen, tx); 45784491Smsmith } 45884491Smsmith } else { 459167802Sjkim if (flag & IS_REPLAY) { 46084491Smsmith err = dmu_object_claim(zfsvfs->z_os, *oid, 46184491Smsmith DMU_OT_PLAIN_FILE_CONTENTS, 0, 46284491Smsmith DMU_OT_ZNODE, sizeof (znode_phys_t) + bonuslen, tx); 46384491Smsmith ASSERT3U(err, ==, 0); 46484491Smsmith } else { 46584491Smsmith *oid = dmu_object_alloc(zfsvfs->z_os, 46684491Smsmith DMU_OT_PLAIN_FILE_CONTENTS, 0, 46784491Smsmith DMU_OT_ZNODE, sizeof (znode_phys_t) + bonuslen, tx); 46884491Smsmith } 46984491Smsmith } 470138287Smarks VERIFY(0 == dmu_bonus_hold(zfsvfs->z_os, *oid, NULL, &dbp)); 47199679Siwasaki dmu_buf_will_dirty(dbp, tx); 47284491Smsmith 47384491Smsmith /* 47484491Smsmith * Initialize the znode physical data to zero. 47584491Smsmith */ 47684491Smsmith ASSERT(dbp->db_size >= sizeof (znode_phys_t)); 477151937Sjkim bzero(dbp->db_data, dbp->db_size); 47884491Smsmith pzp = dbp->db_data; 47984491Smsmith 48084491Smsmith /* 48184491Smsmith * If this is the root, fix up the half-initialized parent pointer 48284491Smsmith * to reference the just-allocated physical data area. 48384491Smsmith */ 48484491Smsmith if (flag & IS_ROOT_NODE) { 48584491Smsmith dzp->z_phys = pzp; 48684491Smsmith dzp->z_id = *oid; 48784491Smsmith } 48884491Smsmith 48984491Smsmith /* 490167802Sjkim * If parent is an xattr, so am I. 49184491Smsmith */ 49284491Smsmith if (dzp->z_phys->zp_flags & ZFS_XATTR) 49384491Smsmith flag |= IS_XATTR; 49484491Smsmith 49584491Smsmith if (vap->va_type == VBLK || vap->va_type == VCHR) { 49684491Smsmith pzp->zp_rdev = zfs_expldev(vap->va_rdev); 49784491Smsmith } 49884491Smsmith 49967754Smsmith if (vap->va_type == VDIR) { 50067754Smsmith pzp->zp_size = 2; /* contents ("." and "..") */ 501138287Smarks pzp->zp_links = (flag & (IS_ROOT_NODE | IS_XATTR)) ? 2 : 1; 50299679Siwasaki } 50367754Smsmith 50467754Smsmith pzp->zp_parent = dzp->z_id; 50567754Smsmith if (flag & IS_XATTR) 50667754Smsmith pzp->zp_flags |= ZFS_XATTR; 50767754Smsmith 508151937Sjkim pzp->zp_gen = gen; 50967754Smsmith 51077424Smsmith ZFS_TIME_ENCODE(&now, pzp->zp_crtime); 51167754Smsmith ZFS_TIME_ENCODE(&now, pzp->zp_ctime); 51267754Smsmith 51367754Smsmith if (vap->va_mask & AT_ATIME) { 51467754Smsmith ZFS_TIME_ENCODE(&vap->va_atime, pzp->zp_atime); 51567754Smsmith } else { 51667754Smsmith ZFS_TIME_ENCODE(&now, pzp->zp_atime); 51767754Smsmith } 51867754Smsmith 519167802Sjkim if (vap->va_mask & AT_MTIME) { 52067754Smsmith ZFS_TIME_ENCODE(&vap->va_mtime, pzp->zp_mtime); 52177424Smsmith } else { 52267754Smsmith ZFS_TIME_ENCODE(&now, pzp->zp_mtime); 52367754Smsmith } 52467754Smsmith 52567754Smsmith pzp->zp_mode = MAKEIMODE(vap->va_type, vap->va_mode); 52667754Smsmith zp = zfs_znode_alloc(zfsvfs, dbp, *oid, 0); 52767754Smsmith 52867754Smsmith zfs_perm_init(zp, dzp, flag, vap, tx, cr); 52967754Smsmith 53067754Smsmith if (zpp) { 53167754Smsmith kmutex_t *hash_mtx = ZFS_OBJ_MUTEX(zp); 53267754Smsmith 53367754Smsmith mutex_enter(hash_mtx); 53467754Smsmith zfs_znode_dmu_init(zp); 53577424Smsmith mutex_exit(hash_mtx); 53667754Smsmith 53767754Smsmith *zpp = zp; 53867754Smsmith } else { 53999679Siwasaki if (ZTOV(zp) != NULL) { 54099679Siwasaki ZTOV(zp)->v_count = 0; 54167754Smsmith VOP_UNLOCK(ZTOV(zp), 0); 54267754Smsmith } 54367754Smsmith dmu_buf_rele(dbp, NULL); 54467754Smsmith zfs_znode_free(zp); 54567754Smsmith } 54667754Smsmith} 54767754Smsmith 54867754Smsmithint 54967754Smsmithzfs_zget(zfsvfs_t *zfsvfs, uint64_t obj_num, znode_t **zpp) 550193267Sjkim{ 55167754Smsmith dmu_object_info_t doi; 552193267Sjkim dmu_buf_t *db; 55367754Smsmith znode_t *zp; 55467754Smsmith vnode_t *vp; 555167802Sjkim int err; 556167802Sjkim 557167802Sjkim *zpp = NULL; 55867754Smsmith 55967754Smsmith ZFS_OBJ_HOLD_ENTER(zfsvfs, obj_num); 56067754Smsmith 56167754Smsmith err = dmu_bonus_hold(zfsvfs->z_os, obj_num, NULL, &db); 56267754Smsmith if (err) { 56367754Smsmith ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 56467754Smsmith return (err); 56577424Smsmith } 56677424Smsmith 56777424Smsmith dmu_object_info_from_db(db, &doi); 56877424Smsmith if (doi.doi_bonus_type != DMU_OT_ZNODE || 56967754Smsmith doi.doi_bonus_size < sizeof (znode_phys_t)) { 57067754Smsmith dmu_buf_rele(db, NULL); 57167754Smsmith ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 572167802Sjkim return (EINVAL); 57367754Smsmith } 57467754Smsmith 57567754Smsmith ASSERT(db->db_object == obj_num); 57667754Smsmith ASSERT(db->db_offset == -1); 57767754Smsmith ASSERT(db->db_data != NULL); 57867754Smsmith 57967754Smsmith zp = dmu_buf_get_user(db); 58087031Smsmith 58187031Smsmith if (zp != NULL) { 58287031Smsmith mutex_enter(&zp->z_lock); 58387031Smsmith 58487031Smsmith ASSERT3U(zp->z_id, ==, obj_num); 585298714Sjkim if (zp->z_unlinked) { 58699679Siwasaki dmu_buf_rele(db, NULL); 587209746Sjkim mutex_exit(&zp->z_lock); 58867754Smsmith ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 58967754Smsmith return (ENOENT); 59067754Smsmith } else if (zp->z_dbuf_held) { 59199679Siwasaki dmu_buf_rele(db, NULL); 592298714Sjkim } else { 59367754Smsmith zp->z_dbuf_held = 1; 59467754Smsmith VFS_HOLD(zfsvfs->z_vfs); 59567754Smsmith } 596117521Snjl 597117521Snjl if (ZTOV(zp) != NULL) 59867754Smsmith VN_HOLD(ZTOV(zp)); 59977424Smsmith else { 60067754Smsmith err = getnewvnode("zfs", zfsvfs->z_vfs, &zfs_vnodeops, 60167754Smsmith &zp->z_vnode); 602117521Snjl ASSERT(err == 0); 603117521Snjl vp = ZTOV(zp); 60467754Smsmith vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 60567754Smsmith vp->v_data = (caddr_t)zp; 60667754Smsmith VN_LOCK_AREC(vp); 60767754Smsmith VN_LOCK_ASHARE(vp); 60867754Smsmith vp->v_type = IFTOVT((mode_t)zp->z_phys->zp_mode); 60967754Smsmith if (vp->v_type == VDIR) 610298714Sjkim zp->z_zn_prefetch = B_TRUE; /* z_prefetch default is enabled */ 61167754Smsmith vp->v_vflag |= VV_FORCEINSMQ; 61299679Siwasaki err = insmntque(vp, zfsvfs->z_vfs); 613200553Sjkim vp->v_vflag &= ~VV_FORCEINSMQ; 614200553Sjkim KASSERT(err == 0, ("insmntque() failed: error %d", err)); 615200553Sjkim VOP_UNLOCK(vp, 0); 616200553Sjkim } 617200553Sjkim mutex_exit(&zp->z_lock); 618200553Sjkim ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 619200553Sjkim *zpp = zp; 620200553Sjkim return (0); 621217365Sjkim } 622200553Sjkim 623217365Sjkim /* 624200553Sjkim * Not found create new znode/vnode 625200553Sjkim */ 626200553Sjkim zp = zfs_znode_alloc(zfsvfs, db, obj_num, doi.doi_data_block_size); 62799679Siwasaki ASSERT3U(zp->z_id, ==, obj_num); 628250838Sjkim zfs_znode_dmu_init(zp); 62999679Siwasaki ZFS_OBJ_HOLD_EXIT(zfsvfs, obj_num); 630250838Sjkim *zpp = zp; 63199679Siwasaki VOP_UNLOCK(vp, 0); 63267754Smsmith return (0); 63367754Smsmith} 634298714Sjkim 635298714Sjkimvoid 63667754Smsmithzfs_znode_delete(znode_t *zp, dmu_tx_t *tx) 637298714Sjkim{ 63899679Siwasaki zfsvfs_t *zfsvfs = zp->z_zfsvfs; 639298714Sjkim int error; 640298714Sjkim 641298714Sjkim ZFS_OBJ_HOLD_ENTER(zfsvfs, zp->z_id); 64299679Siwasaki if (zp->z_phys->zp_acl.z_acl_extern_obj) { 643298714Sjkim error = dmu_object_free(zfsvfs->z_os, 644298714Sjkim zp->z_phys->zp_acl.z_acl_extern_obj, tx); 64567754Smsmith ASSERT3U(error, ==, 0); 646298714Sjkim } 647298714Sjkim error = dmu_object_free(zfsvfs->z_os, zp->z_id, tx); 648298714Sjkim ASSERT3U(error, ==, 0); 649298714Sjkim zp->z_dbuf_held = 0; 650298714Sjkim ZFS_OBJ_HOLD_EXIT(zfsvfs, zp->z_id); 651298714Sjkim dmu_buf_rele(zp->z_dbuf, NULL); 652298714Sjkim} 653298714Sjkim 654123315Snjlvoid 655298714Sjkimzfs_zinactive(znode_t *zp) 656123315Snjl{ 657298714Sjkim vnode_t *vp = ZTOV(zp); 658123315Snjl zfsvfs_t *zfsvfs = zp->z_zfsvfs; 659298714Sjkim uint64_t z_id = zp->z_id; 660123315Snjl 661298714Sjkim ASSERT(zp->z_dbuf_held && zp->z_phys); 662298714Sjkim 663298714Sjkim /* 664298714Sjkim * Don't allow a zfs_zget() while were trying to release this znode 665123315Snjl */ 666298714Sjkim ZFS_OBJ_HOLD_ENTER(zfsvfs, z_id); 667123315Snjl 66867754Smsmith mutex_enter(&zp->z_lock); 66967754Smsmith VI_LOCK(vp); 670298714Sjkim if (vp->v_count > 0) { 67199679Siwasaki /* 67267754Smsmith * If the hold count is greater than zero, somebody has 67367754Smsmith * obtained a new reference on this znode while we were 674193267Sjkim * processing it here, so we are done. 675193267Sjkim */ 676209746Sjkim VI_UNLOCK(vp); 67799679Siwasaki mutex_exit(&zp->z_lock); 67867754Smsmith ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id); 679117521Snjl return; 680117521Snjl } 68182367Smsmith VI_UNLOCK(vp); 68282367Smsmith 68377424Smsmith /* 68467754Smsmith * If this was the last reference to a file with no links, 68567754Smsmith * remove the file from the file system. 68667754Smsmith */ 687 if (zp->z_unlinked) { 688 ZTOV(zp) = NULL; 689 mutex_exit(&zp->z_lock); 690 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id); 691 ASSERT(vp->v_count == 0); 692 vrecycle(vp, curthread); 693 zfs_rmnode(zp); 694 VFS_RELE(zfsvfs->z_vfs); 695 return; 696 } 697 ASSERT(zp->z_phys); 698 ASSERT(zp->z_dbuf_held); 699 mutex_exit(&zp->z_lock); 700 ZFS_OBJ_HOLD_EXIT(zfsvfs, z_id); 701} 702 703void 704zfs_znode_free(znode_t *zp) 705{ 706 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 707 708 mutex_enter(&zfsvfs->z_znodes_lock); 709 list_remove(&zfsvfs->z_all_znodes, zp); 710 mutex_exit(&zfsvfs->z_znodes_lock); 711 712 kmem_cache_free(znode_cache, zp); 713} 714 715void 716zfs_time_stamper_locked(znode_t *zp, uint_t flag, dmu_tx_t *tx) 717{ 718 timestruc_t now; 719 720 ASSERT(MUTEX_HELD(&zp->z_lock)); 721 722 gethrestime(&now); 723 724 if (tx) { 725 dmu_buf_will_dirty(zp->z_dbuf, tx); 726 zp->z_atime_dirty = 0; 727 zp->z_seq++; 728 } else { 729 zp->z_atime_dirty = 1; 730 } 731 732 if (flag & AT_ATIME) 733 ZFS_TIME_ENCODE(&now, zp->z_phys->zp_atime); 734 735 if (flag & AT_MTIME) 736 ZFS_TIME_ENCODE(&now, zp->z_phys->zp_mtime); 737 738 if (flag & AT_CTIME) 739 ZFS_TIME_ENCODE(&now, zp->z_phys->zp_ctime); 740} 741 742/* 743 * Update the requested znode timestamps with the current time. 744 * If we are in a transaction, then go ahead and mark the znode 745 * dirty in the transaction so the timestamps will go to disk. 746 * Otherwise, we will get pushed next time the znode is updated 747 * in a transaction, or when this znode eventually goes inactive. 748 * 749 * Why is this OK? 750 * 1 - Only the ACCESS time is ever updated outside of a transaction. 751 * 2 - Multiple consecutive updates will be collapsed into a single 752 * znode update by the transaction grouping semantics of the DMU. 753 */ 754void 755zfs_time_stamper(znode_t *zp, uint_t flag, dmu_tx_t *tx) 756{ 757 mutex_enter(&zp->z_lock); 758 zfs_time_stamper_locked(zp, flag, tx); 759 mutex_exit(&zp->z_lock); 760} 761 762/* 763 * Grow the block size for a file. 764 * 765 * IN: zp - znode of file to free data in. 766 * size - requested block size 767 * tx - open transaction. 768 * 769 * NOTE: this function assumes that the znode is write locked. 770 */ 771void 772zfs_grow_blocksize(znode_t *zp, uint64_t size, dmu_tx_t *tx) 773{ 774 int error; 775 u_longlong_t dummy; 776 777 if (size <= zp->z_blksz) 778 return; 779 /* 780 * If the file size is already greater than the current blocksize, 781 * we will not grow. If there is more than one block in a file, 782 * the blocksize cannot change. 783 */ 784 if (zp->z_blksz && zp->z_phys->zp_size > zp->z_blksz) 785 return; 786 787 error = dmu_object_set_blocksize(zp->z_zfsvfs->z_os, zp->z_id, 788 size, 0, tx); 789 if (error == ENOTSUP) 790 return; 791 ASSERT3U(error, ==, 0); 792 793 /* What blocksize did we actually get? */ 794 dmu_object_size_from_db(zp->z_dbuf, &zp->z_blksz, &dummy); 795} 796 797/* 798 * Free space in a file. 799 * 800 * IN: zp - znode of file to free data in. 801 * off - start of section to free. 802 * len - length of section to free (0 => to EOF). 803 * flag - current file open mode flags. 804 * 805 * RETURN: 0 if success 806 * error code if failure 807 */ 808int 809zfs_freesp(znode_t *zp, uint64_t off, uint64_t len, int flag, boolean_t log) 810{ 811 vnode_t *vp = ZTOV(zp); 812 dmu_tx_t *tx; 813 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 814 zilog_t *zilog = zfsvfs->z_log; 815 rl_t *rl; 816 uint64_t end = off + len; 817 uint64_t size, new_blksz; 818 int error; 819 820 if (ZTOV(zp)->v_type == VFIFO) 821 return (0); 822 823 /* 824 * If we will change zp_size then lock the whole file, 825 * otherwise just lock the range being freed. 826 */ 827 if (len == 0 || off + len > zp->z_phys->zp_size) { 828 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER); 829 } else { 830 rl = zfs_range_lock(zp, off, len, RL_WRITER); 831 /* recheck, in case zp_size changed */ 832 if (off + len > zp->z_phys->zp_size) { 833 /* lost race: file size changed, lock whole file */ 834 zfs_range_unlock(rl); 835 rl = zfs_range_lock(zp, 0, UINT64_MAX, RL_WRITER); 836 } 837 } 838 839 /* 840 * Nothing to do if file already at desired length. 841 */ 842 size = zp->z_phys->zp_size; 843 if (len == 0 && size == off && off != 0) { 844 zfs_range_unlock(rl); 845 return (0); 846 } 847 848 tx = dmu_tx_create(zfsvfs->z_os); 849 dmu_tx_hold_bonus(tx, zp->z_id); 850 new_blksz = 0; 851 if (end > size && 852 (!ISP2(zp->z_blksz) || zp->z_blksz < zfsvfs->z_max_blksz)) { 853 /* 854 * We are growing the file past the current block size. 855 */ 856 if (zp->z_blksz > zp->z_zfsvfs->z_max_blksz) { 857 ASSERT(!ISP2(zp->z_blksz)); 858 new_blksz = MIN(end, SPA_MAXBLOCKSIZE); 859 } else { 860 new_blksz = MIN(end, zp->z_zfsvfs->z_max_blksz); 861 } 862 dmu_tx_hold_write(tx, zp->z_id, 0, MIN(end, new_blksz)); 863 } else if (off < size) { 864 /* 865 * If len == 0, we are truncating the file. 866 */ 867 dmu_tx_hold_free(tx, zp->z_id, off, len ? len : DMU_OBJECT_END); 868 } 869 870 error = dmu_tx_assign(tx, zfsvfs->z_assign); 871 if (error) { 872 if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) 873 dmu_tx_wait(tx); 874 dmu_tx_abort(tx); 875 zfs_range_unlock(rl); 876 return (error); 877 } 878 879 if (new_blksz) 880 zfs_grow_blocksize(zp, new_blksz, tx); 881 882 if (end > size || len == 0) 883 zp->z_phys->zp_size = end; 884 885 if (off < size) { 886 objset_t *os = zfsvfs->z_os; 887 uint64_t rlen = len; 888 889 if (len == 0) 890 rlen = -1; 891 else if (end > size) 892 rlen = size - off; 893 VERIFY(0 == dmu_free_range(os, zp->z_id, off, rlen, tx)); 894 } 895 896 if (log) { 897 zfs_time_stamper(zp, CONTENT_MODIFIED, tx); 898 zfs_log_truncate(zilog, tx, TX_TRUNCATE, zp, off, len); 899 } 900 901 zfs_range_unlock(rl); 902 903 dmu_tx_commit(tx); 904 905 /* 906 * Clear any mapped pages in the truncated region. This has to 907 * happen outside of the transaction to avoid the possibility of 908 * a deadlock with someone trying to push a page that we are 909 * about to invalidate. 910 */ 911 rw_enter(&zp->z_map_lock, RW_WRITER); 912 if (end > size) 913 vnode_pager_setsize(vp, end); 914 else if (len == 0) { 915#if 0 916 error = vtruncbuf(vp, curthread->td_ucred, curthread, end, PAGE_SIZE); 917#else 918 error = vinvalbuf(vp, V_SAVE, curthread, 0, 0); 919 vnode_pager_setsize(vp, end); 920#endif 921 } 922 rw_exit(&zp->z_map_lock); 923 924 return (0); 925} 926 927void 928zfs_create_fs(objset_t *os, cred_t *cr, dmu_tx_t *tx) 929{ 930 zfsvfs_t zfsvfs; 931 uint64_t moid, doid, roid = 0; 932 uint64_t version = ZPL_VERSION; 933 int error; 934 znode_t *rootzp = NULL; 935 vattr_t vattr; 936 937 /* 938 * First attempt to create master node. 939 */ 940 /* 941 * In an empty objset, there are no blocks to read and thus 942 * there can be no i/o errors (which we assert below). 943 */ 944 moid = MASTER_NODE_OBJ; 945 error = zap_create_claim(os, moid, DMU_OT_MASTER_NODE, 946 DMU_OT_NONE, 0, tx); 947 ASSERT(error == 0); 948 949 /* 950 * Set starting attributes. 951 */ 952 953 error = zap_update(os, moid, ZPL_VERSION_OBJ, 8, 1, &version, tx); 954 ASSERT(error == 0); 955 956 /* 957 * Create a delete queue. 958 */ 959 doid = zap_create(os, DMU_OT_UNLINKED_SET, DMU_OT_NONE, 0, tx); 960 961 error = zap_add(os, moid, ZFS_UNLINKED_SET, 8, 1, &doid, tx); 962 ASSERT(error == 0); 963 964 /* 965 * Create root znode. Create minimal znode/vnode/zfsvfs 966 * to allow zfs_mknode to work. 967 */ 968 vattr.va_mask = AT_MODE|AT_UID|AT_GID|AT_TYPE; 969 vattr.va_type = VDIR; 970 vattr.va_mode = S_IFDIR|0755; 971 vattr.va_uid = UID_ROOT; 972 vattr.va_gid = GID_WHEEL; 973 974 rootzp = kmem_cache_alloc(znode_cache, KM_SLEEP); 975 zfs_znode_cache_constructor(rootzp, NULL, 0); 976 rootzp->z_zfsvfs = &zfsvfs; 977 rootzp->z_unlinked = 0; 978 rootzp->z_atime_dirty = 0; 979 rootzp->z_dbuf_held = 0; 980 981 bzero(&zfsvfs, sizeof (zfsvfs_t)); 982 983 zfsvfs.z_os = os; 984 zfsvfs.z_assign = TXG_NOWAIT; 985 zfsvfs.z_parent = &zfsvfs; 986 987 mutex_init(&zfsvfs.z_znodes_lock, NULL, MUTEX_DEFAULT, NULL); 988 list_create(&zfsvfs.z_all_znodes, sizeof (znode_t), 989 offsetof(znode_t, z_link_node)); 990 991 zfs_mknode(rootzp, &vattr, &roid, tx, cr, IS_ROOT_NODE, NULL, 0); 992 ASSERT3U(rootzp->z_id, ==, roid); 993 error = zap_add(os, moid, ZFS_ROOT_OBJ, 8, 1, &roid, tx); 994 ASSERT(error == 0); 995 996 mutex_destroy(&zfsvfs.z_znodes_lock); 997 kmem_cache_free(znode_cache, rootzp); 998} 999#endif /* _KERNEL */ 1000 1001/* 1002 * Given an object number, return its parent object number and whether 1003 * or not the object is an extended attribute directory. 1004 */ 1005static int 1006zfs_obj_to_pobj(objset_t *osp, uint64_t obj, uint64_t *pobjp, int *is_xattrdir) 1007{ 1008 dmu_buf_t *db; 1009 dmu_object_info_t doi; 1010 znode_phys_t *zp; 1011 int error; 1012 1013 if ((error = dmu_bonus_hold(osp, obj, FTAG, &db)) != 0) 1014 return (error); 1015 1016 dmu_object_info_from_db(db, &doi); 1017 if (doi.doi_bonus_type != DMU_OT_ZNODE || 1018 doi.doi_bonus_size < sizeof (znode_phys_t)) { 1019 dmu_buf_rele(db, FTAG); 1020 return (EINVAL); 1021 } 1022 1023 zp = db->db_data; 1024 *pobjp = zp->zp_parent; 1025 *is_xattrdir = ((zp->zp_flags & ZFS_XATTR) != 0) && 1026 S_ISDIR(zp->zp_mode); 1027 dmu_buf_rele(db, FTAG); 1028 1029 return (0); 1030} 1031 1032int 1033zfs_obj_to_path(objset_t *osp, uint64_t obj, char *buf, int len) 1034{ 1035 char *path = buf + len - 1; 1036 int error; 1037 1038 *path = '\0'; 1039 1040 for (;;) { 1041 uint64_t pobj; 1042 char component[MAXNAMELEN + 2]; 1043 size_t complen; 1044 int is_xattrdir; 1045 1046 if ((error = zfs_obj_to_pobj(osp, obj, &pobj, 1047 &is_xattrdir)) != 0) 1048 break; 1049 1050 if (pobj == obj) { 1051 if (path[0] != '/') 1052 *--path = '/'; 1053 break; 1054 } 1055 1056 component[0] = '/'; 1057 if (is_xattrdir) { 1058 (void) sprintf(component + 1, "<xattrdir>"); 1059 } else { 1060 error = zap_value_search(osp, pobj, obj, component + 1); 1061 if (error != 0) 1062 break; 1063 } 1064 1065 complen = strlen(component); 1066 path -= complen; 1067 ASSERT(path >= buf); 1068 bcopy(component, path, complen); 1069 obj = pobj; 1070 } 1071 1072 if (error == 0) 1073 (void) memmove(buf, path, buf + len - path); 1074 return (error); 1075} 1076