zfs_vnops.c revision 224251
1168404Spjd/* 2168404Spjd * CDDL HEADER START 3168404Spjd * 4168404Spjd * The contents of this file are subject to the terms of the 5168404Spjd * Common Development and Distribution License (the "License"). 6168404Spjd * You may not use this file except in compliance with the License. 7168404Spjd * 8168404Spjd * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9168404Spjd * or http://www.opensolaris.org/os/licensing. 10168404Spjd * See the License for the specific language governing permissions 11168404Spjd * and limitations under the License. 12168404Spjd * 13168404Spjd * When distributing Covered Code, include this CDDL HEADER in each 14168404Spjd * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15168404Spjd * If applicable, add the following below this CDDL HEADER, with the 16168404Spjd * fields enclosed by brackets "[]" replaced with your own identifying 17168404Spjd * information: Portions Copyright [yyyy] [name of copyright owner] 18168404Spjd * 19168404Spjd * CDDL HEADER END 20168404Spjd */ 21168404Spjd/* 22212694Smm * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 23168404Spjd */ 24168404Spjd 25169195Spjd/* Portions Copyright 2007 Jeremy Teo */ 26219089Spjd/* Portions Copyright 2010 Robert Milkowski */ 27169195Spjd 28168404Spjd#include <sys/types.h> 29168404Spjd#include <sys/param.h> 30168404Spjd#include <sys/time.h> 31168404Spjd#include <sys/systm.h> 32168404Spjd#include <sys/sysmacros.h> 33168404Spjd#include <sys/resource.h> 34168404Spjd#include <sys/vfs.h> 35168404Spjd#include <sys/vnode.h> 36168404Spjd#include <sys/file.h> 37168404Spjd#include <sys/stat.h> 38168404Spjd#include <sys/kmem.h> 39168404Spjd#include <sys/taskq.h> 40168404Spjd#include <sys/uio.h> 41168404Spjd#include <sys/atomic.h> 42168404Spjd#include <sys/namei.h> 43168404Spjd#include <sys/mman.h> 44168404Spjd#include <sys/cmn_err.h> 45168404Spjd#include <sys/errno.h> 46168404Spjd#include <sys/unistd.h> 47168404Spjd#include <sys/zfs_dir.h> 48168404Spjd#include <sys/zfs_ioctl.h> 49168404Spjd#include <sys/fs/zfs.h> 50168404Spjd#include <sys/dmu.h> 51219089Spjd#include <sys/dmu_objset.h> 52168404Spjd#include <sys/spa.h> 53168404Spjd#include <sys/txg.h> 54168404Spjd#include <sys/dbuf.h> 55168404Spjd#include <sys/zap.h> 56219089Spjd#include <sys/sa.h> 57168404Spjd#include <sys/dirent.h> 58168962Spjd#include <sys/policy.h> 59168962Spjd#include <sys/sunddi.h> 60168404Spjd#include <sys/filio.h> 61209962Smm#include <sys/sid.h> 62168404Spjd#include <sys/zfs_ctldir.h> 63185029Spjd#include <sys/zfs_fuid.h> 64219089Spjd#include <sys/zfs_sa.h> 65168404Spjd#include <sys/dnlc.h> 66168404Spjd#include <sys/zfs_rlock.h> 67185029Spjd#include <sys/extdirent.h> 68185029Spjd#include <sys/kidmap.h> 69168404Spjd#include <sys/bio.h> 70168404Spjd#include <sys/buf.h> 71168404Spjd#include <sys/sf_buf.h> 72168404Spjd#include <sys/sched.h> 73192800Strasz#include <sys/acl.h> 74215401Savg#include <vm/vm_pageout.h> 75168404Spjd 76168404Spjd/* 77168404Spjd * Programming rules. 78168404Spjd * 79168404Spjd * Each vnode op performs some logical unit of work. To do this, the ZPL must 80168404Spjd * properly lock its in-core state, create a DMU transaction, do the work, 81168404Spjd * record this work in the intent log (ZIL), commit the DMU transaction, 82185029Spjd * and wait for the intent log to commit if it is a synchronous operation. 83185029Spjd * Moreover, the vnode ops must work in both normal and log replay context. 84168404Spjd * The ordering of events is important to avoid deadlocks and references 85168404Spjd * to freed memory. The example below illustrates the following Big Rules: 86168404Spjd * 87168404Spjd * (1) A check must be made in each zfs thread for a mounted file system. 88168404Spjd * This is done avoiding races using ZFS_ENTER(zfsvfs). 89185029Spjd * A ZFS_EXIT(zfsvfs) is needed before all returns. Any znodes 90185029Spjd * must be checked with ZFS_VERIFY_ZP(zp). Both of these macros 91185029Spjd * can return EIO from the calling function. 92168404Spjd * 93168404Spjd * (2) VN_RELE() should always be the last thing except for zil_commit() 94168404Spjd * (if necessary) and ZFS_EXIT(). This is for 3 reasons: 95168404Spjd * First, if it's the last reference, the vnode/znode 96168404Spjd * can be freed, so the zp may point to freed memory. Second, the last 97168404Spjd * reference will call zfs_zinactive(), which may induce a lot of work -- 98168404Spjd * pushing cached pages (which acquires range locks) and syncing out 99168404Spjd * cached atime changes. Third, zfs_zinactive() may require a new tx, 100168404Spjd * which could deadlock the system if you were already holding one. 101191900Skmacy * If you must call VN_RELE() within a tx then use VN_RELE_ASYNC(). 102168404Spjd * 103168404Spjd * (3) All range locks must be grabbed before calling dmu_tx_assign(), 104168404Spjd * as they can span dmu_tx_assign() calls. 105168404Spjd * 106209962Smm * (4) Always pass TXG_NOWAIT as the second argument to dmu_tx_assign(). 107168404Spjd * This is critical because we don't want to block while holding locks. 108168404Spjd * Note, in particular, that if a lock is sometimes acquired before 109168404Spjd * the tx assigns, and sometimes after (e.g. z_lock), then failing to 110168404Spjd * use a non-blocking assign can deadlock the system. The scenario: 111168404Spjd * 112168404Spjd * Thread A has grabbed a lock before calling dmu_tx_assign(). 113168404Spjd * Thread B is in an already-assigned tx, and blocks for this lock. 114168404Spjd * Thread A calls dmu_tx_assign(TXG_WAIT) and blocks in txg_wait_open() 115168404Spjd * forever, because the previous txg can't quiesce until B's tx commits. 116168404Spjd * 117168404Spjd * If dmu_tx_assign() returns ERESTART and zfsvfs->z_assign is TXG_NOWAIT, 118168404Spjd * then drop all locks, call dmu_tx_wait(), and try again. 119168404Spjd * 120168404Spjd * (5) If the operation succeeded, generate the intent log entry for it 121168404Spjd * before dropping locks. This ensures that the ordering of events 122168404Spjd * in the intent log matches the order in which they actually occurred. 123209962Smm * During ZIL replay the zfs_log_* functions will update the sequence 124209962Smm * number to indicate the zil transaction has replayed. 125168404Spjd * 126168404Spjd * (6) At the end of each vnode op, the DMU tx must always commit, 127168404Spjd * regardless of whether there were any errors. 128168404Spjd * 129219089Spjd * (7) After dropping all locks, invoke zil_commit(zilog, foid) 130168404Spjd * to ensure that synchronous semantics are provided when necessary. 131168404Spjd * 132168404Spjd * In general, this is how things should be ordered in each vnode op: 133168404Spjd * 134168404Spjd * ZFS_ENTER(zfsvfs); // exit if unmounted 135168404Spjd * top: 136168404Spjd * zfs_dirent_lock(&dl, ...) // lock directory entry (may VN_HOLD()) 137168404Spjd * rw_enter(...); // grab any other locks you need 138168404Spjd * tx = dmu_tx_create(...); // get DMU tx 139168404Spjd * dmu_tx_hold_*(); // hold each object you might modify 140209962Smm * error = dmu_tx_assign(tx, TXG_NOWAIT); // try to assign 141168404Spjd * if (error) { 142168404Spjd * rw_exit(...); // drop locks 143168404Spjd * zfs_dirent_unlock(dl); // unlock directory entry 144168404Spjd * VN_RELE(...); // release held vnodes 145209962Smm * if (error == ERESTART) { 146168404Spjd * dmu_tx_wait(tx); 147168404Spjd * dmu_tx_abort(tx); 148168404Spjd * goto top; 149168404Spjd * } 150168404Spjd * dmu_tx_abort(tx); // abort DMU tx 151168404Spjd * ZFS_EXIT(zfsvfs); // finished in zfs 152168404Spjd * return (error); // really out of space 153168404Spjd * } 154168404Spjd * error = do_real_work(); // do whatever this VOP does 155168404Spjd * if (error == 0) 156168404Spjd * zfs_log_*(...); // on success, make ZIL entry 157168404Spjd * dmu_tx_commit(tx); // commit DMU tx -- error or not 158168404Spjd * rw_exit(...); // drop locks 159168404Spjd * zfs_dirent_unlock(dl); // unlock directory entry 160168404Spjd * VN_RELE(...); // release held vnodes 161219089Spjd * zil_commit(zilog, foid); // synchronous when necessary 162168404Spjd * ZFS_EXIT(zfsvfs); // finished in zfs 163168404Spjd * return (error); // done, report error 164168404Spjd */ 165185029Spjd 166168404Spjd/* ARGSUSED */ 167168404Spjdstatic int 168185029Spjdzfs_open(vnode_t **vpp, int flag, cred_t *cr, caller_context_t *ct) 169168404Spjd{ 170168962Spjd znode_t *zp = VTOZ(*vpp); 171209962Smm zfsvfs_t *zfsvfs = zp->z_zfsvfs; 172168404Spjd 173209962Smm ZFS_ENTER(zfsvfs); 174209962Smm ZFS_VERIFY_ZP(zp); 175209962Smm 176219089Spjd if ((flag & FWRITE) && (zp->z_pflags & ZFS_APPENDONLY) && 177185029Spjd ((flag & FAPPEND) == 0)) { 178209962Smm ZFS_EXIT(zfsvfs); 179185029Spjd return (EPERM); 180185029Spjd } 181185029Spjd 182185029Spjd if (!zfs_has_ctldir(zp) && zp->z_zfsvfs->z_vscan && 183185029Spjd ZTOV(zp)->v_type == VREG && 184219089Spjd !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0) { 185209962Smm if (fs_vscan(*vpp, cr, 0) != 0) { 186209962Smm ZFS_EXIT(zfsvfs); 187185029Spjd return (EACCES); 188209962Smm } 189209962Smm } 190185029Spjd 191168404Spjd /* Keep a count of the synchronous opens in the znode */ 192168962Spjd if (flag & (FSYNC | FDSYNC)) 193168404Spjd atomic_inc_32(&zp->z_sync_cnt); 194185029Spjd 195209962Smm ZFS_EXIT(zfsvfs); 196168404Spjd return (0); 197168404Spjd} 198168404Spjd 199168404Spjd/* ARGSUSED */ 200168404Spjdstatic int 201185029Spjdzfs_close(vnode_t *vp, int flag, int count, offset_t offset, cred_t *cr, 202185029Spjd caller_context_t *ct) 203168404Spjd{ 204168962Spjd znode_t *zp = VTOZ(vp); 205209962Smm zfsvfs_t *zfsvfs = zp->z_zfsvfs; 206168404Spjd 207210470Smm /* 208210470Smm * Clean up any locks held by this process on the vp. 209210470Smm */ 210210470Smm cleanlocks(vp, ddi_get_pid(), 0); 211210470Smm cleanshares(vp, ddi_get_pid()); 212210470Smm 213209962Smm ZFS_ENTER(zfsvfs); 214209962Smm ZFS_VERIFY_ZP(zp); 215209962Smm 216168404Spjd /* Decrement the synchronous opens in the znode */ 217185029Spjd if ((flag & (FSYNC | FDSYNC)) && (count == 1)) 218168404Spjd atomic_dec_32(&zp->z_sync_cnt); 219168404Spjd 220185029Spjd if (!zfs_has_ctldir(zp) && zp->z_zfsvfs->z_vscan && 221185029Spjd ZTOV(zp)->v_type == VREG && 222219089Spjd !(zp->z_pflags & ZFS_AV_QUARANTINED) && zp->z_size > 0) 223185029Spjd VERIFY(fs_vscan(vp, cr, 1) == 0); 224185029Spjd 225209962Smm ZFS_EXIT(zfsvfs); 226168404Spjd return (0); 227168404Spjd} 228168404Spjd 229168404Spjd/* 230168404Spjd * Lseek support for finding holes (cmd == _FIO_SEEK_HOLE) and 231168404Spjd * data (cmd == _FIO_SEEK_DATA). "off" is an in/out parameter. 232168404Spjd */ 233168404Spjdstatic int 234168978Spjdzfs_holey(vnode_t *vp, u_long cmd, offset_t *off) 235168404Spjd{ 236168404Spjd znode_t *zp = VTOZ(vp); 237168404Spjd uint64_t noff = (uint64_t)*off; /* new offset */ 238168404Spjd uint64_t file_sz; 239168404Spjd int error; 240168404Spjd boolean_t hole; 241168404Spjd 242219089Spjd file_sz = zp->z_size; 243168404Spjd if (noff >= file_sz) { 244168404Spjd return (ENXIO); 245168404Spjd } 246168404Spjd 247168962Spjd if (cmd == _FIO_SEEK_HOLE) 248168404Spjd hole = B_TRUE; 249168404Spjd else 250168404Spjd hole = B_FALSE; 251168404Spjd 252168404Spjd error = dmu_offset_next(zp->z_zfsvfs->z_os, zp->z_id, hole, &noff); 253168404Spjd 254168404Spjd /* end of file? */ 255168404Spjd if ((error == ESRCH) || (noff > file_sz)) { 256168404Spjd /* 257168404Spjd * Handle the virtual hole at the end of file. 258168404Spjd */ 259168404Spjd if (hole) { 260168404Spjd *off = file_sz; 261168404Spjd return (0); 262168404Spjd } 263168404Spjd return (ENXIO); 264168404Spjd } 265168404Spjd 266168404Spjd if (noff < *off) 267168404Spjd return (error); 268168404Spjd *off = noff; 269168404Spjd return (error); 270168404Spjd} 271168404Spjd 272168404Spjd/* ARGSUSED */ 273168404Spjdstatic int 274168978Spjdzfs_ioctl(vnode_t *vp, u_long com, intptr_t data, int flag, cred_t *cred, 275185029Spjd int *rvalp, caller_context_t *ct) 276168404Spjd{ 277168962Spjd offset_t off; 278168962Spjd int error; 279168962Spjd zfsvfs_t *zfsvfs; 280185029Spjd znode_t *zp; 281168404Spjd 282168404Spjd switch (com) { 283185029Spjd case _FIOFFS: 284168962Spjd return (0); 285168404Spjd 286168962Spjd /* 287168962Spjd * The following two ioctls are used by bfu. Faking out, 288168962Spjd * necessary to avoid bfu errors. 289168962Spjd */ 290185029Spjd case _FIOGDIO: 291185029Spjd case _FIOSDIO: 292168962Spjd return (0); 293168962Spjd 294185029Spjd case _FIO_SEEK_DATA: 295185029Spjd case _FIO_SEEK_HOLE: 296168962Spjd if (ddi_copyin((void *)data, &off, sizeof (off), flag)) 297168962Spjd return (EFAULT); 298168962Spjd 299185029Spjd zp = VTOZ(vp); 300185029Spjd zfsvfs = zp->z_zfsvfs; 301168404Spjd ZFS_ENTER(zfsvfs); 302185029Spjd ZFS_VERIFY_ZP(zp); 303168404Spjd 304168404Spjd /* offset parameter is in/out */ 305168404Spjd error = zfs_holey(vp, com, &off); 306168404Spjd ZFS_EXIT(zfsvfs); 307168404Spjd if (error) 308168404Spjd return (error); 309168962Spjd if (ddi_copyout(&off, (void *)data, sizeof (off), flag)) 310168962Spjd return (EFAULT); 311168404Spjd return (0); 312168404Spjd } 313168404Spjd return (ENOTTY); 314168404Spjd} 315168404Spjd 316209962Smmstatic vm_page_t 317209962Smmpage_lookup(vnode_t *vp, int64_t start, int64_t off, int64_t nbytes) 318209962Smm{ 319209962Smm vm_object_t obj; 320209962Smm vm_page_t pp; 321209962Smm 322209962Smm obj = vp->v_object; 323209962Smm VM_OBJECT_LOCK_ASSERT(obj, MA_OWNED); 324209962Smm 325209962Smm for (;;) { 326209962Smm if ((pp = vm_page_lookup(obj, OFF_TO_IDX(start))) != NULL && 327209962Smm vm_page_is_valid(pp, (vm_offset_t)off, nbytes)) { 328212652Savg if ((pp->oflags & VPO_BUSY) != 0) { 329212652Savg /* 330212652Savg * Reference the page before unlocking and 331212652Savg * sleeping so that the page daemon is less 332212652Savg * likely to reclaim it. 333212652Savg */ 334212652Savg vm_page_lock_queues(); 335212652Savg vm_page_flag_set(pp, PG_REFERENCED); 336212652Savg vm_page_sleep(pp, "zfsmwb"); 337209962Smm continue; 338212652Savg } 339209962Smm vm_page_busy(pp); 340209962Smm vm_page_undirty(pp); 341209962Smm } else { 342209962Smm if (__predict_false(obj->cache != NULL)) { 343209962Smm vm_page_cache_free(obj, OFF_TO_IDX(start), 344209962Smm OFF_TO_IDX(start) + 1); 345209962Smm } 346209962Smm pp = NULL; 347209962Smm } 348209962Smm break; 349209962Smm } 350209962Smm return (pp); 351209962Smm} 352209962Smm 353209962Smmstatic void 354209962Smmpage_unlock(vm_page_t pp) 355209962Smm{ 356209962Smm 357209962Smm vm_page_wakeup(pp); 358209962Smm} 359209962Smm 360209962Smmstatic caddr_t 361209962Smmzfs_map_page(vm_page_t pp, struct sf_buf **sfp) 362209962Smm{ 363209962Smm 364212951Savg *sfp = sf_buf_alloc(pp, 0); 365209962Smm return ((caddr_t)sf_buf_kva(*sfp)); 366209962Smm} 367209962Smm 368209962Smmstatic void 369209962Smmzfs_unmap_page(struct sf_buf *sf) 370209962Smm{ 371209962Smm 372209962Smm sf_buf_free(sf); 373209962Smm} 374209962Smm 375168404Spjd/* 376168404Spjd * When a file is memory mapped, we must keep the IO data synchronized 377168404Spjd * between the DMU cache and the memory mapped pages. What this means: 378168404Spjd * 379168404Spjd * On Write: If we find a memory mapped page, we write to *both* 380168404Spjd * the page and the dmu buffer. 381168404Spjd */ 382209962Smmstatic void 383209962Smmupdate_pages(vnode_t *vp, int64_t start, int len, objset_t *os, uint64_t oid, 384209962Smm int segflg, dmu_tx_t *tx) 385168404Spjd{ 386168404Spjd vm_object_t obj; 387168404Spjd struct sf_buf *sf; 388212655Savg int off; 389168404Spjd 390168404Spjd ASSERT(vp->v_mount != NULL); 391168404Spjd obj = vp->v_object; 392168404Spjd ASSERT(obj != NULL); 393168404Spjd 394168404Spjd off = start & PAGEOFFSET; 395168404Spjd VM_OBJECT_LOCK(obj); 396168404Spjd for (start &= PAGEMASK; len > 0; start += PAGESIZE) { 397209962Smm vm_page_t pp; 398212655Savg int nbytes = MIN(PAGESIZE - off, len); 399168404Spjd 400209962Smm if ((pp = page_lookup(vp, start, off, nbytes)) != NULL) { 401168404Spjd caddr_t va; 402168404Spjd 403168404Spjd VM_OBJECT_UNLOCK(obj); 404209962Smm va = zfs_map_page(pp, &sf); 405209962Smm if (segflg == UIO_NOCOPY) { 406209962Smm (void) dmu_write(os, oid, start+off, nbytes, 407209962Smm va+off, tx); 408209962Smm } else { 409209962Smm (void) dmu_read(os, oid, start+off, nbytes, 410216378Spjd va+off, DMU_READ_PREFETCH); 411169059Spjd } 412209962Smm zfs_unmap_page(sf); 413168404Spjd VM_OBJECT_LOCK(obj); 414209962Smm page_unlock(pp); 415168404Spjd } 416209962Smm len -= nbytes; 417168404Spjd off = 0; 418168404Spjd } 419168404Spjd VM_OBJECT_UNLOCK(obj); 420168404Spjd} 421168404Spjd 422168404Spjd/* 423219089Spjd * Read with UIO_NOCOPY flag means that sendfile(2) requests 424219089Spjd * ZFS to populate a range of page cache pages with data. 425219089Spjd * 426219089Spjd * NOTE: this function could be optimized to pre-allocate 427219089Spjd * all pages in advance, drain VPO_BUSY on all of them, 428219089Spjd * map them into contiguous KVA region and populate them 429219089Spjd * in one single dmu_read() call. 430219089Spjd */ 431219089Spjdstatic int 432219089Spjdmappedread_sf(vnode_t *vp, int nbytes, uio_t *uio) 433219089Spjd{ 434219089Spjd znode_t *zp = VTOZ(vp); 435219089Spjd objset_t *os = zp->z_zfsvfs->z_os; 436219089Spjd struct sf_buf *sf; 437219089Spjd vm_object_t obj; 438219089Spjd vm_page_t pp; 439219089Spjd int64_t start; 440219089Spjd caddr_t va; 441219089Spjd int len = nbytes; 442219089Spjd int off; 443219089Spjd int error = 0; 444219089Spjd 445219089Spjd ASSERT(uio->uio_segflg == UIO_NOCOPY); 446219089Spjd ASSERT(vp->v_mount != NULL); 447219089Spjd obj = vp->v_object; 448219089Spjd ASSERT(obj != NULL); 449219089Spjd ASSERT((uio->uio_loffset & PAGEOFFSET) == 0); 450219089Spjd 451219089Spjd VM_OBJECT_LOCK(obj); 452219089Spjd for (start = uio->uio_loffset; len > 0; start += PAGESIZE) { 453219089Spjd int bytes = MIN(PAGESIZE, len); 454219089Spjd 455219089Spjd pp = vm_page_grab(obj, OFF_TO_IDX(start), VM_ALLOC_NOBUSY | 456219089Spjd VM_ALLOC_NORMAL | VM_ALLOC_RETRY | VM_ALLOC_IGN_SBUSY); 457219089Spjd if (pp->valid == 0) { 458219089Spjd vm_page_io_start(pp); 459219089Spjd VM_OBJECT_UNLOCK(obj); 460219089Spjd va = zfs_map_page(pp, &sf); 461219089Spjd error = dmu_read(os, zp->z_id, start, bytes, va, 462219089Spjd DMU_READ_PREFETCH); 463219089Spjd if (bytes != PAGESIZE && error == 0) 464219089Spjd bzero(va + bytes, PAGESIZE - bytes); 465219089Spjd zfs_unmap_page(sf); 466219089Spjd VM_OBJECT_LOCK(obj); 467219089Spjd vm_page_io_finish(pp); 468219089Spjd vm_page_lock(pp); 469219089Spjd if (error) { 470219089Spjd vm_page_free(pp); 471219089Spjd } else { 472219089Spjd pp->valid = VM_PAGE_BITS_ALL; 473219089Spjd vm_page_activate(pp); 474219089Spjd } 475219089Spjd vm_page_unlock(pp); 476219089Spjd } 477219089Spjd if (error) 478219089Spjd break; 479219089Spjd uio->uio_resid -= bytes; 480219089Spjd uio->uio_offset += bytes; 481219089Spjd len -= bytes; 482219089Spjd } 483219089Spjd VM_OBJECT_UNLOCK(obj); 484219089Spjd return (error); 485219089Spjd} 486219089Spjd 487219089Spjd/* 488168404Spjd * When a file is memory mapped, we must keep the IO data synchronized 489168404Spjd * between the DMU cache and the memory mapped pages. What this means: 490168404Spjd * 491168404Spjd * On Read: We "read" preferentially from memory mapped pages, 492168404Spjd * else we default from the dmu buffer. 493168404Spjd * 494168404Spjd * NOTE: We will always "break up" the IO into PAGESIZE uiomoves when 495168404Spjd * the file is memory mapped. 496168404Spjd */ 497168404Spjdstatic int 498168404Spjdmappedread(vnode_t *vp, int nbytes, uio_t *uio) 499168404Spjd{ 500168404Spjd znode_t *zp = VTOZ(vp); 501168404Spjd objset_t *os = zp->z_zfsvfs->z_os; 502168404Spjd vm_object_t obj; 503212655Savg int64_t start; 504168926Spjd caddr_t va; 505168404Spjd int len = nbytes; 506212655Savg int off; 507168404Spjd int error = 0; 508168404Spjd 509168404Spjd ASSERT(vp->v_mount != NULL); 510168404Spjd obj = vp->v_object; 511168404Spjd ASSERT(obj != NULL); 512168404Spjd 513168404Spjd start = uio->uio_loffset; 514168404Spjd off = start & PAGEOFFSET; 515168404Spjd VM_OBJECT_LOCK(obj); 516168404Spjd for (start &= PAGEMASK; len > 0; start += PAGESIZE) { 517219089Spjd vm_page_t pp; 518219089Spjd uint64_t bytes = MIN(PAGESIZE - off, len); 519168404Spjd 520219089Spjd if (pp = page_lookup(vp, start, off, bytes)) { 521219089Spjd struct sf_buf *sf; 522219089Spjd caddr_t va; 523212652Savg 524168404Spjd VM_OBJECT_UNLOCK(obj); 525219089Spjd va = zfs_map_page(pp, &sf); 526219089Spjd error = uiomove(va + off, bytes, UIO_READ, uio); 527219089Spjd zfs_unmap_page(sf); 528168404Spjd VM_OBJECT_LOCK(obj); 529219089Spjd page_unlock(pp); 530219089Spjd } else { 531168926Spjd VM_OBJECT_UNLOCK(obj); 532219089Spjd error = dmu_read_uio(os, zp->z_id, uio, bytes); 533168926Spjd VM_OBJECT_LOCK(obj); 534168404Spjd } 535168404Spjd len -= bytes; 536168404Spjd off = 0; 537168404Spjd if (error) 538168404Spjd break; 539168404Spjd } 540168404Spjd VM_OBJECT_UNLOCK(obj); 541168404Spjd return (error); 542168404Spjd} 543168404Spjd 544168404Spjdoffset_t zfs_read_chunk_size = 1024 * 1024; /* Tunable */ 545168404Spjd 546168404Spjd/* 547168404Spjd * Read bytes from specified file into supplied buffer. 548168404Spjd * 549168404Spjd * IN: vp - vnode of file to be read from. 550168404Spjd * uio - structure supplying read location, range info, 551168404Spjd * and return buffer. 552168404Spjd * ioflag - SYNC flags; used to provide FRSYNC semantics. 553168404Spjd * cr - credentials of caller. 554185029Spjd * ct - caller context 555168404Spjd * 556168404Spjd * OUT: uio - updated offset and range, buffer filled. 557168404Spjd * 558168404Spjd * RETURN: 0 if success 559168404Spjd * error code if failure 560168404Spjd * 561168404Spjd * Side Effects: 562168404Spjd * vp - atime updated if byte count > 0 563168404Spjd */ 564168404Spjd/* ARGSUSED */ 565168404Spjdstatic int 566168962Spjdzfs_read(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct) 567168404Spjd{ 568168404Spjd znode_t *zp = VTOZ(vp); 569168404Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 570185029Spjd objset_t *os; 571168404Spjd ssize_t n, nbytes; 572168404Spjd int error; 573168404Spjd rl_t *rl; 574219089Spjd xuio_t *xuio = NULL; 575168404Spjd 576168404Spjd ZFS_ENTER(zfsvfs); 577185029Spjd ZFS_VERIFY_ZP(zp); 578185029Spjd os = zfsvfs->z_os; 579168404Spjd 580219089Spjd if (zp->z_pflags & ZFS_AV_QUARANTINED) { 581185029Spjd ZFS_EXIT(zfsvfs); 582185029Spjd return (EACCES); 583185029Spjd } 584185029Spjd 585168404Spjd /* 586168404Spjd * Validate file offset 587168404Spjd */ 588168404Spjd if (uio->uio_loffset < (offset_t)0) { 589168404Spjd ZFS_EXIT(zfsvfs); 590168404Spjd return (EINVAL); 591168404Spjd } 592168404Spjd 593168404Spjd /* 594168404Spjd * Fasttrack empty reads 595168404Spjd */ 596168404Spjd if (uio->uio_resid == 0) { 597168404Spjd ZFS_EXIT(zfsvfs); 598168404Spjd return (0); 599168404Spjd } 600168404Spjd 601168404Spjd /* 602168962Spjd * Check for mandatory locks 603168962Spjd */ 604219089Spjd if (MANDMODE(zp->z_mode)) { 605168962Spjd if (error = chklock(vp, FREAD, 606168962Spjd uio->uio_loffset, uio->uio_resid, uio->uio_fmode, ct)) { 607168962Spjd ZFS_EXIT(zfsvfs); 608168962Spjd return (error); 609168962Spjd } 610168962Spjd } 611168962Spjd 612168962Spjd /* 613168404Spjd * If we're in FRSYNC mode, sync out this znode before reading it. 614168404Spjd */ 615219089Spjd if (ioflag & FRSYNC || zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) 616219089Spjd zil_commit(zfsvfs->z_log, zp->z_id); 617168404Spjd 618168404Spjd /* 619168404Spjd * Lock the range against changes. 620168404Spjd */ 621168404Spjd rl = zfs_range_lock(zp, uio->uio_loffset, uio->uio_resid, RL_READER); 622168404Spjd 623168404Spjd /* 624168404Spjd * If we are reading past end-of-file we can skip 625168404Spjd * to the end; but we might still need to set atime. 626168404Spjd */ 627219089Spjd if (uio->uio_loffset >= zp->z_size) { 628168404Spjd error = 0; 629168404Spjd goto out; 630168404Spjd } 631168404Spjd 632219089Spjd ASSERT(uio->uio_loffset < zp->z_size); 633219089Spjd n = MIN(uio->uio_resid, zp->z_size - uio->uio_loffset); 634168404Spjd 635219089Spjd#ifdef sun 636219089Spjd if ((uio->uio_extflg == UIO_XUIO) && 637219089Spjd (((xuio_t *)uio)->xu_type == UIOTYPE_ZEROCOPY)) { 638219089Spjd int nblk; 639219089Spjd int blksz = zp->z_blksz; 640219089Spjd uint64_t offset = uio->uio_loffset; 641219089Spjd 642219089Spjd xuio = (xuio_t *)uio; 643219089Spjd if ((ISP2(blksz))) { 644219089Spjd nblk = (P2ROUNDUP(offset + n, blksz) - P2ALIGN(offset, 645219089Spjd blksz)) / blksz; 646219089Spjd } else { 647219089Spjd ASSERT(offset + n <= blksz); 648219089Spjd nblk = 1; 649219089Spjd } 650219089Spjd (void) dmu_xuio_init(xuio, nblk); 651219089Spjd 652219089Spjd if (vn_has_cached_data(vp)) { 653219089Spjd /* 654219089Spjd * For simplicity, we always allocate a full buffer 655219089Spjd * even if we only expect to read a portion of a block. 656219089Spjd */ 657219089Spjd while (--nblk >= 0) { 658219089Spjd (void) dmu_xuio_add(xuio, 659219089Spjd dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), 660219089Spjd blksz), 0, blksz); 661219089Spjd } 662219089Spjd } 663219089Spjd } 664219089Spjd#endif /* sun */ 665219089Spjd 666168404Spjd while (n > 0) { 667168404Spjd nbytes = MIN(n, zfs_read_chunk_size - 668168404Spjd P2PHASE(uio->uio_loffset, zfs_read_chunk_size)); 669168404Spjd 670219089Spjd#ifdef __FreeBSD__ 671219089Spjd if (uio->uio_segflg == UIO_NOCOPY) 672219089Spjd error = mappedread_sf(vp, nbytes, uio); 673219089Spjd else 674219089Spjd#endif /* __FreeBSD__ */ 675168404Spjd if (vn_has_cached_data(vp)) 676168404Spjd error = mappedread(vp, nbytes, uio); 677168404Spjd else 678168404Spjd error = dmu_read_uio(os, zp->z_id, uio, nbytes); 679185029Spjd if (error) { 680185029Spjd /* convert checksum errors into IO errors */ 681185029Spjd if (error == ECKSUM) 682185029Spjd error = EIO; 683168404Spjd break; 684185029Spjd } 685168962Spjd 686168404Spjd n -= nbytes; 687168404Spjd } 688168404Spjdout: 689168404Spjd zfs_range_unlock(rl); 690168404Spjd 691168404Spjd ZFS_ACCESSTIME_STAMP(zfsvfs, zp); 692168404Spjd ZFS_EXIT(zfsvfs); 693168404Spjd return (error); 694168404Spjd} 695168404Spjd 696168404Spjd/* 697168404Spjd * Write the bytes to a file. 698168404Spjd * 699168404Spjd * IN: vp - vnode of file to be written to. 700168404Spjd * uio - structure supplying write location, range info, 701168404Spjd * and data buffer. 702213673Spjd * ioflag - FAPPEND flag set if in append mode. 703168404Spjd * cr - credentials of caller. 704185029Spjd * ct - caller context (NFS/CIFS fem monitor only) 705168404Spjd * 706168404Spjd * OUT: uio - updated offset and range. 707168404Spjd * 708168404Spjd * RETURN: 0 if success 709168404Spjd * error code if failure 710168404Spjd * 711168404Spjd * Timestamps: 712168404Spjd * vp - ctime|mtime updated if byte count > 0 713168404Spjd */ 714219089Spjd 715168404Spjd/* ARGSUSED */ 716168404Spjdstatic int 717168962Spjdzfs_write(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct) 718168404Spjd{ 719168404Spjd znode_t *zp = VTOZ(vp); 720168962Spjd rlim64_t limit = MAXOFFSET_T; 721168404Spjd ssize_t start_resid = uio->uio_resid; 722168404Spjd ssize_t tx_bytes; 723168404Spjd uint64_t end_size; 724168404Spjd dmu_tx_t *tx; 725168404Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 726185029Spjd zilog_t *zilog; 727168404Spjd offset_t woff; 728168404Spjd ssize_t n, nbytes; 729168404Spjd rl_t *rl; 730168404Spjd int max_blksz = zfsvfs->z_max_blksz; 731168404Spjd int error; 732209962Smm arc_buf_t *abuf; 733219089Spjd iovec_t *aiov; 734219089Spjd xuio_t *xuio = NULL; 735219089Spjd int i_iov = 0; 736219089Spjd int iovcnt = uio->uio_iovcnt; 737219089Spjd iovec_t *iovp = uio->uio_iov; 738219089Spjd int write_eof; 739219089Spjd int count = 0; 740219089Spjd sa_bulk_attr_t bulk[4]; 741219089Spjd uint64_t mtime[2], ctime[2]; 742168404Spjd 743168404Spjd /* 744168404Spjd * Fasttrack empty write 745168404Spjd */ 746168404Spjd n = start_resid; 747168404Spjd if (n == 0) 748168404Spjd return (0); 749168404Spjd 750168962Spjd if (limit == RLIM64_INFINITY || limit > MAXOFFSET_T) 751168962Spjd limit = MAXOFFSET_T; 752168962Spjd 753168404Spjd ZFS_ENTER(zfsvfs); 754185029Spjd ZFS_VERIFY_ZP(zp); 755168404Spjd 756219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); 757219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); 758219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_SIZE(zfsvfs), NULL, 759219089Spjd &zp->z_size, 8); 760219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, 761219089Spjd &zp->z_pflags, 8); 762219089Spjd 763168404Spjd /* 764185029Spjd * If immutable or not appending then return EPERM 765185029Spjd */ 766219089Spjd if ((zp->z_pflags & (ZFS_IMMUTABLE | ZFS_READONLY)) || 767219089Spjd ((zp->z_pflags & ZFS_APPENDONLY) && !(ioflag & FAPPEND) && 768219089Spjd (uio->uio_loffset < zp->z_size))) { 769185029Spjd ZFS_EXIT(zfsvfs); 770185029Spjd return (EPERM); 771185029Spjd } 772185029Spjd 773185029Spjd zilog = zfsvfs->z_log; 774185029Spjd 775185029Spjd /* 776219089Spjd * Validate file offset 777219089Spjd */ 778219089Spjd woff = ioflag & FAPPEND ? zp->z_size : uio->uio_loffset; 779219089Spjd if (woff < 0) { 780219089Spjd ZFS_EXIT(zfsvfs); 781219089Spjd return (EINVAL); 782219089Spjd } 783219089Spjd 784219089Spjd /* 785219089Spjd * Check for mandatory locks before calling zfs_range_lock() 786219089Spjd * in order to prevent a deadlock with locks set via fcntl(). 787219089Spjd */ 788219089Spjd if (MANDMODE((mode_t)zp->z_mode) && 789219089Spjd (error = chklock(vp, FWRITE, woff, n, uio->uio_fmode, ct)) != 0) { 790219089Spjd ZFS_EXIT(zfsvfs); 791219089Spjd return (error); 792219089Spjd } 793219089Spjd 794219089Spjd#ifdef sun 795219089Spjd /* 796168404Spjd * Pre-fault the pages to ensure slow (eg NFS) pages 797168404Spjd * don't hold up txg. 798219089Spjd * Skip this if uio contains loaned arc_buf. 799168404Spjd */ 800219089Spjd if ((uio->uio_extflg == UIO_XUIO) && 801219089Spjd (((xuio_t *)uio)->xu_type == UIOTYPE_ZEROCOPY)) 802219089Spjd xuio = (xuio_t *)uio; 803219089Spjd else 804219089Spjd uio_prefaultpages(MIN(n, max_blksz), uio); 805219089Spjd#endif /* sun */ 806168404Spjd 807168404Spjd /* 808168404Spjd * If in append mode, set the io offset pointer to eof. 809168404Spjd */ 810213673Spjd if (ioflag & FAPPEND) { 811168404Spjd /* 812219089Spjd * Obtain an appending range lock to guarantee file append 813219089Spjd * semantics. We reset the write offset once we have the lock. 814168404Spjd */ 815168404Spjd rl = zfs_range_lock(zp, 0, n, RL_APPEND); 816219089Spjd woff = rl->r_off; 817168404Spjd if (rl->r_len == UINT64_MAX) { 818219089Spjd /* 819219089Spjd * We overlocked the file because this write will cause 820219089Spjd * the file block size to increase. 821219089Spjd * Note that zp_size cannot change with this lock held. 822219089Spjd */ 823219089Spjd woff = zp->z_size; 824168404Spjd } 825219089Spjd uio->uio_loffset = woff; 826168404Spjd } else { 827168404Spjd /* 828219089Spjd * Note that if the file block size will change as a result of 829219089Spjd * this write, then this range lock will lock the entire file 830219089Spjd * so that we can re-write the block safely. 831168404Spjd */ 832168404Spjd rl = zfs_range_lock(zp, woff, n, RL_WRITER); 833168404Spjd } 834168404Spjd 835168962Spjd if (woff >= limit) { 836168962Spjd zfs_range_unlock(rl); 837168962Spjd ZFS_EXIT(zfsvfs); 838168962Spjd return (EFBIG); 839168962Spjd } 840168962Spjd 841168962Spjd if ((woff + n) > limit || woff > (limit - n)) 842168962Spjd n = limit - woff; 843168962Spjd 844219089Spjd /* Will this write extend the file length? */ 845219089Spjd write_eof = (woff + n > zp->z_size); 846168404Spjd 847219089Spjd end_size = MAX(zp->z_size, woff + n); 848219089Spjd 849168404Spjd /* 850168404Spjd * Write the file in reasonable size chunks. Each chunk is written 851168404Spjd * in a separate transaction; this keeps the intent log records small 852168404Spjd * and allows us to do more fine-grained space accounting. 853168404Spjd */ 854168404Spjd while (n > 0) { 855209962Smm abuf = NULL; 856209962Smm woff = uio->uio_loffset; 857209962Smmagain: 858219089Spjd if (zfs_owner_overquota(zfsvfs, zp, B_FALSE) || 859219089Spjd zfs_owner_overquota(zfsvfs, zp, B_TRUE)) { 860209962Smm if (abuf != NULL) 861209962Smm dmu_return_arcbuf(abuf); 862209962Smm error = EDQUOT; 863209962Smm break; 864209962Smm } 865209962Smm 866219089Spjd if (xuio && abuf == NULL) { 867219089Spjd ASSERT(i_iov < iovcnt); 868219089Spjd aiov = &iovp[i_iov]; 869219089Spjd abuf = dmu_xuio_arcbuf(xuio, i_iov); 870219089Spjd dmu_xuio_clear(xuio, i_iov); 871219089Spjd DTRACE_PROBE3(zfs_cp_write, int, i_iov, 872219089Spjd iovec_t *, aiov, arc_buf_t *, abuf); 873219089Spjd ASSERT((aiov->iov_base == abuf->b_data) || 874219089Spjd ((char *)aiov->iov_base - (char *)abuf->b_data + 875219089Spjd aiov->iov_len == arc_buf_size(abuf))); 876219089Spjd i_iov++; 877219089Spjd } else if (abuf == NULL && n >= max_blksz && 878219089Spjd woff >= zp->z_size && 879209962Smm P2PHASE(woff, max_blksz) == 0 && 880209962Smm zp->z_blksz == max_blksz) { 881219089Spjd /* 882219089Spjd * This write covers a full block. "Borrow" a buffer 883219089Spjd * from the dmu so that we can fill it before we enter 884219089Spjd * a transaction. This avoids the possibility of 885219089Spjd * holding up the transaction if the data copy hangs 886219089Spjd * up on a pagefault (e.g., from an NFS server mapping). 887219089Spjd */ 888209962Smm size_t cbytes; 889209962Smm 890219089Spjd abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), 891219089Spjd max_blksz); 892209962Smm ASSERT(abuf != NULL); 893209962Smm ASSERT(arc_buf_size(abuf) == max_blksz); 894209962Smm if (error = uiocopy(abuf->b_data, max_blksz, 895209962Smm UIO_WRITE, uio, &cbytes)) { 896209962Smm dmu_return_arcbuf(abuf); 897209962Smm break; 898209962Smm } 899209962Smm ASSERT(cbytes == max_blksz); 900209962Smm } 901209962Smm 902209962Smm /* 903168404Spjd * Start a transaction. 904168404Spjd */ 905168404Spjd tx = dmu_tx_create(zfsvfs->z_os); 906219089Spjd dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); 907168404Spjd dmu_tx_hold_write(tx, zp->z_id, woff, MIN(n, max_blksz)); 908219089Spjd zfs_sa_upgrade_txholds(tx, zp); 909209962Smm error = dmu_tx_assign(tx, TXG_NOWAIT); 910168404Spjd if (error) { 911209962Smm if (error == ERESTART) { 912168404Spjd dmu_tx_wait(tx); 913168404Spjd dmu_tx_abort(tx); 914209962Smm goto again; 915168404Spjd } 916168404Spjd dmu_tx_abort(tx); 917209962Smm if (abuf != NULL) 918209962Smm dmu_return_arcbuf(abuf); 919168404Spjd break; 920168404Spjd } 921168404Spjd 922168404Spjd /* 923168404Spjd * If zfs_range_lock() over-locked we grow the blocksize 924168404Spjd * and then reduce the lock range. This will only happen 925168404Spjd * on the first iteration since zfs_range_reduce() will 926168404Spjd * shrink down r_len to the appropriate size. 927168404Spjd */ 928168404Spjd if (rl->r_len == UINT64_MAX) { 929168404Spjd uint64_t new_blksz; 930168404Spjd 931168404Spjd if (zp->z_blksz > max_blksz) { 932168404Spjd ASSERT(!ISP2(zp->z_blksz)); 933168404Spjd new_blksz = MIN(end_size, SPA_MAXBLOCKSIZE); 934168404Spjd } else { 935168404Spjd new_blksz = MIN(end_size, max_blksz); 936168404Spjd } 937168404Spjd zfs_grow_blocksize(zp, new_blksz, tx); 938168404Spjd zfs_range_reduce(rl, woff, n); 939168404Spjd } 940168404Spjd 941168404Spjd /* 942168404Spjd * XXX - should we really limit each write to z_max_blksz? 943168404Spjd * Perhaps we should use SPA_MAXBLOCKSIZE chunks? 944168404Spjd */ 945168404Spjd nbytes = MIN(n, max_blksz - P2PHASE(woff, max_blksz)); 946168404Spjd 947219089Spjd if (woff + nbytes > zp->z_size) 948168404Spjd vnode_pager_setsize(vp, woff + nbytes); 949168404Spjd 950209962Smm if (abuf == NULL) { 951209962Smm tx_bytes = uio->uio_resid; 952219089Spjd error = dmu_write_uio_dbuf(sa_get_db(zp->z_sa_hdl), 953219089Spjd uio, nbytes, tx); 954209962Smm tx_bytes -= uio->uio_resid; 955168404Spjd } else { 956209962Smm tx_bytes = nbytes; 957219089Spjd ASSERT(xuio == NULL || tx_bytes == aiov->iov_len); 958219089Spjd /* 959219089Spjd * If this is not a full block write, but we are 960219089Spjd * extending the file past EOF and this data starts 961219089Spjd * block-aligned, use assign_arcbuf(). Otherwise, 962219089Spjd * write via dmu_write(). 963219089Spjd */ 964219089Spjd if (tx_bytes < max_blksz && (!write_eof || 965219089Spjd aiov->iov_base != abuf->b_data)) { 966219089Spjd ASSERT(xuio); 967219089Spjd dmu_write(zfsvfs->z_os, zp->z_id, woff, 968219089Spjd aiov->iov_len, aiov->iov_base, tx); 969219089Spjd dmu_return_arcbuf(abuf); 970219089Spjd xuio_stat_wbuf_copied(); 971219089Spjd } else { 972219089Spjd ASSERT(xuio || tx_bytes == max_blksz); 973219089Spjd dmu_assign_arcbuf(sa_get_db(zp->z_sa_hdl), 974219089Spjd woff, abuf, tx); 975219089Spjd } 976209962Smm ASSERT(tx_bytes <= uio->uio_resid); 977209962Smm uioskip(uio, tx_bytes); 978168404Spjd } 979212657Savg if (tx_bytes && vn_has_cached_data(vp)) { 980209962Smm update_pages(vp, woff, tx_bytes, zfsvfs->z_os, 981209962Smm zp->z_id, uio->uio_segflg, tx); 982209962Smm } 983209962Smm 984209962Smm /* 985168404Spjd * If we made no progress, we're done. If we made even 986168404Spjd * partial progress, update the znode and ZIL accordingly. 987168404Spjd */ 988168404Spjd if (tx_bytes == 0) { 989219089Spjd (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs), 990219089Spjd (void *)&zp->z_size, sizeof (uint64_t), tx); 991168404Spjd dmu_tx_commit(tx); 992168404Spjd ASSERT(error != 0); 993168404Spjd break; 994168404Spjd } 995168404Spjd 996168404Spjd /* 997168404Spjd * Clear Set-UID/Set-GID bits on successful write if not 998168404Spjd * privileged and at least one of the excute bits is set. 999168404Spjd * 1000168404Spjd * It would be nice to to this after all writes have 1001168404Spjd * been done, but that would still expose the ISUID/ISGID 1002168404Spjd * to another app after the partial write is committed. 1003185029Spjd * 1004185029Spjd * Note: we don't call zfs_fuid_map_id() here because 1005185029Spjd * user 0 is not an ephemeral uid. 1006168404Spjd */ 1007168404Spjd mutex_enter(&zp->z_acl_lock); 1008219089Spjd if ((zp->z_mode & (S_IXUSR | (S_IXUSR >> 3) | 1009168404Spjd (S_IXUSR >> 6))) != 0 && 1010219089Spjd (zp->z_mode & (S_ISUID | S_ISGID)) != 0 && 1011185029Spjd secpolicy_vnode_setid_retain(vp, cr, 1012219089Spjd (zp->z_mode & S_ISUID) != 0 && zp->z_uid == 0) != 0) { 1013219089Spjd uint64_t newmode; 1014219089Spjd zp->z_mode &= ~(S_ISUID | S_ISGID); 1015219089Spjd newmode = zp->z_mode; 1016219089Spjd (void) sa_update(zp->z_sa_hdl, SA_ZPL_MODE(zfsvfs), 1017219089Spjd (void *)&newmode, sizeof (uint64_t), tx); 1018168404Spjd } 1019168404Spjd mutex_exit(&zp->z_acl_lock); 1020168404Spjd 1021219089Spjd zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, 1022219089Spjd B_TRUE); 1023168404Spjd 1024168404Spjd /* 1025168404Spjd * Update the file size (zp_size) if it has changed; 1026168404Spjd * account for possible concurrent updates. 1027168404Spjd */ 1028219089Spjd while ((end_size = zp->z_size) < uio->uio_loffset) { 1029219089Spjd (void) atomic_cas_64(&zp->z_size, end_size, 1030168404Spjd uio->uio_loffset); 1031219089Spjd ASSERT(error == 0); 1032219089Spjd } 1033219089Spjd /* 1034219089Spjd * If we are replaying and eof is non zero then force 1035219089Spjd * the file size to the specified eof. Note, there's no 1036219089Spjd * concurrency during replay. 1037219089Spjd */ 1038219089Spjd if (zfsvfs->z_replay && zfsvfs->z_replay_eof != 0) 1039219089Spjd zp->z_size = zfsvfs->z_replay_eof; 1040219089Spjd 1041219089Spjd error = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); 1042219089Spjd 1043168404Spjd zfs_log_write(zilog, tx, TX_WRITE, zp, woff, tx_bytes, ioflag); 1044168404Spjd dmu_tx_commit(tx); 1045168404Spjd 1046168404Spjd if (error != 0) 1047168404Spjd break; 1048168404Spjd ASSERT(tx_bytes == nbytes); 1049168404Spjd n -= nbytes; 1050219089Spjd 1051219089Spjd#ifdef sun 1052219089Spjd if (!xuio && n > 0) 1053219089Spjd uio_prefaultpages(MIN(n, max_blksz), uio); 1054219089Spjd#endif /* sun */ 1055168404Spjd } 1056168404Spjd 1057168404Spjd zfs_range_unlock(rl); 1058168404Spjd 1059168404Spjd /* 1060168404Spjd * If we're in replay mode, or we made no progress, return error. 1061168404Spjd * Otherwise, it's at least a partial write, so it's successful. 1062168404Spjd */ 1063209962Smm if (zfsvfs->z_replay || uio->uio_resid == start_resid) { 1064168404Spjd ZFS_EXIT(zfsvfs); 1065168404Spjd return (error); 1066168404Spjd } 1067168404Spjd 1068219089Spjd if (ioflag & (FSYNC | FDSYNC) || 1069219089Spjd zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) 1070219089Spjd zil_commit(zilog, zp->z_id); 1071168404Spjd 1072168404Spjd ZFS_EXIT(zfsvfs); 1073168404Spjd return (0); 1074168404Spjd} 1075168404Spjd 1076168404Spjdvoid 1077219089Spjdzfs_get_done(zgd_t *zgd, int error) 1078168404Spjd{ 1079219089Spjd znode_t *zp = zgd->zgd_private; 1080219089Spjd objset_t *os = zp->z_zfsvfs->z_os; 1081168404Spjd int vfslocked; 1082168404Spjd 1083219089Spjd if (zgd->zgd_db) 1084219089Spjd dmu_buf_rele(zgd->zgd_db, zgd); 1085219089Spjd 1086219089Spjd zfs_range_unlock(zgd->zgd_rl); 1087219089Spjd 1088219089Spjd vfslocked = VFS_LOCK_GIANT(zp->z_zfsvfs->z_vfs); 1089191900Skmacy /* 1090191900Skmacy * Release the vnode asynchronously as we currently have the 1091191900Skmacy * txg stopped from syncing. 1092191900Skmacy */ 1093219089Spjd VN_RELE_ASYNC(ZTOV(zp), dsl_pool_vnrele_taskq(dmu_objset_pool(os))); 1094219089Spjd 1095219089Spjd if (error == 0 && zgd->zgd_bp) 1096219089Spjd zil_add_block(zgd->zgd_zilog, zgd->zgd_bp); 1097219089Spjd 1098168404Spjd kmem_free(zgd, sizeof (zgd_t)); 1099168404Spjd VFS_UNLOCK_GIANT(vfslocked); 1100168404Spjd} 1101168404Spjd 1102214378Smm#ifdef DEBUG 1103214378Smmstatic int zil_fault_io = 0; 1104214378Smm#endif 1105214378Smm 1106168404Spjd/* 1107168404Spjd * Get data to generate a TX_WRITE intent log record. 1108168404Spjd */ 1109168404Spjdint 1110168404Spjdzfs_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio) 1111168404Spjd{ 1112168404Spjd zfsvfs_t *zfsvfs = arg; 1113168404Spjd objset_t *os = zfsvfs->z_os; 1114168404Spjd znode_t *zp; 1115219089Spjd uint64_t object = lr->lr_foid; 1116219089Spjd uint64_t offset = lr->lr_offset; 1117219089Spjd uint64_t size = lr->lr_length; 1118219089Spjd blkptr_t *bp = &lr->lr_blkptr; 1119168404Spjd dmu_buf_t *db; 1120168404Spjd zgd_t *zgd; 1121168404Spjd int error = 0; 1122168404Spjd 1123219089Spjd ASSERT(zio != NULL); 1124219089Spjd ASSERT(size != 0); 1125168404Spjd 1126168404Spjd /* 1127168404Spjd * Nothing to do if the file has been removed 1128168404Spjd */ 1129219089Spjd if (zfs_zget(zfsvfs, object, &zp) != 0) 1130168404Spjd return (ENOENT); 1131168404Spjd if (zp->z_unlinked) { 1132191900Skmacy /* 1133191900Skmacy * Release the vnode asynchronously as we currently have the 1134191900Skmacy * txg stopped from syncing. 1135191900Skmacy */ 1136196307Spjd VN_RELE_ASYNC(ZTOV(zp), 1137196307Spjd dsl_pool_vnrele_taskq(dmu_objset_pool(os))); 1138168404Spjd return (ENOENT); 1139168404Spjd } 1140168404Spjd 1141219089Spjd zgd = (zgd_t *)kmem_zalloc(sizeof (zgd_t), KM_SLEEP); 1142219089Spjd zgd->zgd_zilog = zfsvfs->z_log; 1143219089Spjd zgd->zgd_private = zp; 1144219089Spjd 1145168404Spjd /* 1146168404Spjd * Write records come in two flavors: immediate and indirect. 1147168404Spjd * For small writes it's cheaper to store the data with the 1148168404Spjd * log record (immediate); for large writes it's cheaper to 1149168404Spjd * sync the data and get a pointer to it (indirect) so that 1150168404Spjd * we don't have to write the data twice. 1151168404Spjd */ 1152168404Spjd if (buf != NULL) { /* immediate write */ 1153219089Spjd zgd->zgd_rl = zfs_range_lock(zp, offset, size, RL_READER); 1154168404Spjd /* test for truncation needs to be done while range locked */ 1155219089Spjd if (offset >= zp->z_size) { 1156168404Spjd error = ENOENT; 1157219089Spjd } else { 1158219089Spjd error = dmu_read(os, object, offset, size, buf, 1159219089Spjd DMU_READ_NO_PREFETCH); 1160168404Spjd } 1161219089Spjd ASSERT(error == 0 || error == ENOENT); 1162168404Spjd } else { /* indirect write */ 1163168404Spjd /* 1164168404Spjd * Have to lock the whole block to ensure when it's 1165168404Spjd * written out and it's checksum is being calculated 1166168404Spjd * that no one can change the data. We need to re-check 1167168404Spjd * blocksize after we get the lock in case it's changed! 1168168404Spjd */ 1169168404Spjd for (;;) { 1170219089Spjd uint64_t blkoff; 1171219089Spjd size = zp->z_blksz; 1172219089Spjd blkoff = ISP2(size) ? P2PHASE(offset, size) : offset; 1173219089Spjd offset -= blkoff; 1174219089Spjd zgd->zgd_rl = zfs_range_lock(zp, offset, size, 1175219089Spjd RL_READER); 1176219089Spjd if (zp->z_blksz == size) 1177168404Spjd break; 1178219089Spjd offset += blkoff; 1179219089Spjd zfs_range_unlock(zgd->zgd_rl); 1180168404Spjd } 1181168404Spjd /* test for truncation needs to be done while range locked */ 1182219089Spjd if (lr->lr_offset >= zp->z_size) 1183168404Spjd error = ENOENT; 1184214378Smm#ifdef DEBUG 1185214378Smm if (zil_fault_io) { 1186214378Smm error = EIO; 1187214378Smm zil_fault_io = 0; 1188214378Smm } 1189214378Smm#endif 1190219089Spjd if (error == 0) 1191219089Spjd error = dmu_buf_hold(os, object, offset, zgd, &db, 1192219089Spjd DMU_READ_NO_PREFETCH); 1193214378Smm 1194209962Smm if (error == 0) { 1195219089Spjd zgd->zgd_db = db; 1196219089Spjd zgd->zgd_bp = bp; 1197219089Spjd 1198219089Spjd ASSERT(db->db_offset == offset); 1199219089Spjd ASSERT(db->db_size == size); 1200219089Spjd 1201219089Spjd error = dmu_sync(zio, lr->lr_common.lrc_txg, 1202219089Spjd zfs_get_done, zgd); 1203219089Spjd ASSERT(error || lr->lr_length <= zp->z_blksz); 1204219089Spjd 1205209962Smm /* 1206219089Spjd * On success, we need to wait for the write I/O 1207219089Spjd * initiated by dmu_sync() to complete before we can 1208219089Spjd * release this dbuf. We will finish everything up 1209219089Spjd * in the zfs_get_done() callback. 1210209962Smm */ 1211219089Spjd if (error == 0) 1212219089Spjd return (0); 1213209962Smm 1214219089Spjd if (error == EALREADY) { 1215219089Spjd lr->lr_common.lrc_txtype = TX_WRITE2; 1216219089Spjd error = 0; 1217219089Spjd } 1218209962Smm } 1219168404Spjd } 1220219089Spjd 1221219089Spjd zfs_get_done(zgd, error); 1222219089Spjd 1223168404Spjd return (error); 1224168404Spjd} 1225168404Spjd 1226168404Spjd/*ARGSUSED*/ 1227168404Spjdstatic int 1228185029Spjdzfs_access(vnode_t *vp, int mode, int flag, cred_t *cr, 1229185029Spjd caller_context_t *ct) 1230168404Spjd{ 1231168404Spjd znode_t *zp = VTOZ(vp); 1232168404Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1233168404Spjd int error; 1234168404Spjd 1235168404Spjd ZFS_ENTER(zfsvfs); 1236185029Spjd ZFS_VERIFY_ZP(zp); 1237185029Spjd 1238185029Spjd if (flag & V_ACE_MASK) 1239185029Spjd error = zfs_zaccess(zp, mode, flag, B_FALSE, cr); 1240185029Spjd else 1241185029Spjd error = zfs_zaccess_rwx(zp, mode, flag, cr); 1242185029Spjd 1243168404Spjd ZFS_EXIT(zfsvfs); 1244168404Spjd return (error); 1245168404Spjd} 1246168404Spjd 1247168404Spjd/* 1248211932Smm * If vnode is for a device return a specfs vnode instead. 1249211932Smm */ 1250211932Smmstatic int 1251211932Smmspecvp_check(vnode_t **vpp, cred_t *cr) 1252211932Smm{ 1253211932Smm int error = 0; 1254211932Smm 1255211932Smm if (IS_DEVVP(*vpp)) { 1256211932Smm struct vnode *svp; 1257211932Smm 1258211932Smm svp = specvp(*vpp, (*vpp)->v_rdev, (*vpp)->v_type, cr); 1259211932Smm VN_RELE(*vpp); 1260211932Smm if (svp == NULL) 1261211932Smm error = ENOSYS; 1262211932Smm *vpp = svp; 1263211932Smm } 1264211932Smm return (error); 1265211932Smm} 1266211932Smm 1267211932Smm 1268211932Smm/* 1269168404Spjd * Lookup an entry in a directory, or an extended attribute directory. 1270168404Spjd * If it exists, return a held vnode reference for it. 1271168404Spjd * 1272168404Spjd * IN: dvp - vnode of directory to search. 1273168404Spjd * nm - name of entry to lookup. 1274168404Spjd * pnp - full pathname to lookup [UNUSED]. 1275168404Spjd * flags - LOOKUP_XATTR set if looking for an attribute. 1276168404Spjd * rdir - root directory vnode [UNUSED]. 1277168404Spjd * cr - credentials of caller. 1278185029Spjd * ct - caller context 1279185029Spjd * direntflags - directory lookup flags 1280185029Spjd * realpnp - returned pathname. 1281168404Spjd * 1282168404Spjd * OUT: vpp - vnode of located entry, NULL if not found. 1283168404Spjd * 1284168404Spjd * RETURN: 0 if success 1285168404Spjd * error code if failure 1286168404Spjd * 1287168404Spjd * Timestamps: 1288168404Spjd * NA 1289168404Spjd */ 1290168404Spjd/* ARGSUSED */ 1291168962Spjdstatic int 1292168962Spjdzfs_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, struct componentname *cnp, 1293185029Spjd int nameiop, cred_t *cr, kthread_t *td, int flags) 1294168404Spjd{ 1295168962Spjd znode_t *zdp = VTOZ(dvp); 1296168962Spjd zfsvfs_t *zfsvfs = zdp->z_zfsvfs; 1297211932Smm int error = 0; 1298185029Spjd int *direntflags = NULL; 1299185029Spjd void *realpnp = NULL; 1300168404Spjd 1301211932Smm /* fast path */ 1302211932Smm if (!(flags & (LOOKUP_XATTR | FIGNORECASE))) { 1303211932Smm 1304211932Smm if (dvp->v_type != VDIR) { 1305211932Smm return (ENOTDIR); 1306219089Spjd } else if (zdp->z_sa_hdl == NULL) { 1307211932Smm return (EIO); 1308211932Smm } 1309211932Smm 1310211932Smm if (nm[0] == 0 || (nm[0] == '.' && nm[1] == '\0')) { 1311211932Smm error = zfs_fastaccesschk_execute(zdp, cr); 1312211932Smm if (!error) { 1313211932Smm *vpp = dvp; 1314211932Smm VN_HOLD(*vpp); 1315211932Smm return (0); 1316211932Smm } 1317211932Smm return (error); 1318211932Smm } else { 1319211932Smm vnode_t *tvp = dnlc_lookup(dvp, nm); 1320211932Smm 1321211932Smm if (tvp) { 1322211932Smm error = zfs_fastaccesschk_execute(zdp, cr); 1323211932Smm if (error) { 1324211932Smm VN_RELE(tvp); 1325211932Smm return (error); 1326211932Smm } 1327211932Smm if (tvp == DNLC_NO_VNODE) { 1328211932Smm VN_RELE(tvp); 1329211932Smm return (ENOENT); 1330211932Smm } else { 1331211932Smm *vpp = tvp; 1332211932Smm return (specvp_check(vpp, cr)); 1333211932Smm } 1334211932Smm } 1335211932Smm } 1336211932Smm } 1337211932Smm 1338211932Smm DTRACE_PROBE2(zfs__fastpath__lookup__miss, vnode_t *, dvp, char *, nm); 1339211932Smm 1340168404Spjd ZFS_ENTER(zfsvfs); 1341185029Spjd ZFS_VERIFY_ZP(zdp); 1342168404Spjd 1343168404Spjd *vpp = NULL; 1344168404Spjd 1345185029Spjd if (flags & LOOKUP_XATTR) { 1346168404Spjd#ifdef TODO 1347168404Spjd /* 1348168404Spjd * If the xattr property is off, refuse the lookup request. 1349168404Spjd */ 1350168404Spjd if (!(zfsvfs->z_vfs->vfs_flag & VFS_XATTR)) { 1351168404Spjd ZFS_EXIT(zfsvfs); 1352168404Spjd return (EINVAL); 1353168404Spjd } 1354185029Spjd#endif 1355168404Spjd 1356168404Spjd /* 1357168404Spjd * We don't allow recursive attributes.. 1358168404Spjd * Maybe someday we will. 1359168404Spjd */ 1360219089Spjd if (zdp->z_pflags & ZFS_XATTR) { 1361168404Spjd ZFS_EXIT(zfsvfs); 1362168404Spjd return (EINVAL); 1363168404Spjd } 1364168404Spjd 1365168404Spjd if (error = zfs_get_xattrdir(VTOZ(dvp), vpp, cr, flags)) { 1366168404Spjd ZFS_EXIT(zfsvfs); 1367168404Spjd return (error); 1368168404Spjd } 1369168404Spjd 1370168404Spjd /* 1371168404Spjd * Do we have permission to get into attribute directory? 1372168404Spjd */ 1373168404Spjd 1374185029Spjd if (error = zfs_zaccess(VTOZ(*vpp), ACE_EXECUTE, 0, 1375185029Spjd B_FALSE, cr)) { 1376168404Spjd VN_RELE(*vpp); 1377185029Spjd *vpp = NULL; 1378168404Spjd } 1379168404Spjd 1380168404Spjd ZFS_EXIT(zfsvfs); 1381168404Spjd return (error); 1382168404Spjd } 1383168404Spjd 1384168404Spjd if (dvp->v_type != VDIR) { 1385168404Spjd ZFS_EXIT(zfsvfs); 1386168404Spjd return (ENOTDIR); 1387168404Spjd } 1388168404Spjd 1389168404Spjd /* 1390168404Spjd * Check accessibility of directory. 1391168404Spjd */ 1392168404Spjd 1393185029Spjd if (error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr)) { 1394168404Spjd ZFS_EXIT(zfsvfs); 1395168404Spjd return (error); 1396168404Spjd } 1397168404Spjd 1398185029Spjd if (zfsvfs->z_utf8 && u8_validate(nm, strlen(nm), 1399185029Spjd NULL, U8_VALIDATE_ENTIRE, &error) < 0) { 1400185029Spjd ZFS_EXIT(zfsvfs); 1401185029Spjd return (EILSEQ); 1402185029Spjd } 1403168404Spjd 1404185029Spjd error = zfs_dirlook(zdp, nm, vpp, flags, direntflags, realpnp); 1405211932Smm if (error == 0) 1406211932Smm error = specvp_check(vpp, cr); 1407168962Spjd 1408168404Spjd /* Translate errors and add SAVENAME when needed. */ 1409168404Spjd if (cnp->cn_flags & ISLASTCN) { 1410168404Spjd switch (nameiop) { 1411168404Spjd case CREATE: 1412168404Spjd case RENAME: 1413168404Spjd if (error == ENOENT) { 1414168404Spjd error = EJUSTRETURN; 1415168404Spjd cnp->cn_flags |= SAVENAME; 1416168404Spjd break; 1417168404Spjd } 1418168404Spjd /* FALLTHROUGH */ 1419168404Spjd case DELETE: 1420168404Spjd if (error == 0) 1421168404Spjd cnp->cn_flags |= SAVENAME; 1422168404Spjd break; 1423168404Spjd } 1424168404Spjd } 1425168404Spjd if (error == 0 && (nm[0] != '.' || nm[1] != '\0')) { 1426169198Spjd int ltype = 0; 1427169198Spjd 1428169198Spjd if (cnp->cn_flags & ISDOTDOT) { 1429176559Sattilio ltype = VOP_ISLOCKED(dvp); 1430175294Sattilio VOP_UNLOCK(dvp, 0); 1431169198Spjd } 1432206667Spjd ZFS_EXIT(zfsvfs); 1433219089Spjd error = zfs_vnode_lock(*vpp, cnp->cn_lkflags); 1434168962Spjd if (cnp->cn_flags & ISDOTDOT) 1435175202Sattilio vn_lock(dvp, ltype | LK_RETRY); 1436169172Spjd if (error != 0) { 1437169172Spjd VN_RELE(*vpp); 1438169172Spjd *vpp = NULL; 1439169172Spjd return (error); 1440169172Spjd } 1441206667Spjd } else { 1442206667Spjd ZFS_EXIT(zfsvfs); 1443168404Spjd } 1444168404Spjd 1445168404Spjd#ifdef FREEBSD_NAMECACHE 1446168404Spjd /* 1447168404Spjd * Insert name into cache (as non-existent) if appropriate. 1448168404Spjd */ 1449168404Spjd if (error == ENOENT && (cnp->cn_flags & MAKEENTRY) && nameiop != CREATE) 1450168404Spjd cache_enter(dvp, *vpp, cnp); 1451169170Spjd /* 1452169170Spjd * Insert name into cache if appropriate. 1453169170Spjd */ 1454168404Spjd if (error == 0 && (cnp->cn_flags & MAKEENTRY)) { 1455168404Spjd if (!(cnp->cn_flags & ISLASTCN) || 1456168404Spjd (nameiop != DELETE && nameiop != RENAME)) { 1457168404Spjd cache_enter(dvp, *vpp, cnp); 1458168404Spjd } 1459168404Spjd } 1460168404Spjd#endif 1461168404Spjd 1462168404Spjd return (error); 1463168404Spjd} 1464168404Spjd 1465168404Spjd/* 1466168404Spjd * Attempt to create a new entry in a directory. If the entry 1467168404Spjd * already exists, truncate the file if permissible, else return 1468168404Spjd * an error. Return the vp of the created or trunc'd file. 1469168404Spjd * 1470168404Spjd * IN: dvp - vnode of directory to put new file entry in. 1471168404Spjd * name - name of new file entry. 1472168404Spjd * vap - attributes of new file. 1473168404Spjd * excl - flag indicating exclusive or non-exclusive mode. 1474168404Spjd * mode - mode to open file with. 1475168404Spjd * cr - credentials of caller. 1476168404Spjd * flag - large file flag [UNUSED]. 1477185029Spjd * ct - caller context 1478185029Spjd * vsecp - ACL to be set 1479168404Spjd * 1480168404Spjd * OUT: vpp - vnode of created or trunc'd entry. 1481168404Spjd * 1482168404Spjd * RETURN: 0 if success 1483168404Spjd * error code if failure 1484168404Spjd * 1485168404Spjd * Timestamps: 1486168404Spjd * dvp - ctime|mtime updated if new entry created 1487168404Spjd * vp - ctime|mtime always, atime if new 1488168404Spjd */ 1489185029Spjd 1490168404Spjd/* ARGSUSED */ 1491168404Spjdstatic int 1492168962Spjdzfs_create(vnode_t *dvp, char *name, vattr_t *vap, int excl, int mode, 1493185029Spjd vnode_t **vpp, cred_t *cr, kthread_t *td) 1494168404Spjd{ 1495168404Spjd znode_t *zp, *dzp = VTOZ(dvp); 1496168404Spjd zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 1497185029Spjd zilog_t *zilog; 1498185029Spjd objset_t *os; 1499168404Spjd zfs_dirlock_t *dl; 1500168404Spjd dmu_tx_t *tx; 1501168404Spjd int error; 1502209962Smm ksid_t *ksid; 1503209962Smm uid_t uid; 1504209962Smm gid_t gid = crgetgid(cr); 1505219089Spjd zfs_acl_ids_t acl_ids; 1506209962Smm boolean_t fuid_dirtied; 1507219089Spjd boolean_t have_acl = B_FALSE; 1508185029Spjd void *vsecp = NULL; 1509185029Spjd int flag = 0; 1510168404Spjd 1511185029Spjd /* 1512185029Spjd * If we have an ephemeral id, ACL, or XVATTR then 1513185029Spjd * make sure file system is at proper version 1514185029Spjd */ 1515185029Spjd 1516209962Smm ksid = crgetsid(cr, KSID_OWNER); 1517209962Smm if (ksid) 1518209962Smm uid = ksid_getid(ksid); 1519209962Smm else 1520209962Smm uid = crgetuid(cr); 1521219089Spjd 1522185029Spjd if (zfsvfs->z_use_fuids == B_FALSE && 1523185029Spjd (vsecp || (vap->va_mask & AT_XVATTR) || 1524219089Spjd IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) 1525185029Spjd return (EINVAL); 1526185029Spjd 1527168404Spjd ZFS_ENTER(zfsvfs); 1528185029Spjd ZFS_VERIFY_ZP(dzp); 1529185029Spjd os = zfsvfs->z_os; 1530185029Spjd zilog = zfsvfs->z_log; 1531168404Spjd 1532185029Spjd if (zfsvfs->z_utf8 && u8_validate(name, strlen(name), 1533185029Spjd NULL, U8_VALIDATE_ENTIRE, &error) < 0) { 1534185029Spjd ZFS_EXIT(zfsvfs); 1535185029Spjd return (EILSEQ); 1536185029Spjd } 1537185029Spjd 1538185029Spjd if (vap->va_mask & AT_XVATTR) { 1539197861Spjd if ((error = secpolicy_xvattr(dvp, (xvattr_t *)vap, 1540185029Spjd crgetuid(cr), cr, vap->va_type)) != 0) { 1541185029Spjd ZFS_EXIT(zfsvfs); 1542185029Spjd return (error); 1543185029Spjd } 1544185029Spjd } 1545168404Spjdtop: 1546168404Spjd *vpp = NULL; 1547168404Spjd 1548182905Strasz if ((vap->va_mode & S_ISVTX) && secpolicy_vnode_stky_modify(cr)) 1549182905Strasz vap->va_mode &= ~S_ISVTX; 1550168404Spjd 1551168404Spjd if (*name == '\0') { 1552168404Spjd /* 1553168404Spjd * Null component name refers to the directory itself. 1554168404Spjd */ 1555168404Spjd VN_HOLD(dvp); 1556168404Spjd zp = dzp; 1557168404Spjd dl = NULL; 1558168404Spjd error = 0; 1559168404Spjd } else { 1560168404Spjd /* possible VN_HOLD(zp) */ 1561185029Spjd int zflg = 0; 1562185029Spjd 1563185029Spjd if (flag & FIGNORECASE) 1564185029Spjd zflg |= ZCILOOK; 1565185029Spjd 1566185029Spjd error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, 1567185029Spjd NULL, NULL); 1568185029Spjd if (error) { 1569219089Spjd if (have_acl) 1570219089Spjd zfs_acl_ids_free(&acl_ids); 1571168404Spjd if (strcmp(name, "..") == 0) 1572168404Spjd error = EISDIR; 1573168404Spjd ZFS_EXIT(zfsvfs); 1574168404Spjd return (error); 1575168404Spjd } 1576168404Spjd } 1577219089Spjd 1578185029Spjd if (zp == NULL) { 1579185029Spjd uint64_t txtype; 1580168404Spjd 1581168404Spjd /* 1582168404Spjd * Create a new file object and update the directory 1583168404Spjd * to reference it. 1584168404Spjd */ 1585185029Spjd if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) { 1586219089Spjd if (have_acl) 1587219089Spjd zfs_acl_ids_free(&acl_ids); 1588168404Spjd goto out; 1589168404Spjd } 1590168404Spjd 1591168404Spjd /* 1592168404Spjd * We only support the creation of regular files in 1593168404Spjd * extended attribute directories. 1594168404Spjd */ 1595219089Spjd 1596219089Spjd if ((dzp->z_pflags & ZFS_XATTR) && 1597168404Spjd (vap->va_type != VREG)) { 1598219089Spjd if (have_acl) 1599219089Spjd zfs_acl_ids_free(&acl_ids); 1600168404Spjd error = EINVAL; 1601168404Spjd goto out; 1602168404Spjd } 1603168404Spjd 1604219089Spjd if (!have_acl && (error = zfs_acl_ids_create(dzp, 0, vap, 1605219089Spjd cr, vsecp, &acl_ids)) != 0) 1606219089Spjd goto out; 1607219089Spjd have_acl = B_TRUE; 1608209962Smm 1609209962Smm if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) { 1610211932Smm zfs_acl_ids_free(&acl_ids); 1611209962Smm error = EDQUOT; 1612209962Smm goto out; 1613209962Smm } 1614209962Smm 1615168404Spjd tx = dmu_tx_create(os); 1616219089Spjd 1617219089Spjd dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + 1618219089Spjd ZFS_SA_BASE_ATTR_SIZE); 1619219089Spjd 1620209962Smm fuid_dirtied = zfsvfs->z_fuid_dirty; 1621209962Smm if (fuid_dirtied) 1622209962Smm zfs_fuid_txhold(zfsvfs, tx); 1623168404Spjd dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); 1624219089Spjd dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); 1625219089Spjd if (!zfsvfs->z_use_sa && 1626219089Spjd acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { 1627168404Spjd dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 1628219089Spjd 0, acl_ids.z_aclp->z_acl_bytes); 1629185029Spjd } 1630209962Smm error = dmu_tx_assign(tx, TXG_NOWAIT); 1631168404Spjd if (error) { 1632168404Spjd zfs_dirent_unlock(dl); 1633209962Smm if (error == ERESTART) { 1634168404Spjd dmu_tx_wait(tx); 1635168404Spjd dmu_tx_abort(tx); 1636168404Spjd goto top; 1637168404Spjd } 1638219089Spjd zfs_acl_ids_free(&acl_ids); 1639168404Spjd dmu_tx_abort(tx); 1640168404Spjd ZFS_EXIT(zfsvfs); 1641168404Spjd return (error); 1642168404Spjd } 1643219089Spjd zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); 1644209962Smm 1645209962Smm if (fuid_dirtied) 1646209962Smm zfs_fuid_sync(zfsvfs, tx); 1647209962Smm 1648168404Spjd (void) zfs_link_create(dl, zp, tx, ZNEW); 1649185029Spjd txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap); 1650185029Spjd if (flag & FIGNORECASE) 1651185029Spjd txtype |= TX_CI; 1652185029Spjd zfs_log_create(zilog, tx, txtype, dzp, zp, name, 1653209962Smm vsecp, acl_ids.z_fuidp, vap); 1654209962Smm zfs_acl_ids_free(&acl_ids); 1655168404Spjd dmu_tx_commit(tx); 1656168404Spjd } else { 1657185029Spjd int aflags = (flag & FAPPEND) ? V_APPEND : 0; 1658185029Spjd 1659219089Spjd if (have_acl) 1660219089Spjd zfs_acl_ids_free(&acl_ids); 1661219089Spjd have_acl = B_FALSE; 1662219089Spjd 1663168404Spjd /* 1664168404Spjd * A directory entry already exists for this name. 1665168404Spjd */ 1666168404Spjd /* 1667168962Spjd * Can't truncate an existing file if in exclusive mode. 1668168962Spjd */ 1669168962Spjd if (excl == EXCL) { 1670168962Spjd error = EEXIST; 1671168962Spjd goto out; 1672168962Spjd } 1673168962Spjd /* 1674168404Spjd * Can't open a directory for writing. 1675168404Spjd */ 1676168404Spjd if ((ZTOV(zp)->v_type == VDIR) && (mode & S_IWRITE)) { 1677168404Spjd error = EISDIR; 1678168404Spjd goto out; 1679168404Spjd } 1680168404Spjd /* 1681168404Spjd * Verify requested access to file. 1682168404Spjd */ 1683185029Spjd if (mode && (error = zfs_zaccess_rwx(zp, mode, aflags, cr))) { 1684168404Spjd goto out; 1685168404Spjd } 1686168404Spjd 1687168404Spjd mutex_enter(&dzp->z_lock); 1688168404Spjd dzp->z_seq++; 1689168404Spjd mutex_exit(&dzp->z_lock); 1690168404Spjd 1691168404Spjd /* 1692168404Spjd * Truncate regular files if requested. 1693168404Spjd */ 1694168404Spjd if ((ZTOV(zp)->v_type == VREG) && 1695168404Spjd (vap->va_mask & AT_SIZE) && (vap->va_size == 0)) { 1696185029Spjd /* we can't hold any locks when calling zfs_freesp() */ 1697185029Spjd zfs_dirent_unlock(dl); 1698185029Spjd dl = NULL; 1699168404Spjd error = zfs_freesp(zp, 0, 0, mode, TRUE); 1700185029Spjd if (error == 0) { 1701185029Spjd vnevent_create(ZTOV(zp), ct); 1702168404Spjd } 1703168404Spjd } 1704168404Spjd } 1705168404Spjdout: 1706168404Spjd if (dl) 1707168404Spjd zfs_dirent_unlock(dl); 1708168404Spjd 1709168404Spjd if (error) { 1710168404Spjd if (zp) 1711168404Spjd VN_RELE(ZTOV(zp)); 1712168962Spjd } else { 1713168962Spjd *vpp = ZTOV(zp); 1714211932Smm error = specvp_check(vpp, cr); 1715168404Spjd } 1716168404Spjd 1717219089Spjd if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) 1718219089Spjd zil_commit(zilog, 0); 1719219089Spjd 1720168404Spjd ZFS_EXIT(zfsvfs); 1721168404Spjd return (error); 1722168404Spjd} 1723168404Spjd 1724168404Spjd/* 1725168404Spjd * Remove an entry from a directory. 1726168404Spjd * 1727168404Spjd * IN: dvp - vnode of directory to remove entry from. 1728168404Spjd * name - name of entry to remove. 1729168404Spjd * cr - credentials of caller. 1730185029Spjd * ct - caller context 1731185029Spjd * flags - case flags 1732168404Spjd * 1733168404Spjd * RETURN: 0 if success 1734168404Spjd * error code if failure 1735168404Spjd * 1736168404Spjd * Timestamps: 1737168404Spjd * dvp - ctime|mtime 1738168404Spjd * vp - ctime (if nlink > 0) 1739168404Spjd */ 1740219089Spjd 1741219089Spjduint64_t null_xattr = 0; 1742219089Spjd 1743185029Spjd/*ARGSUSED*/ 1744168404Spjdstatic int 1745185029Spjdzfs_remove(vnode_t *dvp, char *name, cred_t *cr, caller_context_t *ct, 1746185029Spjd int flags) 1747168404Spjd{ 1748168404Spjd znode_t *zp, *dzp = VTOZ(dvp); 1749219089Spjd znode_t *xzp; 1750168404Spjd vnode_t *vp; 1751168404Spjd zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 1752185029Spjd zilog_t *zilog; 1753168962Spjd uint64_t acl_obj, xattr_obj; 1754219089Spjd uint64_t xattr_obj_unlinked = 0; 1755219089Spjd uint64_t obj = 0; 1756168404Spjd zfs_dirlock_t *dl; 1757168404Spjd dmu_tx_t *tx; 1758168962Spjd boolean_t may_delete_now, delete_now = FALSE; 1759185029Spjd boolean_t unlinked, toobig = FALSE; 1760185029Spjd uint64_t txtype; 1761185029Spjd pathname_t *realnmp = NULL; 1762185029Spjd pathname_t realnm; 1763168404Spjd int error; 1764185029Spjd int zflg = ZEXISTS; 1765168404Spjd 1766168404Spjd ZFS_ENTER(zfsvfs); 1767185029Spjd ZFS_VERIFY_ZP(dzp); 1768185029Spjd zilog = zfsvfs->z_log; 1769168404Spjd 1770185029Spjd if (flags & FIGNORECASE) { 1771185029Spjd zflg |= ZCILOOK; 1772185029Spjd pn_alloc(&realnm); 1773185029Spjd realnmp = &realnm; 1774185029Spjd } 1775185029Spjd 1776168404Spjdtop: 1777219089Spjd xattr_obj = 0; 1778219089Spjd xzp = NULL; 1779168404Spjd /* 1780168404Spjd * Attempt to lock directory; fail if entry doesn't exist. 1781168404Spjd */ 1782185029Spjd if (error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, 1783185029Spjd NULL, realnmp)) { 1784185029Spjd if (realnmp) 1785185029Spjd pn_free(realnmp); 1786168404Spjd ZFS_EXIT(zfsvfs); 1787168404Spjd return (error); 1788168404Spjd } 1789168404Spjd 1790168404Spjd vp = ZTOV(zp); 1791168404Spjd 1792168962Spjd if (error = zfs_zaccess_delete(dzp, zp, cr)) { 1793168404Spjd goto out; 1794168962Spjd } 1795168404Spjd 1796168962Spjd /* 1797168962Spjd * Need to use rmdir for removing directories. 1798168962Spjd */ 1799168962Spjd if (vp->v_type == VDIR) { 1800168962Spjd error = EPERM; 1801168962Spjd goto out; 1802168962Spjd } 1803168962Spjd 1804185029Spjd vnevent_remove(vp, dvp, name, ct); 1805168962Spjd 1806185029Spjd if (realnmp) 1807185029Spjd dnlc_remove(dvp, realnmp->pn_buf); 1808185029Spjd else 1809185029Spjd dnlc_remove(dvp, name); 1810168404Spjd 1811219089Spjd VI_LOCK(vp); 1812219089Spjd may_delete_now = vp->v_count == 1 && !vn_has_cached_data(vp); 1813219089Spjd VI_UNLOCK(vp); 1814168962Spjd 1815168404Spjd /* 1816168404Spjd * We may delete the znode now, or we may put it in the unlinked set; 1817168404Spjd * it depends on whether we're the last link, and on whether there are 1818168404Spjd * other holds on the vnode. So we dmu_tx_hold() the right things to 1819168404Spjd * allow for either case. 1820168404Spjd */ 1821219089Spjd obj = zp->z_id; 1822168404Spjd tx = dmu_tx_create(zfsvfs->z_os); 1823168404Spjd dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); 1824219089Spjd dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); 1825219089Spjd zfs_sa_upgrade_txholds(tx, zp); 1826219089Spjd zfs_sa_upgrade_txholds(tx, dzp); 1827185029Spjd if (may_delete_now) { 1828185029Spjd toobig = 1829219089Spjd zp->z_size > zp->z_blksz * DMU_MAX_DELETEBLKCNT; 1830185029Spjd /* if the file is too big, only hold_free a token amount */ 1831185029Spjd dmu_tx_hold_free(tx, zp->z_id, 0, 1832185029Spjd (toobig ? DMU_MAX_ACCESS : DMU_OBJECT_END)); 1833185029Spjd } 1834168404Spjd 1835168404Spjd /* are there any extended attributes? */ 1836219089Spjd error = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), 1837219089Spjd &xattr_obj, sizeof (xattr_obj)); 1838219089Spjd if (error == 0 && xattr_obj) { 1839219089Spjd error = zfs_zget(zfsvfs, xattr_obj, &xzp); 1840219089Spjd ASSERT3U(error, ==, 0); 1841219089Spjd dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); 1842219089Spjd dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE); 1843168404Spjd } 1844168404Spjd 1845219089Spjd mutex_enter(&zp->z_lock); 1846219089Spjd if ((acl_obj = zfs_external_acl(zp)) != 0 && may_delete_now) 1847168962Spjd dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END); 1848219089Spjd mutex_exit(&zp->z_lock); 1849168962Spjd 1850168404Spjd /* charge as an update -- would be nice not to charge at all */ 1851168404Spjd dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); 1852168404Spjd 1853209962Smm error = dmu_tx_assign(tx, TXG_NOWAIT); 1854168404Spjd if (error) { 1855168404Spjd zfs_dirent_unlock(dl); 1856168962Spjd VN_RELE(vp); 1857219089Spjd if (xzp) 1858219089Spjd VN_RELE(ZTOV(xzp)); 1859209962Smm if (error == ERESTART) { 1860168404Spjd dmu_tx_wait(tx); 1861168404Spjd dmu_tx_abort(tx); 1862168404Spjd goto top; 1863168404Spjd } 1864185029Spjd if (realnmp) 1865185029Spjd pn_free(realnmp); 1866168404Spjd dmu_tx_abort(tx); 1867168404Spjd ZFS_EXIT(zfsvfs); 1868168404Spjd return (error); 1869168404Spjd } 1870168404Spjd 1871168404Spjd /* 1872168404Spjd * Remove the directory entry. 1873168404Spjd */ 1874185029Spjd error = zfs_link_destroy(dl, zp, tx, zflg, &unlinked); 1875168404Spjd 1876168404Spjd if (error) { 1877168404Spjd dmu_tx_commit(tx); 1878168404Spjd goto out; 1879168404Spjd } 1880168404Spjd 1881219089Spjd if (unlinked) { 1882219089Spjd 1883219089Spjd /* 1884219089Spjd * Hold z_lock so that we can make sure that the ACL obj 1885219089Spjd * hasn't changed. Could have been deleted due to 1886219089Spjd * zfs_sa_upgrade(). 1887219089Spjd */ 1888219089Spjd mutex_enter(&zp->z_lock); 1889168962Spjd VI_LOCK(vp); 1890219089Spjd (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), 1891219089Spjd &xattr_obj_unlinked, sizeof (xattr_obj_unlinked)); 1892185029Spjd delete_now = may_delete_now && !toobig && 1893168962Spjd vp->v_count == 1 && !vn_has_cached_data(vp) && 1894219089Spjd xattr_obj == xattr_obj_unlinked && zfs_external_acl(zp) == 1895219089Spjd acl_obj; 1896168962Spjd VI_UNLOCK(vp); 1897168962Spjd } 1898168962Spjd 1899168962Spjd if (delete_now) { 1900219089Spjd if (xattr_obj_unlinked) { 1901219089Spjd ASSERT3U(xzp->z_links, ==, 2); 1902168962Spjd mutex_enter(&xzp->z_lock); 1903168962Spjd xzp->z_unlinked = 1; 1904219089Spjd xzp->z_links = 0; 1905219089Spjd error = sa_update(xzp->z_sa_hdl, SA_ZPL_LINKS(zfsvfs), 1906219089Spjd &xzp->z_links, sizeof (xzp->z_links), tx); 1907219089Spjd ASSERT3U(error, ==, 0); 1908168962Spjd mutex_exit(&xzp->z_lock); 1909168962Spjd zfs_unlinked_add(xzp, tx); 1910219089Spjd 1911219089Spjd if (zp->z_is_sa) 1912219089Spjd error = sa_remove(zp->z_sa_hdl, 1913219089Spjd SA_ZPL_XATTR(zfsvfs), tx); 1914219089Spjd else 1915219089Spjd error = sa_update(zp->z_sa_hdl, 1916219089Spjd SA_ZPL_XATTR(zfsvfs), &null_xattr, 1917219089Spjd sizeof (uint64_t), tx); 1918219089Spjd ASSERT3U(error, ==, 0); 1919168962Spjd } 1920168962Spjd VI_LOCK(vp); 1921168962Spjd vp->v_count--; 1922168962Spjd ASSERT3U(vp->v_count, ==, 0); 1923168962Spjd VI_UNLOCK(vp); 1924168962Spjd mutex_exit(&zp->z_lock); 1925168962Spjd zfs_znode_delete(zp, tx); 1926168962Spjd } else if (unlinked) { 1927219089Spjd mutex_exit(&zp->z_lock); 1928168404Spjd zfs_unlinked_add(zp, tx); 1929168962Spjd } 1930168404Spjd 1931185029Spjd txtype = TX_REMOVE; 1932185029Spjd if (flags & FIGNORECASE) 1933185029Spjd txtype |= TX_CI; 1934219089Spjd zfs_log_remove(zilog, tx, txtype, dzp, name, obj); 1935168404Spjd 1936168404Spjd dmu_tx_commit(tx); 1937168404Spjdout: 1938185029Spjd if (realnmp) 1939185029Spjd pn_free(realnmp); 1940185029Spjd 1941168404Spjd zfs_dirent_unlock(dl); 1942168404Spjd 1943219089Spjd if (!delete_now) 1944168962Spjd VN_RELE(vp); 1945219089Spjd if (xzp) 1946168962Spjd VN_RELE(ZTOV(xzp)); 1947168962Spjd 1948219089Spjd if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) 1949219089Spjd zil_commit(zilog, 0); 1950219089Spjd 1951168404Spjd ZFS_EXIT(zfsvfs); 1952168404Spjd return (error); 1953168404Spjd} 1954168404Spjd 1955168404Spjd/* 1956168404Spjd * Create a new directory and insert it into dvp using the name 1957168404Spjd * provided. Return a pointer to the inserted directory. 1958168404Spjd * 1959168404Spjd * IN: dvp - vnode of directory to add subdir to. 1960168404Spjd * dirname - name of new directory. 1961168404Spjd * vap - attributes of new directory. 1962168404Spjd * cr - credentials of caller. 1963185029Spjd * ct - caller context 1964185029Spjd * vsecp - ACL to be set 1965168404Spjd * 1966168404Spjd * OUT: vpp - vnode of created directory. 1967168404Spjd * 1968168404Spjd * RETURN: 0 if success 1969168404Spjd * error code if failure 1970168404Spjd * 1971168404Spjd * Timestamps: 1972168404Spjd * dvp - ctime|mtime updated 1973168404Spjd * vp - ctime|mtime|atime updated 1974168404Spjd */ 1975185029Spjd/*ARGSUSED*/ 1976168404Spjdstatic int 1977185029Spjdzfs_mkdir(vnode_t *dvp, char *dirname, vattr_t *vap, vnode_t **vpp, cred_t *cr, 1978185029Spjd caller_context_t *ct, int flags, vsecattr_t *vsecp) 1979168404Spjd{ 1980168404Spjd znode_t *zp, *dzp = VTOZ(dvp); 1981168404Spjd zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 1982185029Spjd zilog_t *zilog; 1983168404Spjd zfs_dirlock_t *dl; 1984185029Spjd uint64_t txtype; 1985168404Spjd dmu_tx_t *tx; 1986168404Spjd int error; 1987185029Spjd int zf = ZNEW; 1988209962Smm ksid_t *ksid; 1989209962Smm uid_t uid; 1990209962Smm gid_t gid = crgetgid(cr); 1991219089Spjd zfs_acl_ids_t acl_ids; 1992209962Smm boolean_t fuid_dirtied; 1993168404Spjd 1994168404Spjd ASSERT(vap->va_type == VDIR); 1995168404Spjd 1996185029Spjd /* 1997185029Spjd * If we have an ephemeral id, ACL, or XVATTR then 1998185029Spjd * make sure file system is at proper version 1999185029Spjd */ 2000185029Spjd 2001209962Smm ksid = crgetsid(cr, KSID_OWNER); 2002209962Smm if (ksid) 2003209962Smm uid = ksid_getid(ksid); 2004209962Smm else 2005209962Smm uid = crgetuid(cr); 2006185029Spjd if (zfsvfs->z_use_fuids == B_FALSE && 2007219089Spjd (vsecp || (vap->va_mask & AT_XVATTR) || 2008219089Spjd IS_EPHEMERAL(uid) || IS_EPHEMERAL(gid))) 2009185029Spjd return (EINVAL); 2010185029Spjd 2011168404Spjd ZFS_ENTER(zfsvfs); 2012185029Spjd ZFS_VERIFY_ZP(dzp); 2013185029Spjd zilog = zfsvfs->z_log; 2014168404Spjd 2015219089Spjd if (dzp->z_pflags & ZFS_XATTR) { 2016168404Spjd ZFS_EXIT(zfsvfs); 2017168404Spjd return (EINVAL); 2018168404Spjd } 2019168404Spjd 2020185029Spjd if (zfsvfs->z_utf8 && u8_validate(dirname, 2021185029Spjd strlen(dirname), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { 2022185029Spjd ZFS_EXIT(zfsvfs); 2023185029Spjd return (EILSEQ); 2024185029Spjd } 2025185029Spjd if (flags & FIGNORECASE) 2026185029Spjd zf |= ZCILOOK; 2027185029Spjd 2028219089Spjd if (vap->va_mask & AT_XVATTR) { 2029197861Spjd if ((error = secpolicy_xvattr(dvp, (xvattr_t *)vap, 2030185029Spjd crgetuid(cr), cr, vap->va_type)) != 0) { 2031185029Spjd ZFS_EXIT(zfsvfs); 2032185029Spjd return (error); 2033185029Spjd } 2034219089Spjd } 2035185029Spjd 2036219089Spjd if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, 2037219089Spjd vsecp, &acl_ids)) != 0) { 2038219089Spjd ZFS_EXIT(zfsvfs); 2039219089Spjd return (error); 2040219089Spjd } 2041168404Spjd /* 2042168404Spjd * First make sure the new directory doesn't exist. 2043219089Spjd * 2044219089Spjd * Existence is checked first to make sure we don't return 2045219089Spjd * EACCES instead of EEXIST which can cause some applications 2046219089Spjd * to fail. 2047168404Spjd */ 2048185029Spjdtop: 2049185029Spjd *vpp = NULL; 2050185029Spjd 2051185029Spjd if (error = zfs_dirent_lock(&dl, dzp, dirname, &zp, zf, 2052185029Spjd NULL, NULL)) { 2053219089Spjd zfs_acl_ids_free(&acl_ids); 2054168404Spjd ZFS_EXIT(zfsvfs); 2055168404Spjd return (error); 2056168404Spjd } 2057168404Spjd 2058185029Spjd if (error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr)) { 2059219089Spjd zfs_acl_ids_free(&acl_ids); 2060168404Spjd zfs_dirent_unlock(dl); 2061168404Spjd ZFS_EXIT(zfsvfs); 2062168404Spjd return (error); 2063168404Spjd } 2064168404Spjd 2065209962Smm if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) { 2066211932Smm zfs_acl_ids_free(&acl_ids); 2067209962Smm zfs_dirent_unlock(dl); 2068209962Smm ZFS_EXIT(zfsvfs); 2069209962Smm return (EDQUOT); 2070209962Smm } 2071209962Smm 2072168404Spjd /* 2073168404Spjd * Add a new entry to the directory. 2074168404Spjd */ 2075168404Spjd tx = dmu_tx_create(zfsvfs->z_os); 2076168404Spjd dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname); 2077168404Spjd dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); 2078209962Smm fuid_dirtied = zfsvfs->z_fuid_dirty; 2079209962Smm if (fuid_dirtied) 2080209962Smm zfs_fuid_txhold(zfsvfs, tx); 2081219089Spjd if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { 2082219089Spjd dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, 2083219089Spjd acl_ids.z_aclp->z_acl_bytes); 2084219089Spjd } 2085219089Spjd 2086219089Spjd dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + 2087219089Spjd ZFS_SA_BASE_ATTR_SIZE); 2088219089Spjd 2089209962Smm error = dmu_tx_assign(tx, TXG_NOWAIT); 2090168404Spjd if (error) { 2091168404Spjd zfs_dirent_unlock(dl); 2092209962Smm if (error == ERESTART) { 2093168404Spjd dmu_tx_wait(tx); 2094168404Spjd dmu_tx_abort(tx); 2095168404Spjd goto top; 2096168404Spjd } 2097219089Spjd zfs_acl_ids_free(&acl_ids); 2098168404Spjd dmu_tx_abort(tx); 2099168404Spjd ZFS_EXIT(zfsvfs); 2100168404Spjd return (error); 2101168404Spjd } 2102168404Spjd 2103168404Spjd /* 2104168404Spjd * Create new node. 2105168404Spjd */ 2106219089Spjd zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); 2107168404Spjd 2108209962Smm if (fuid_dirtied) 2109209962Smm zfs_fuid_sync(zfsvfs, tx); 2110219089Spjd 2111168404Spjd /* 2112168404Spjd * Now put new name in parent dir. 2113168404Spjd */ 2114168404Spjd (void) zfs_link_create(dl, zp, tx, ZNEW); 2115168404Spjd 2116168404Spjd *vpp = ZTOV(zp); 2117168404Spjd 2118185029Spjd txtype = zfs_log_create_txtype(Z_DIR, vsecp, vap); 2119185029Spjd if (flags & FIGNORECASE) 2120185029Spjd txtype |= TX_CI; 2121209962Smm zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, vsecp, 2122209962Smm acl_ids.z_fuidp, vap); 2123185029Spjd 2124209962Smm zfs_acl_ids_free(&acl_ids); 2125219089Spjd 2126168404Spjd dmu_tx_commit(tx); 2127168404Spjd 2128168404Spjd zfs_dirent_unlock(dl); 2129168404Spjd 2130219089Spjd if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) 2131219089Spjd zil_commit(zilog, 0); 2132219089Spjd 2133168404Spjd ZFS_EXIT(zfsvfs); 2134168404Spjd return (0); 2135168404Spjd} 2136168404Spjd 2137168404Spjd/* 2138168404Spjd * Remove a directory subdir entry. If the current working 2139168404Spjd * directory is the same as the subdir to be removed, the 2140168404Spjd * remove will fail. 2141168404Spjd * 2142168404Spjd * IN: dvp - vnode of directory to remove from. 2143168404Spjd * name - name of directory to be removed. 2144168404Spjd * cwd - vnode of current working directory. 2145168404Spjd * cr - credentials of caller. 2146185029Spjd * ct - caller context 2147185029Spjd * flags - case flags 2148168404Spjd * 2149168404Spjd * RETURN: 0 if success 2150168404Spjd * error code if failure 2151168404Spjd * 2152168404Spjd * Timestamps: 2153168404Spjd * dvp - ctime|mtime updated 2154168404Spjd */ 2155185029Spjd/*ARGSUSED*/ 2156168404Spjdstatic int 2157185029Spjdzfs_rmdir(vnode_t *dvp, char *name, vnode_t *cwd, cred_t *cr, 2158185029Spjd caller_context_t *ct, int flags) 2159168404Spjd{ 2160168404Spjd znode_t *dzp = VTOZ(dvp); 2161168404Spjd znode_t *zp; 2162168404Spjd vnode_t *vp; 2163168404Spjd zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 2164185029Spjd zilog_t *zilog; 2165168404Spjd zfs_dirlock_t *dl; 2166168404Spjd dmu_tx_t *tx; 2167168404Spjd int error; 2168185029Spjd int zflg = ZEXISTS; 2169168404Spjd 2170168962Spjd ZFS_ENTER(zfsvfs); 2171185029Spjd ZFS_VERIFY_ZP(dzp); 2172185029Spjd zilog = zfsvfs->z_log; 2173168404Spjd 2174185029Spjd if (flags & FIGNORECASE) 2175185029Spjd zflg |= ZCILOOK; 2176168404Spjdtop: 2177168404Spjd zp = NULL; 2178168404Spjd 2179168404Spjd /* 2180168404Spjd * Attempt to lock directory; fail if entry doesn't exist. 2181168404Spjd */ 2182185029Spjd if (error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, 2183185029Spjd NULL, NULL)) { 2184168404Spjd ZFS_EXIT(zfsvfs); 2185168404Spjd return (error); 2186168404Spjd } 2187168404Spjd 2188168404Spjd vp = ZTOV(zp); 2189168404Spjd 2190168404Spjd if (error = zfs_zaccess_delete(dzp, zp, cr)) { 2191168404Spjd goto out; 2192168404Spjd } 2193168404Spjd 2194168962Spjd if (vp->v_type != VDIR) { 2195168962Spjd error = ENOTDIR; 2196168962Spjd goto out; 2197168962Spjd } 2198168962Spjd 2199168962Spjd if (vp == cwd) { 2200168962Spjd error = EINVAL; 2201168962Spjd goto out; 2202168962Spjd } 2203168962Spjd 2204185029Spjd vnevent_rmdir(vp, dvp, name, ct); 2205168962Spjd 2206168404Spjd /* 2207168404Spjd * Grab a lock on the directory to make sure that noone is 2208168404Spjd * trying to add (or lookup) entries while we are removing it. 2209168404Spjd */ 2210168404Spjd rw_enter(&zp->z_name_lock, RW_WRITER); 2211168404Spjd 2212168404Spjd /* 2213168404Spjd * Grab a lock on the parent pointer to make sure we play well 2214168404Spjd * with the treewalk and directory rename code. 2215168404Spjd */ 2216168404Spjd rw_enter(&zp->z_parent_lock, RW_WRITER); 2217168404Spjd 2218168404Spjd tx = dmu_tx_create(zfsvfs->z_os); 2219168404Spjd dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); 2220219089Spjd dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); 2221168404Spjd dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); 2222219089Spjd zfs_sa_upgrade_txholds(tx, zp); 2223219089Spjd zfs_sa_upgrade_txholds(tx, dzp); 2224209962Smm error = dmu_tx_assign(tx, TXG_NOWAIT); 2225168404Spjd if (error) { 2226168404Spjd rw_exit(&zp->z_parent_lock); 2227168404Spjd rw_exit(&zp->z_name_lock); 2228168404Spjd zfs_dirent_unlock(dl); 2229168962Spjd VN_RELE(vp); 2230209962Smm if (error == ERESTART) { 2231168404Spjd dmu_tx_wait(tx); 2232168404Spjd dmu_tx_abort(tx); 2233168404Spjd goto top; 2234168404Spjd } 2235168404Spjd dmu_tx_abort(tx); 2236168404Spjd ZFS_EXIT(zfsvfs); 2237168404Spjd return (error); 2238168404Spjd } 2239168404Spjd 2240168404Spjd#ifdef FREEBSD_NAMECACHE 2241168404Spjd cache_purge(dvp); 2242168404Spjd#endif 2243168404Spjd 2244185029Spjd error = zfs_link_destroy(dl, zp, tx, zflg, NULL); 2245168404Spjd 2246185029Spjd if (error == 0) { 2247185029Spjd uint64_t txtype = TX_RMDIR; 2248185029Spjd if (flags & FIGNORECASE) 2249185029Spjd txtype |= TX_CI; 2250219089Spjd zfs_log_remove(zilog, tx, txtype, dzp, name, ZFS_NO_OBJECT); 2251185029Spjd } 2252168404Spjd 2253168404Spjd dmu_tx_commit(tx); 2254168404Spjd 2255168404Spjd rw_exit(&zp->z_parent_lock); 2256168404Spjd rw_exit(&zp->z_name_lock); 2257168404Spjd#ifdef FREEBSD_NAMECACHE 2258168404Spjd cache_purge(vp); 2259168404Spjd#endif 2260168404Spjdout: 2261168404Spjd zfs_dirent_unlock(dl); 2262168404Spjd 2263168962Spjd VN_RELE(vp); 2264168962Spjd 2265219089Spjd if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) 2266219089Spjd zil_commit(zilog, 0); 2267219089Spjd 2268168404Spjd ZFS_EXIT(zfsvfs); 2269168404Spjd return (error); 2270168404Spjd} 2271168404Spjd 2272168404Spjd/* 2273168404Spjd * Read as many directory entries as will fit into the provided 2274168404Spjd * buffer from the given directory cursor position (specified in 2275168404Spjd * the uio structure. 2276168404Spjd * 2277168404Spjd * IN: vp - vnode of directory to read. 2278168404Spjd * uio - structure supplying read location, range info, 2279168404Spjd * and return buffer. 2280168404Spjd * cr - credentials of caller. 2281185029Spjd * ct - caller context 2282185029Spjd * flags - case flags 2283168404Spjd * 2284168404Spjd * OUT: uio - updated offset and range, buffer filled. 2285168404Spjd * eofp - set to true if end-of-file detected. 2286168404Spjd * 2287168404Spjd * RETURN: 0 if success 2288168404Spjd * error code if failure 2289168404Spjd * 2290168404Spjd * Timestamps: 2291168404Spjd * vp - atime updated 2292168404Spjd * 2293168404Spjd * Note that the low 4 bits of the cookie returned by zap is always zero. 2294168404Spjd * This allows us to use the low range for "special" directory entries: 2295168404Spjd * We use 0 for '.', and 1 for '..'. If this is the root of the filesystem, 2296168404Spjd * we use the offset 2 for the '.zfs' directory. 2297168404Spjd */ 2298168404Spjd/* ARGSUSED */ 2299168404Spjdstatic int 2300168962Spjdzfs_readdir(vnode_t *vp, uio_t *uio, cred_t *cr, int *eofp, int *ncookies, u_long **cookies) 2301168404Spjd{ 2302168404Spjd znode_t *zp = VTOZ(vp); 2303168404Spjd iovec_t *iovp; 2304185029Spjd edirent_t *eodp; 2305168404Spjd dirent64_t *odp; 2306168404Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 2307168404Spjd objset_t *os; 2308168404Spjd caddr_t outbuf; 2309168404Spjd size_t bufsize; 2310168404Spjd zap_cursor_t zc; 2311168404Spjd zap_attribute_t zap; 2312168404Spjd uint_t bytes_wanted; 2313168404Spjd uint64_t offset; /* must be unsigned; checks for < 1 */ 2314219089Spjd uint64_t parent; 2315168404Spjd int local_eof; 2316168404Spjd int outcount; 2317168404Spjd int error; 2318168404Spjd uint8_t prefetch; 2319185029Spjd boolean_t check_sysattrs; 2320168404Spjd uint8_t type; 2321168962Spjd int ncooks; 2322168962Spjd u_long *cooks = NULL; 2323185029Spjd int flags = 0; 2324168404Spjd 2325168404Spjd ZFS_ENTER(zfsvfs); 2326185029Spjd ZFS_VERIFY_ZP(zp); 2327168404Spjd 2328219089Spjd if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs), 2329219089Spjd &parent, sizeof (parent))) != 0) { 2330219089Spjd ZFS_EXIT(zfsvfs); 2331219089Spjd return (error); 2332219089Spjd } 2333219089Spjd 2334168404Spjd /* 2335168404Spjd * If we are not given an eof variable, 2336168404Spjd * use a local one. 2337168404Spjd */ 2338168404Spjd if (eofp == NULL) 2339168404Spjd eofp = &local_eof; 2340168404Spjd 2341168404Spjd /* 2342168404Spjd * Check for valid iov_len. 2343168404Spjd */ 2344168404Spjd if (uio->uio_iov->iov_len <= 0) { 2345168404Spjd ZFS_EXIT(zfsvfs); 2346168404Spjd return (EINVAL); 2347168404Spjd } 2348168404Spjd 2349168404Spjd /* 2350168404Spjd * Quit if directory has been removed (posix) 2351168404Spjd */ 2352168404Spjd if ((*eofp = zp->z_unlinked) != 0) { 2353168404Spjd ZFS_EXIT(zfsvfs); 2354168404Spjd return (0); 2355168404Spjd } 2356168404Spjd 2357168404Spjd error = 0; 2358168404Spjd os = zfsvfs->z_os; 2359168404Spjd offset = uio->uio_loffset; 2360168404Spjd prefetch = zp->z_zn_prefetch; 2361168404Spjd 2362168404Spjd /* 2363168404Spjd * Initialize the iterator cursor. 2364168404Spjd */ 2365168404Spjd if (offset <= 3) { 2366168404Spjd /* 2367168404Spjd * Start iteration from the beginning of the directory. 2368168404Spjd */ 2369168404Spjd zap_cursor_init(&zc, os, zp->z_id); 2370168404Spjd } else { 2371168404Spjd /* 2372168404Spjd * The offset is a serialized cursor. 2373168404Spjd */ 2374168404Spjd zap_cursor_init_serialized(&zc, os, zp->z_id, offset); 2375168404Spjd } 2376168404Spjd 2377168404Spjd /* 2378168404Spjd * Get space to change directory entries into fs independent format. 2379168404Spjd */ 2380168404Spjd iovp = uio->uio_iov; 2381168404Spjd bytes_wanted = iovp->iov_len; 2382168404Spjd if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1) { 2383168404Spjd bufsize = bytes_wanted; 2384168404Spjd outbuf = kmem_alloc(bufsize, KM_SLEEP); 2385168404Spjd odp = (struct dirent64 *)outbuf; 2386168404Spjd } else { 2387168404Spjd bufsize = bytes_wanted; 2388168404Spjd odp = (struct dirent64 *)iovp->iov_base; 2389168404Spjd } 2390185029Spjd eodp = (struct edirent *)odp; 2391168404Spjd 2392169170Spjd if (ncookies != NULL) { 2393168404Spjd /* 2394168404Spjd * Minimum entry size is dirent size and 1 byte for a file name. 2395168404Spjd */ 2396168962Spjd ncooks = uio->uio_resid / (sizeof(struct dirent) - sizeof(((struct dirent *)NULL)->d_name) + 1); 2397219404Spjd cooks = malloc(ncooks * sizeof(u_long), M_TEMP, M_WAITOK); 2398219404Spjd *cookies = cooks; 2399168962Spjd *ncookies = ncooks; 2400168404Spjd } 2401185029Spjd /* 2402185029Spjd * If this VFS supports the system attribute view interface; and 2403185029Spjd * we're looking at an extended attribute directory; and we care 2404185029Spjd * about normalization conflicts on this vfs; then we must check 2405185029Spjd * for normalization conflicts with the sysattr name space. 2406185029Spjd */ 2407185029Spjd#ifdef TODO 2408185029Spjd check_sysattrs = vfs_has_feature(vp->v_vfsp, VFSFT_SYSATTR_VIEWS) && 2409185029Spjd (vp->v_flag & V_XATTRDIR) && zfsvfs->z_norm && 2410185029Spjd (flags & V_RDDIR_ENTFLAGS); 2411185029Spjd#else 2412185029Spjd check_sysattrs = 0; 2413185029Spjd#endif 2414168404Spjd 2415168404Spjd /* 2416168404Spjd * Transform to file-system independent format 2417168404Spjd */ 2418168404Spjd outcount = 0; 2419168404Spjd while (outcount < bytes_wanted) { 2420168404Spjd ino64_t objnum; 2421168404Spjd ushort_t reclen; 2422219089Spjd off64_t *next = NULL; 2423168404Spjd 2424168404Spjd /* 2425168404Spjd * Special case `.', `..', and `.zfs'. 2426168404Spjd */ 2427168404Spjd if (offset == 0) { 2428168404Spjd (void) strcpy(zap.za_name, "."); 2429185029Spjd zap.za_normalization_conflict = 0; 2430168404Spjd objnum = zp->z_id; 2431169108Spjd type = DT_DIR; 2432168404Spjd } else if (offset == 1) { 2433168404Spjd (void) strcpy(zap.za_name, ".."); 2434185029Spjd zap.za_normalization_conflict = 0; 2435219089Spjd objnum = parent; 2436169108Spjd type = DT_DIR; 2437168404Spjd } else if (offset == 2 && zfs_show_ctldir(zp)) { 2438168404Spjd (void) strcpy(zap.za_name, ZFS_CTLDIR_NAME); 2439185029Spjd zap.za_normalization_conflict = 0; 2440168404Spjd objnum = ZFSCTL_INO_ROOT; 2441169108Spjd type = DT_DIR; 2442168404Spjd } else { 2443168404Spjd /* 2444168404Spjd * Grab next entry. 2445168404Spjd */ 2446168404Spjd if (error = zap_cursor_retrieve(&zc, &zap)) { 2447168404Spjd if ((*eofp = (error == ENOENT)) != 0) 2448168404Spjd break; 2449168404Spjd else 2450168404Spjd goto update; 2451168404Spjd } 2452168404Spjd 2453168404Spjd if (zap.za_integer_length != 8 || 2454168404Spjd zap.za_num_integers != 1) { 2455168404Spjd cmn_err(CE_WARN, "zap_readdir: bad directory " 2456168404Spjd "entry, obj = %lld, offset = %lld\n", 2457168404Spjd (u_longlong_t)zp->z_id, 2458168404Spjd (u_longlong_t)offset); 2459168404Spjd error = ENXIO; 2460168404Spjd goto update; 2461168404Spjd } 2462168404Spjd 2463168404Spjd objnum = ZFS_DIRENT_OBJ(zap.za_first_integer); 2464168404Spjd /* 2465168404Spjd * MacOS X can extract the object type here such as: 2466168404Spjd * uint8_t type = ZFS_DIRENT_TYPE(zap.za_first_integer); 2467168404Spjd */ 2468168404Spjd type = ZFS_DIRENT_TYPE(zap.za_first_integer); 2469185029Spjd 2470185029Spjd if (check_sysattrs && !zap.za_normalization_conflict) { 2471185029Spjd#ifdef TODO 2472185029Spjd zap.za_normalization_conflict = 2473185029Spjd xattr_sysattr_casechk(zap.za_name); 2474185029Spjd#else 2475185029Spjd panic("%s:%u: TODO", __func__, __LINE__); 2476185029Spjd#endif 2477185029Spjd } 2478168404Spjd } 2479168404Spjd 2480211932Smm if (flags & V_RDDIR_ACCFILTER) { 2481211932Smm /* 2482211932Smm * If we have no access at all, don't include 2483211932Smm * this entry in the returned information 2484211932Smm */ 2485211932Smm znode_t *ezp; 2486211932Smm if (zfs_zget(zp->z_zfsvfs, objnum, &ezp) != 0) 2487211932Smm goto skip_entry; 2488211932Smm if (!zfs_has_access(ezp, cr)) { 2489211932Smm VN_RELE(ZTOV(ezp)); 2490211932Smm goto skip_entry; 2491211932Smm } 2492211932Smm VN_RELE(ZTOV(ezp)); 2493211932Smm } 2494211932Smm 2495185029Spjd if (flags & V_RDDIR_ENTFLAGS) 2496185029Spjd reclen = EDIRENT_RECLEN(strlen(zap.za_name)); 2497185029Spjd else 2498185029Spjd reclen = DIRENT64_RECLEN(strlen(zap.za_name)); 2499185029Spjd 2500168404Spjd /* 2501168404Spjd * Will this entry fit in the buffer? 2502168404Spjd */ 2503168404Spjd if (outcount + reclen > bufsize) { 2504168404Spjd /* 2505168404Spjd * Did we manage to fit anything in the buffer? 2506168404Spjd */ 2507168404Spjd if (!outcount) { 2508168404Spjd error = EINVAL; 2509168404Spjd goto update; 2510168404Spjd } 2511168404Spjd break; 2512168404Spjd } 2513185029Spjd if (flags & V_RDDIR_ENTFLAGS) { 2514185029Spjd /* 2515185029Spjd * Add extended flag entry: 2516185029Spjd */ 2517185029Spjd eodp->ed_ino = objnum; 2518185029Spjd eodp->ed_reclen = reclen; 2519185029Spjd /* NOTE: ed_off is the offset for the *next* entry */ 2520185029Spjd next = &(eodp->ed_off); 2521185029Spjd eodp->ed_eflags = zap.za_normalization_conflict ? 2522185029Spjd ED_CASE_CONFLICT : 0; 2523185029Spjd (void) strncpy(eodp->ed_name, zap.za_name, 2524185029Spjd EDIRENT_NAMELEN(reclen)); 2525185029Spjd eodp = (edirent_t *)((intptr_t)eodp + reclen); 2526185029Spjd } else { 2527185029Spjd /* 2528185029Spjd * Add normal entry: 2529185029Spjd */ 2530185029Spjd odp->d_ino = objnum; 2531185029Spjd odp->d_reclen = reclen; 2532185029Spjd odp->d_namlen = strlen(zap.za_name); 2533185029Spjd (void) strlcpy(odp->d_name, zap.za_name, odp->d_namlen + 1); 2534185029Spjd odp->d_type = type; 2535185029Spjd odp = (dirent64_t *)((intptr_t)odp + reclen); 2536185029Spjd } 2537168404Spjd outcount += reclen; 2538168404Spjd 2539168404Spjd ASSERT(outcount <= bufsize); 2540168404Spjd 2541168404Spjd /* Prefetch znode */ 2542168404Spjd if (prefetch) 2543168404Spjd dmu_prefetch(os, objnum, 0, 0); 2544168404Spjd 2545211932Smm skip_entry: 2546168404Spjd /* 2547168404Spjd * Move to the next entry, fill in the previous offset. 2548168404Spjd */ 2549168404Spjd if (offset > 2 || (offset == 2 && !zfs_show_ctldir(zp))) { 2550168404Spjd zap_cursor_advance(&zc); 2551168404Spjd offset = zap_cursor_serialize(&zc); 2552168404Spjd } else { 2553168404Spjd offset += 1; 2554168404Spjd } 2555219404Spjd 2556219404Spjd if (cooks != NULL) { 2557219404Spjd *cooks++ = offset; 2558219404Spjd ncooks--; 2559219404Spjd KASSERT(ncooks >= 0, ("ncookies=%d", ncooks)); 2560219404Spjd } 2561168404Spjd } 2562168404Spjd zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */ 2563168404Spjd 2564168404Spjd /* Subtract unused cookies */ 2565168962Spjd if (ncookies != NULL) 2566168962Spjd *ncookies -= ncooks; 2567168404Spjd 2568168404Spjd if (uio->uio_segflg == UIO_SYSSPACE && uio->uio_iovcnt == 1) { 2569168404Spjd iovp->iov_base += outcount; 2570168404Spjd iovp->iov_len -= outcount; 2571168404Spjd uio->uio_resid -= outcount; 2572168404Spjd } else if (error = uiomove(outbuf, (long)outcount, UIO_READ, uio)) { 2573168404Spjd /* 2574168404Spjd * Reset the pointer. 2575168404Spjd */ 2576168404Spjd offset = uio->uio_loffset; 2577168404Spjd } 2578168404Spjd 2579168404Spjdupdate: 2580168404Spjd zap_cursor_fini(&zc); 2581168404Spjd if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1) 2582168404Spjd kmem_free(outbuf, bufsize); 2583168404Spjd 2584168404Spjd if (error == ENOENT) 2585168404Spjd error = 0; 2586168404Spjd 2587168404Spjd ZFS_ACCESSTIME_STAMP(zfsvfs, zp); 2588168404Spjd 2589168404Spjd uio->uio_loffset = offset; 2590168404Spjd ZFS_EXIT(zfsvfs); 2591169107Spjd if (error != 0 && cookies != NULL) { 2592168962Spjd free(*cookies, M_TEMP); 2593168962Spjd *cookies = NULL; 2594168962Spjd *ncookies = 0; 2595168404Spjd } 2596168404Spjd return (error); 2597168404Spjd} 2598168404Spjd 2599185029Spjdulong_t zfs_fsync_sync_cnt = 4; 2600185029Spjd 2601168404Spjdstatic int 2602185029Spjdzfs_fsync(vnode_t *vp, int syncflag, cred_t *cr, caller_context_t *ct) 2603168404Spjd{ 2604168962Spjd znode_t *zp = VTOZ(vp); 2605168962Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 2606168404Spjd 2607185029Spjd (void) tsd_set(zfs_fsyncer_key, (void *)zfs_fsync_sync_cnt); 2608185029Spjd 2609219089Spjd if (zfsvfs->z_os->os_sync != ZFS_SYNC_DISABLED) { 2610219089Spjd ZFS_ENTER(zfsvfs); 2611219089Spjd ZFS_VERIFY_ZP(zp); 2612219089Spjd zil_commit(zfsvfs->z_log, zp->z_id); 2613219089Spjd ZFS_EXIT(zfsvfs); 2614219089Spjd } 2615168404Spjd return (0); 2616168404Spjd} 2617168404Spjd 2618185029Spjd 2619168404Spjd/* 2620168404Spjd * Get the requested file attributes and place them in the provided 2621168404Spjd * vattr structure. 2622168404Spjd * 2623168404Spjd * IN: vp - vnode of file. 2624168404Spjd * vap - va_mask identifies requested attributes. 2625185029Spjd * If AT_XVATTR set, then optional attrs are requested 2626185029Spjd * flags - ATTR_NOACLCHECK (CIFS server context) 2627168404Spjd * cr - credentials of caller. 2628185029Spjd * ct - caller context 2629168404Spjd * 2630168404Spjd * OUT: vap - attribute values. 2631168404Spjd * 2632168404Spjd * RETURN: 0 (always succeeds) 2633168404Spjd */ 2634168404Spjd/* ARGSUSED */ 2635168404Spjdstatic int 2636185029Spjdzfs_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr, 2637185029Spjd caller_context_t *ct) 2638168404Spjd{ 2639168962Spjd znode_t *zp = VTOZ(vp); 2640168962Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 2641185029Spjd int error = 0; 2642168962Spjd uint32_t blksize; 2643168962Spjd u_longlong_t nblocks; 2644185029Spjd uint64_t links; 2645224251Sdelphij uint64_t mtime[2], ctime[2], crtime[2], rdev; 2646185029Spjd xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */ 2647185029Spjd xoptattr_t *xoap = NULL; 2648185029Spjd boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; 2649224251Sdelphij sa_bulk_attr_t bulk[4]; 2650219089Spjd int count = 0; 2651168404Spjd 2652168404Spjd ZFS_ENTER(zfsvfs); 2653185029Spjd ZFS_VERIFY_ZP(zp); 2654168404Spjd 2655219089Spjd zfs_fuid_map_ids(zp, cr, &vap->va_uid, &vap->va_gid); 2656219089Spjd 2657219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, &mtime, 16); 2658219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &ctime, 16); 2659219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, &crtime, 16); 2660224251Sdelphij if (vp->v_type == VBLK || vp->v_type == VCHR) 2661224251Sdelphij SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_RDEV(zfsvfs), NULL, 2662224251Sdelphij &rdev, 8); 2663219089Spjd 2664219089Spjd if ((error = sa_bulk_lookup(zp->z_sa_hdl, bulk, count)) != 0) { 2665219089Spjd ZFS_EXIT(zfsvfs); 2666219089Spjd return (error); 2667219089Spjd } 2668219089Spjd 2669168404Spjd /* 2670185029Spjd * If ACL is trivial don't bother looking for ACE_READ_ATTRIBUTES. 2671185029Spjd * Also, if we are the owner don't bother, since owner should 2672185029Spjd * always be allowed to read basic attributes of file. 2673185029Spjd */ 2674219089Spjd if (!(zp->z_pflags & ZFS_ACL_TRIVIAL) && 2675219089Spjd (vap->va_uid != crgetuid(cr))) { 2676185029Spjd if (error = zfs_zaccess(zp, ACE_READ_ATTRIBUTES, 0, 2677185029Spjd skipaclchk, cr)) { 2678185029Spjd ZFS_EXIT(zfsvfs); 2679185029Spjd return (error); 2680185029Spjd } 2681185029Spjd } 2682185029Spjd 2683185029Spjd /* 2684168404Spjd * Return all attributes. It's cheaper to provide the answer 2685168404Spjd * than to determine whether we were asked the question. 2686168404Spjd */ 2687168404Spjd 2688209097Smm mutex_enter(&zp->z_lock); 2689219089Spjd vap->va_type = IFTOVT(zp->z_mode); 2690219089Spjd vap->va_mode = zp->z_mode & ~S_IFMT; 2691185029Spjd// vap->va_fsid = zp->z_zfsvfs->z_vfs->vfs_dev; 2692168404Spjd vap->va_nodeid = zp->z_id; 2693185029Spjd if ((vp->v_flag & VROOT) && zfs_show_ctldir(zp)) 2694219089Spjd links = zp->z_links + 1; 2695185029Spjd else 2696219089Spjd links = zp->z_links; 2697185029Spjd vap->va_nlink = MIN(links, UINT32_MAX); /* nlink_t limit! */ 2698219089Spjd vap->va_size = zp->z_size; 2699168404Spjd vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0]; 2700224251Sdelphij if (vp->v_type == VBLK || vp->v_type == VCHR) 2701224251Sdelphij vap->va_rdev = zfs_cmpldev(rdev); 2702168404Spjd vap->va_seq = zp->z_seq; 2703168404Spjd vap->va_flags = 0; /* FreeBSD: Reset chflags(2) flags. */ 2704168404Spjd 2705185029Spjd /* 2706185029Spjd * Add in any requested optional attributes and the create time. 2707185029Spjd * Also set the corresponding bits in the returned attribute bitmap. 2708185029Spjd */ 2709185029Spjd if ((xoap = xva_getxoptattr(xvap)) != NULL && zfsvfs->z_use_fuids) { 2710185029Spjd if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) { 2711185029Spjd xoap->xoa_archive = 2712219089Spjd ((zp->z_pflags & ZFS_ARCHIVE) != 0); 2713185029Spjd XVA_SET_RTN(xvap, XAT_ARCHIVE); 2714185029Spjd } 2715185029Spjd 2716185029Spjd if (XVA_ISSET_REQ(xvap, XAT_READONLY)) { 2717185029Spjd xoap->xoa_readonly = 2718219089Spjd ((zp->z_pflags & ZFS_READONLY) != 0); 2719185029Spjd XVA_SET_RTN(xvap, XAT_READONLY); 2720185029Spjd } 2721185029Spjd 2722185029Spjd if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) { 2723185029Spjd xoap->xoa_system = 2724219089Spjd ((zp->z_pflags & ZFS_SYSTEM) != 0); 2725185029Spjd XVA_SET_RTN(xvap, XAT_SYSTEM); 2726185029Spjd } 2727185029Spjd 2728185029Spjd if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) { 2729185029Spjd xoap->xoa_hidden = 2730219089Spjd ((zp->z_pflags & ZFS_HIDDEN) != 0); 2731185029Spjd XVA_SET_RTN(xvap, XAT_HIDDEN); 2732185029Spjd } 2733185029Spjd 2734185029Spjd if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { 2735185029Spjd xoap->xoa_nounlink = 2736219089Spjd ((zp->z_pflags & ZFS_NOUNLINK) != 0); 2737185029Spjd XVA_SET_RTN(xvap, XAT_NOUNLINK); 2738185029Spjd } 2739185029Spjd 2740185029Spjd if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { 2741185029Spjd xoap->xoa_immutable = 2742219089Spjd ((zp->z_pflags & ZFS_IMMUTABLE) != 0); 2743185029Spjd XVA_SET_RTN(xvap, XAT_IMMUTABLE); 2744185029Spjd } 2745185029Spjd 2746185029Spjd if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { 2747185029Spjd xoap->xoa_appendonly = 2748219089Spjd ((zp->z_pflags & ZFS_APPENDONLY) != 0); 2749185029Spjd XVA_SET_RTN(xvap, XAT_APPENDONLY); 2750185029Spjd } 2751185029Spjd 2752185029Spjd if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { 2753185029Spjd xoap->xoa_nodump = 2754219089Spjd ((zp->z_pflags & ZFS_NODUMP) != 0); 2755185029Spjd XVA_SET_RTN(xvap, XAT_NODUMP); 2756185029Spjd } 2757185029Spjd 2758185029Spjd if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) { 2759185029Spjd xoap->xoa_opaque = 2760219089Spjd ((zp->z_pflags & ZFS_OPAQUE) != 0); 2761185029Spjd XVA_SET_RTN(xvap, XAT_OPAQUE); 2762185029Spjd } 2763185029Spjd 2764185029Spjd if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { 2765185029Spjd xoap->xoa_av_quarantined = 2766219089Spjd ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0); 2767185029Spjd XVA_SET_RTN(xvap, XAT_AV_QUARANTINED); 2768185029Spjd } 2769185029Spjd 2770185029Spjd if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { 2771185029Spjd xoap->xoa_av_modified = 2772219089Spjd ((zp->z_pflags & ZFS_AV_MODIFIED) != 0); 2773185029Spjd XVA_SET_RTN(xvap, XAT_AV_MODIFIED); 2774185029Spjd } 2775185029Spjd 2776185029Spjd if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) && 2777219089Spjd vp->v_type == VREG) { 2778219089Spjd zfs_sa_get_scanstamp(zp, xvap); 2779185029Spjd } 2780185029Spjd 2781185029Spjd if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) { 2782219089Spjd uint64_t times[2]; 2783219089Spjd 2784219089Spjd (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_CRTIME(zfsvfs), 2785219089Spjd times, sizeof (times)); 2786219089Spjd ZFS_TIME_DECODE(&xoap->xoa_createtime, times); 2787185029Spjd XVA_SET_RTN(xvap, XAT_CREATETIME); 2788185029Spjd } 2789219089Spjd 2790219089Spjd if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { 2791219089Spjd xoap->xoa_reparse = ((zp->z_pflags & ZFS_REPARSE) != 0); 2792219089Spjd XVA_SET_RTN(xvap, XAT_REPARSE); 2793219089Spjd } 2794219089Spjd if (XVA_ISSET_REQ(xvap, XAT_GEN)) { 2795219089Spjd xoap->xoa_generation = zp->z_gen; 2796219089Spjd XVA_SET_RTN(xvap, XAT_GEN); 2797219089Spjd } 2798219089Spjd 2799219089Spjd if (XVA_ISSET_REQ(xvap, XAT_OFFLINE)) { 2800219089Spjd xoap->xoa_offline = 2801219089Spjd ((zp->z_pflags & ZFS_OFFLINE) != 0); 2802219089Spjd XVA_SET_RTN(xvap, XAT_OFFLINE); 2803219089Spjd } 2804219089Spjd 2805219089Spjd if (XVA_ISSET_REQ(xvap, XAT_SPARSE)) { 2806219089Spjd xoap->xoa_sparse = 2807219089Spjd ((zp->z_pflags & ZFS_SPARSE) != 0); 2808219089Spjd XVA_SET_RTN(xvap, XAT_SPARSE); 2809219089Spjd } 2810185029Spjd } 2811185029Spjd 2812219089Spjd ZFS_TIME_DECODE(&vap->va_atime, zp->z_atime); 2813219089Spjd ZFS_TIME_DECODE(&vap->va_mtime, mtime); 2814219089Spjd ZFS_TIME_DECODE(&vap->va_ctime, ctime); 2815219089Spjd ZFS_TIME_DECODE(&vap->va_birthtime, crtime); 2816168404Spjd 2817168404Spjd mutex_exit(&zp->z_lock); 2818168404Spjd 2819219089Spjd sa_object_size(zp->z_sa_hdl, &blksize, &nblocks); 2820168404Spjd vap->va_blksize = blksize; 2821168404Spjd vap->va_bytes = nblocks << 9; /* nblocks * 512 */ 2822168404Spjd 2823168404Spjd if (zp->z_blksz == 0) { 2824168404Spjd /* 2825168404Spjd * Block size hasn't been set; suggest maximal I/O transfers. 2826168404Spjd */ 2827168404Spjd vap->va_blksize = zfsvfs->z_max_blksz; 2828168404Spjd } 2829168404Spjd 2830168404Spjd ZFS_EXIT(zfsvfs); 2831168404Spjd return (0); 2832168404Spjd} 2833168404Spjd 2834168404Spjd/* 2835168404Spjd * Set the file attributes to the values contained in the 2836168404Spjd * vattr structure. 2837168404Spjd * 2838168404Spjd * IN: vp - vnode of file to be modified. 2839168404Spjd * vap - new attribute values. 2840185029Spjd * If AT_XVATTR set, then optional attrs are being set 2841168404Spjd * flags - ATTR_UTIME set if non-default time values provided. 2842185029Spjd * - ATTR_NOACLCHECK (CIFS context only). 2843168404Spjd * cr - credentials of caller. 2844185029Spjd * ct - caller context 2845168404Spjd * 2846168404Spjd * RETURN: 0 if success 2847168404Spjd * error code if failure 2848168404Spjd * 2849168404Spjd * Timestamps: 2850168404Spjd * vp - ctime updated, mtime updated if size changed. 2851168404Spjd */ 2852168404Spjd/* ARGSUSED */ 2853168404Spjdstatic int 2854168962Spjdzfs_setattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr, 2855168962Spjd caller_context_t *ct) 2856168404Spjd{ 2857185029Spjd znode_t *zp = VTOZ(vp); 2858168404Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 2859185029Spjd zilog_t *zilog; 2860168404Spjd dmu_tx_t *tx; 2861168404Spjd vattr_t oldva; 2862209962Smm xvattr_t tmpxvattr; 2863168962Spjd uint_t mask = vap->va_mask; 2864168404Spjd uint_t saved_mask; 2865197831Spjd uint64_t saved_mode; 2866168404Spjd int trim_mask = 0; 2867168404Spjd uint64_t new_mode; 2868209962Smm uint64_t new_uid, new_gid; 2869219089Spjd uint64_t xattr_obj; 2870219089Spjd uint64_t mtime[2], ctime[2]; 2871168404Spjd znode_t *attrzp; 2872168404Spjd int need_policy = FALSE; 2873219089Spjd int err, err2; 2874185029Spjd zfs_fuid_info_t *fuidp = NULL; 2875185029Spjd xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */ 2876185029Spjd xoptattr_t *xoap; 2877219089Spjd zfs_acl_t *aclp; 2878185029Spjd boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; 2879219089Spjd boolean_t fuid_dirtied = B_FALSE; 2880219089Spjd sa_bulk_attr_t bulk[7], xattr_bulk[7]; 2881219089Spjd int count = 0, xattr_count = 0; 2882168404Spjd 2883168404Spjd if (mask == 0) 2884168404Spjd return (0); 2885168404Spjd 2886168962Spjd if (mask & AT_NOSET) 2887168962Spjd return (EINVAL); 2888168962Spjd 2889185029Spjd ZFS_ENTER(zfsvfs); 2890185029Spjd ZFS_VERIFY_ZP(zp); 2891185029Spjd 2892185029Spjd zilog = zfsvfs->z_log; 2893185029Spjd 2894185029Spjd /* 2895185029Spjd * Make sure that if we have ephemeral uid/gid or xvattr specified 2896185029Spjd * that file system is at proper version level 2897185029Spjd */ 2898185029Spjd 2899185029Spjd if (zfsvfs->z_use_fuids == B_FALSE && 2900185029Spjd (((mask & AT_UID) && IS_EPHEMERAL(vap->va_uid)) || 2901185029Spjd ((mask & AT_GID) && IS_EPHEMERAL(vap->va_gid)) || 2902185029Spjd (mask & AT_XVATTR))) { 2903185029Spjd ZFS_EXIT(zfsvfs); 2904185029Spjd return (EINVAL); 2905185029Spjd } 2906185029Spjd 2907185029Spjd if (mask & AT_SIZE && vp->v_type == VDIR) { 2908185029Spjd ZFS_EXIT(zfsvfs); 2909168404Spjd return (EISDIR); 2910185029Spjd } 2911168404Spjd 2912185029Spjd if (mask & AT_SIZE && vp->v_type != VREG && vp->v_type != VFIFO) { 2913185029Spjd ZFS_EXIT(zfsvfs); 2914168404Spjd return (EINVAL); 2915185029Spjd } 2916168404Spjd 2917185029Spjd /* 2918185029Spjd * If this is an xvattr_t, then get a pointer to the structure of 2919185029Spjd * optional attributes. If this is NULL, then we have a vattr_t. 2920185029Spjd */ 2921185029Spjd xoap = xva_getxoptattr(xvap); 2922168404Spjd 2923209962Smm xva_init(&tmpxvattr); 2924209962Smm 2925185029Spjd /* 2926185029Spjd * Immutable files can only alter immutable bit and atime 2927185029Spjd */ 2928219089Spjd if ((zp->z_pflags & ZFS_IMMUTABLE) && 2929185029Spjd ((mask & (AT_SIZE|AT_UID|AT_GID|AT_MTIME|AT_MODE)) || 2930185029Spjd ((mask & AT_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) { 2931185029Spjd ZFS_EXIT(zfsvfs); 2932185029Spjd return (EPERM); 2933185029Spjd } 2934185029Spjd 2935219089Spjd if ((mask & AT_SIZE) && (zp->z_pflags & ZFS_READONLY)) { 2936185029Spjd ZFS_EXIT(zfsvfs); 2937185029Spjd return (EPERM); 2938185029Spjd } 2939185029Spjd 2940185029Spjd /* 2941185029Spjd * Verify timestamps doesn't overflow 32 bits. 2942185029Spjd * ZFS can handle large timestamps, but 32bit syscalls can't 2943185029Spjd * handle times greater than 2039. This check should be removed 2944185029Spjd * once large timestamps are fully supported. 2945185029Spjd */ 2946185029Spjd if (mask & (AT_ATIME | AT_MTIME)) { 2947185029Spjd if (((mask & AT_ATIME) && TIMESPEC_OVERFLOW(&vap->va_atime)) || 2948185029Spjd ((mask & AT_MTIME) && TIMESPEC_OVERFLOW(&vap->va_mtime))) { 2949185029Spjd ZFS_EXIT(zfsvfs); 2950185029Spjd return (EOVERFLOW); 2951185029Spjd } 2952185029Spjd } 2953185029Spjd 2954168404Spjdtop: 2955168404Spjd attrzp = NULL; 2956219089Spjd aclp = NULL; 2957168404Spjd 2958211932Smm /* Can this be moved to before the top label? */ 2959168404Spjd if (zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) { 2960168404Spjd ZFS_EXIT(zfsvfs); 2961168404Spjd return (EROFS); 2962168404Spjd } 2963168404Spjd 2964168404Spjd /* 2965168404Spjd * First validate permissions 2966168404Spjd */ 2967168404Spjd 2968168404Spjd if (mask & AT_SIZE) { 2969168404Spjd /* 2970168404Spjd * XXX - Note, we are not providing any open 2971168404Spjd * mode flags here (like FNDELAY), so we may 2972168404Spjd * block if there are locks present... this 2973168404Spjd * should be addressed in openat(). 2974168404Spjd */ 2975185029Spjd /* XXX - would it be OK to generate a log record here? */ 2976185029Spjd err = zfs_freesp(zp, vap->va_size, 0, 0, FALSE); 2977168404Spjd if (err) { 2978168404Spjd ZFS_EXIT(zfsvfs); 2979168404Spjd return (err); 2980168404Spjd } 2981168404Spjd } 2982168404Spjd 2983185029Spjd if (mask & (AT_ATIME|AT_MTIME) || 2984185029Spjd ((mask & AT_XVATTR) && (XVA_ISSET_REQ(xvap, XAT_HIDDEN) || 2985185029Spjd XVA_ISSET_REQ(xvap, XAT_READONLY) || 2986185029Spjd XVA_ISSET_REQ(xvap, XAT_ARCHIVE) || 2987219089Spjd XVA_ISSET_REQ(xvap, XAT_OFFLINE) || 2988219089Spjd XVA_ISSET_REQ(xvap, XAT_SPARSE) || 2989185029Spjd XVA_ISSET_REQ(xvap, XAT_CREATETIME) || 2990219089Spjd XVA_ISSET_REQ(xvap, XAT_SYSTEM)))) { 2991185029Spjd need_policy = zfs_zaccess(zp, ACE_WRITE_ATTRIBUTES, 0, 2992185029Spjd skipaclchk, cr); 2993219089Spjd } 2994168404Spjd 2995168404Spjd if (mask & (AT_UID|AT_GID)) { 2996168404Spjd int idmask = (mask & (AT_UID|AT_GID)); 2997168404Spjd int take_owner; 2998168404Spjd int take_group; 2999168404Spjd 3000168404Spjd /* 3001168404Spjd * NOTE: even if a new mode is being set, 3002168404Spjd * we may clear S_ISUID/S_ISGID bits. 3003168404Spjd */ 3004168404Spjd 3005168404Spjd if (!(mask & AT_MODE)) 3006219089Spjd vap->va_mode = zp->z_mode; 3007168404Spjd 3008168404Spjd /* 3009168404Spjd * Take ownership or chgrp to group we are a member of 3010168404Spjd */ 3011168404Spjd 3012168404Spjd take_owner = (mask & AT_UID) && (vap->va_uid == crgetuid(cr)); 3013185029Spjd take_group = (mask & AT_GID) && 3014185029Spjd zfs_groupmember(zfsvfs, vap->va_gid, cr); 3015168404Spjd 3016168404Spjd /* 3017168404Spjd * If both AT_UID and AT_GID are set then take_owner and 3018168404Spjd * take_group must both be set in order to allow taking 3019168404Spjd * ownership. 3020168404Spjd * 3021168404Spjd * Otherwise, send the check through secpolicy_vnode_setattr() 3022168404Spjd * 3023168404Spjd */ 3024168404Spjd 3025168404Spjd if (((idmask == (AT_UID|AT_GID)) && take_owner && take_group) || 3026168404Spjd ((idmask == AT_UID) && take_owner) || 3027168404Spjd ((idmask == AT_GID) && take_group)) { 3028185029Spjd if (zfs_zaccess(zp, ACE_WRITE_OWNER, 0, 3029185029Spjd skipaclchk, cr) == 0) { 3030168404Spjd /* 3031168404Spjd * Remove setuid/setgid for non-privileged users 3032168404Spjd */ 3033185029Spjd secpolicy_setid_clear(vap, vp, cr); 3034168404Spjd trim_mask = (mask & (AT_UID|AT_GID)); 3035168404Spjd } else { 3036168404Spjd need_policy = TRUE; 3037168404Spjd } 3038168404Spjd } else { 3039168404Spjd need_policy = TRUE; 3040168404Spjd } 3041168404Spjd } 3042168404Spjd 3043168404Spjd mutex_enter(&zp->z_lock); 3044219089Spjd oldva.va_mode = zp->z_mode; 3045185029Spjd zfs_fuid_map_ids(zp, cr, &oldva.va_uid, &oldva.va_gid); 3046185029Spjd if (mask & AT_XVATTR) { 3047209962Smm /* 3048209962Smm * Update xvattr mask to include only those attributes 3049209962Smm * that are actually changing. 3050209962Smm * 3051209962Smm * the bits will be restored prior to actually setting 3052209962Smm * the attributes so the caller thinks they were set. 3053209962Smm */ 3054209962Smm if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { 3055209962Smm if (xoap->xoa_appendonly != 3056219089Spjd ((zp->z_pflags & ZFS_APPENDONLY) != 0)) { 3057209962Smm need_policy = TRUE; 3058209962Smm } else { 3059209962Smm XVA_CLR_REQ(xvap, XAT_APPENDONLY); 3060209962Smm XVA_SET_REQ(&tmpxvattr, XAT_APPENDONLY); 3061209962Smm } 3062209962Smm } 3063209962Smm 3064209962Smm if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { 3065209962Smm if (xoap->xoa_nounlink != 3066219089Spjd ((zp->z_pflags & ZFS_NOUNLINK) != 0)) { 3067209962Smm need_policy = TRUE; 3068209962Smm } else { 3069209962Smm XVA_CLR_REQ(xvap, XAT_NOUNLINK); 3070209962Smm XVA_SET_REQ(&tmpxvattr, XAT_NOUNLINK); 3071209962Smm } 3072209962Smm } 3073209962Smm 3074209962Smm if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { 3075209962Smm if (xoap->xoa_immutable != 3076219089Spjd ((zp->z_pflags & ZFS_IMMUTABLE) != 0)) { 3077209962Smm need_policy = TRUE; 3078209962Smm } else { 3079209962Smm XVA_CLR_REQ(xvap, XAT_IMMUTABLE); 3080209962Smm XVA_SET_REQ(&tmpxvattr, XAT_IMMUTABLE); 3081209962Smm } 3082209962Smm } 3083209962Smm 3084209962Smm if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { 3085209962Smm if (xoap->xoa_nodump != 3086219089Spjd ((zp->z_pflags & ZFS_NODUMP) != 0)) { 3087209962Smm need_policy = TRUE; 3088209962Smm } else { 3089209962Smm XVA_CLR_REQ(xvap, XAT_NODUMP); 3090209962Smm XVA_SET_REQ(&tmpxvattr, XAT_NODUMP); 3091209962Smm } 3092209962Smm } 3093209962Smm 3094209962Smm if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { 3095209962Smm if (xoap->xoa_av_modified != 3096219089Spjd ((zp->z_pflags & ZFS_AV_MODIFIED) != 0)) { 3097209962Smm need_policy = TRUE; 3098209962Smm } else { 3099209962Smm XVA_CLR_REQ(xvap, XAT_AV_MODIFIED); 3100209962Smm XVA_SET_REQ(&tmpxvattr, XAT_AV_MODIFIED); 3101209962Smm } 3102209962Smm } 3103209962Smm 3104209962Smm if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { 3105209962Smm if ((vp->v_type != VREG && 3106209962Smm xoap->xoa_av_quarantined) || 3107209962Smm xoap->xoa_av_quarantined != 3108219089Spjd ((zp->z_pflags & ZFS_AV_QUARANTINED) != 0)) { 3109209962Smm need_policy = TRUE; 3110209962Smm } else { 3111209962Smm XVA_CLR_REQ(xvap, XAT_AV_QUARANTINED); 3112209962Smm XVA_SET_REQ(&tmpxvattr, XAT_AV_QUARANTINED); 3113209962Smm } 3114209962Smm } 3115209962Smm 3116219089Spjd if (XVA_ISSET_REQ(xvap, XAT_REPARSE)) { 3117219089Spjd mutex_exit(&zp->z_lock); 3118219089Spjd ZFS_EXIT(zfsvfs); 3119219089Spjd return (EPERM); 3120219089Spjd } 3121219089Spjd 3122209962Smm if (need_policy == FALSE && 3123209962Smm (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) || 3124209962Smm XVA_ISSET_REQ(xvap, XAT_OPAQUE))) { 3125185029Spjd need_policy = TRUE; 3126185029Spjd } 3127185029Spjd } 3128185029Spjd 3129168404Spjd mutex_exit(&zp->z_lock); 3130168404Spjd 3131168404Spjd if (mask & AT_MODE) { 3132185029Spjd if (zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr) == 0) { 3133168962Spjd err = secpolicy_setid_setsticky_clear(vp, vap, 3134168962Spjd &oldva, cr); 3135168962Spjd if (err) { 3136168962Spjd ZFS_EXIT(zfsvfs); 3137168962Spjd return (err); 3138168962Spjd } 3139168404Spjd trim_mask |= AT_MODE; 3140168404Spjd } else { 3141168404Spjd need_policy = TRUE; 3142168404Spjd } 3143168404Spjd } 3144168404Spjd 3145168404Spjd if (need_policy) { 3146168404Spjd /* 3147168404Spjd * If trim_mask is set then take ownership 3148168404Spjd * has been granted or write_acl is present and user 3149168404Spjd * has the ability to modify mode. In that case remove 3150168404Spjd * UID|GID and or MODE from mask so that 3151168404Spjd * secpolicy_vnode_setattr() doesn't revoke it. 3152168404Spjd */ 3153168404Spjd 3154168404Spjd if (trim_mask) { 3155168404Spjd saved_mask = vap->va_mask; 3156168404Spjd vap->va_mask &= ~trim_mask; 3157197831Spjd if (trim_mask & AT_MODE) { 3158197831Spjd /* 3159197831Spjd * Save the mode, as secpolicy_vnode_setattr() 3160197831Spjd * will overwrite it with ova.va_mode. 3161197831Spjd */ 3162197831Spjd saved_mode = vap->va_mode; 3163197831Spjd } 3164168404Spjd } 3165168404Spjd err = secpolicy_vnode_setattr(cr, vp, vap, &oldva, flags, 3166185029Spjd (int (*)(void *, int, cred_t *))zfs_zaccess_unix, zp); 3167168404Spjd if (err) { 3168168404Spjd ZFS_EXIT(zfsvfs); 3169168404Spjd return (err); 3170168404Spjd } 3171168404Spjd 3172197831Spjd if (trim_mask) { 3173168404Spjd vap->va_mask |= saved_mask; 3174197831Spjd if (trim_mask & AT_MODE) { 3175197831Spjd /* 3176197831Spjd * Recover the mode after 3177197831Spjd * secpolicy_vnode_setattr(). 3178197831Spjd */ 3179197831Spjd vap->va_mode = saved_mode; 3180197831Spjd } 3181197831Spjd } 3182168404Spjd } 3183168404Spjd 3184168404Spjd /* 3185168404Spjd * secpolicy_vnode_setattr, or take ownership may have 3186168404Spjd * changed va_mask 3187168404Spjd */ 3188168404Spjd mask = vap->va_mask; 3189168404Spjd 3190219089Spjd if ((mask & (AT_UID | AT_GID))) { 3191219089Spjd err = sa_lookup(zp->z_sa_hdl, SA_ZPL_XATTR(zfsvfs), 3192219089Spjd &xattr_obj, sizeof (xattr_obj)); 3193168404Spjd 3194219089Spjd if (err == 0 && xattr_obj) { 3195219089Spjd err = zfs_zget(zp->z_zfsvfs, xattr_obj, &attrzp); 3196209962Smm if (err) 3197219089Spjd goto out2; 3198168404Spjd } 3199209962Smm if (mask & AT_UID) { 3200209962Smm new_uid = zfs_fuid_create(zfsvfs, 3201209962Smm (uint64_t)vap->va_uid, cr, ZFS_OWNER, &fuidp); 3202219089Spjd if (new_uid != zp->z_uid && 3203219089Spjd zfs_fuid_overquota(zfsvfs, B_FALSE, new_uid)) { 3204219089Spjd if (attrzp) 3205219089Spjd VN_RELE(ZTOV(attrzp)); 3206209962Smm err = EDQUOT; 3207219089Spjd goto out2; 3208209962Smm } 3209209962Smm } 3210209962Smm 3211209962Smm if (mask & AT_GID) { 3212209962Smm new_gid = zfs_fuid_create(zfsvfs, (uint64_t)vap->va_gid, 3213209962Smm cr, ZFS_GROUP, &fuidp); 3214219089Spjd if (new_gid != zp->z_gid && 3215219089Spjd zfs_fuid_overquota(zfsvfs, B_TRUE, new_gid)) { 3216219089Spjd if (attrzp) 3217219089Spjd VN_RELE(ZTOV(attrzp)); 3218209962Smm err = EDQUOT; 3219219089Spjd goto out2; 3220209962Smm } 3221209962Smm } 3222219089Spjd } 3223219089Spjd tx = dmu_tx_create(zfsvfs->z_os); 3224219089Spjd 3225219089Spjd if (mask & AT_MODE) { 3226219089Spjd uint64_t pmode = zp->z_mode; 3227219089Spjd uint64_t acl_obj; 3228219089Spjd new_mode = (pmode & S_IFMT) | (vap->va_mode & ~S_IFMT); 3229219089Spjd 3230224174Smm if (err = zfs_acl_chmod_setattr(zp, &aclp, new_mode)) 3231224174Smm goto out; 3232219089Spjd 3233219089Spjd mutex_enter(&zp->z_lock); 3234219089Spjd if (!zp->z_is_sa && ((acl_obj = zfs_external_acl(zp)) != 0)) { 3235219089Spjd /* 3236219089Spjd * Are we upgrading ACL from old V0 format 3237219089Spjd * to V1 format? 3238219089Spjd */ 3239219089Spjd if (zfsvfs->z_version >= ZPL_VERSION_FUID && 3240219089Spjd zfs_znode_acl_version(zp) == 3241219089Spjd ZFS_ACL_VERSION_INITIAL) { 3242219089Spjd dmu_tx_hold_free(tx, acl_obj, 0, 3243219089Spjd DMU_OBJECT_END); 3244219089Spjd dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 3245219089Spjd 0, aclp->z_acl_bytes); 3246209962Smm } else { 3247219089Spjd dmu_tx_hold_write(tx, acl_obj, 0, 3248219089Spjd aclp->z_acl_bytes); 3249209962Smm } 3250219089Spjd } else if (!zp->z_is_sa && aclp->z_acl_bytes > ZFS_ACE_SPACE) { 3251219089Spjd dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 3252219089Spjd 0, aclp->z_acl_bytes); 3253209962Smm } 3254219089Spjd mutex_exit(&zp->z_lock); 3255219089Spjd dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); 3256219089Spjd } else { 3257219089Spjd if ((mask & AT_XVATTR) && 3258219089Spjd XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) 3259219089Spjd dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_TRUE); 3260219089Spjd else 3261219089Spjd dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); 3262168404Spjd } 3263168404Spjd 3264219089Spjd if (attrzp) { 3265219089Spjd dmu_tx_hold_sa(tx, attrzp->z_sa_hdl, B_FALSE); 3266219089Spjd } 3267219089Spjd 3268219089Spjd fuid_dirtied = zfsvfs->z_fuid_dirty; 3269219089Spjd if (fuid_dirtied) 3270219089Spjd zfs_fuid_txhold(zfsvfs, tx); 3271219089Spjd 3272219089Spjd zfs_sa_upgrade_txholds(tx, zp); 3273219089Spjd 3274209962Smm err = dmu_tx_assign(tx, TXG_NOWAIT); 3275168404Spjd if (err) { 3276209962Smm if (err == ERESTART) 3277168404Spjd dmu_tx_wait(tx); 3278209962Smm goto out; 3279168404Spjd } 3280168404Spjd 3281219089Spjd count = 0; 3282168404Spjd /* 3283168404Spjd * Set each attribute requested. 3284168404Spjd * We group settings according to the locks they need to acquire. 3285168404Spjd * 3286168404Spjd * Note: you cannot set ctime directly, although it will be 3287168404Spjd * updated as a side-effect of calling this function. 3288168404Spjd */ 3289168404Spjd 3290219089Spjd 3291219089Spjd if (mask & (AT_UID|AT_GID|AT_MODE)) 3292219089Spjd mutex_enter(&zp->z_acl_lock); 3293168404Spjd mutex_enter(&zp->z_lock); 3294168404Spjd 3295219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, 3296219089Spjd &zp->z_pflags, sizeof (zp->z_pflags)); 3297219089Spjd 3298219089Spjd if (attrzp) { 3299219089Spjd if (mask & (AT_UID|AT_GID|AT_MODE)) 3300219089Spjd mutex_enter(&attrzp->z_acl_lock); 3301219089Spjd mutex_enter(&attrzp->z_lock); 3302219089Spjd SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, 3303219089Spjd SA_ZPL_FLAGS(zfsvfs), NULL, &attrzp->z_pflags, 3304219089Spjd sizeof (attrzp->z_pflags)); 3305219089Spjd } 3306219089Spjd 3307219089Spjd if (mask & (AT_UID|AT_GID)) { 3308219089Spjd 3309219089Spjd if (mask & AT_UID) { 3310219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_UID(zfsvfs), NULL, 3311219089Spjd &new_uid, sizeof (new_uid)); 3312219089Spjd zp->z_uid = new_uid; 3313219089Spjd if (attrzp) { 3314219089Spjd SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, 3315219089Spjd SA_ZPL_UID(zfsvfs), NULL, &new_uid, 3316219089Spjd sizeof (new_uid)); 3317219089Spjd attrzp->z_uid = new_uid; 3318219089Spjd } 3319219089Spjd } 3320219089Spjd 3321219089Spjd if (mask & AT_GID) { 3322219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_GID(zfsvfs), 3323219089Spjd NULL, &new_gid, sizeof (new_gid)); 3324219089Spjd zp->z_gid = new_gid; 3325219089Spjd if (attrzp) { 3326219089Spjd SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, 3327219089Spjd SA_ZPL_GID(zfsvfs), NULL, &new_gid, 3328219089Spjd sizeof (new_gid)); 3329219089Spjd attrzp->z_gid = new_gid; 3330219089Spjd } 3331219089Spjd } 3332219089Spjd if (!(mask & AT_MODE)) { 3333219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), 3334219089Spjd NULL, &new_mode, sizeof (new_mode)); 3335219089Spjd new_mode = zp->z_mode; 3336219089Spjd } 3337219089Spjd err = zfs_acl_chown_setattr(zp); 3338219089Spjd ASSERT(err == 0); 3339219089Spjd if (attrzp) { 3340219089Spjd err = zfs_acl_chown_setattr(attrzp); 3341219089Spjd ASSERT(err == 0); 3342219089Spjd } 3343219089Spjd } 3344219089Spjd 3345168404Spjd if (mask & AT_MODE) { 3346219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MODE(zfsvfs), NULL, 3347219089Spjd &new_mode, sizeof (new_mode)); 3348219089Spjd zp->z_mode = new_mode; 3349219089Spjd ASSERT3U((uintptr_t)aclp, !=, 0); 3350209962Smm err = zfs_aclset_common(zp, aclp, cr, tx); 3351168404Spjd ASSERT3U(err, ==, 0); 3352219089Spjd if (zp->z_acl_cached) 3353219089Spjd zfs_acl_free(zp->z_acl_cached); 3354211932Smm zp->z_acl_cached = aclp; 3355211932Smm aclp = NULL; 3356168404Spjd } 3357168404Spjd 3358168404Spjd 3359219089Spjd if (mask & AT_ATIME) { 3360219089Spjd ZFS_TIME_ENCODE(&vap->va_atime, zp->z_atime); 3361219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_ATIME(zfsvfs), NULL, 3362219089Spjd &zp->z_atime, sizeof (zp->z_atime)); 3363168404Spjd } 3364168404Spjd 3365219089Spjd if (mask & AT_MTIME) { 3366219089Spjd ZFS_TIME_ENCODE(&vap->va_mtime, mtime); 3367219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, 3368219089Spjd mtime, sizeof (mtime)); 3369168404Spjd } 3370168404Spjd 3371185029Spjd /* XXX - shouldn't this be done *before* the ATIME/MTIME checks? */ 3372219089Spjd if (mask & AT_SIZE && !(mask & AT_MTIME)) { 3373219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), 3374219089Spjd NULL, mtime, sizeof (mtime)); 3375219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, 3376219089Spjd &ctime, sizeof (ctime)); 3377219089Spjd zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, 3378219089Spjd B_TRUE); 3379219089Spjd } else if (mask != 0) { 3380219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, 3381219089Spjd &ctime, sizeof (ctime)); 3382219089Spjd zfs_tstamp_update_setup(zp, STATE_CHANGED, mtime, ctime, 3383219089Spjd B_TRUE); 3384219089Spjd if (attrzp) { 3385219089Spjd SA_ADD_BULK_ATTR(xattr_bulk, xattr_count, 3386219089Spjd SA_ZPL_CTIME(zfsvfs), NULL, 3387219089Spjd &ctime, sizeof (ctime)); 3388219089Spjd zfs_tstamp_update_setup(attrzp, STATE_CHANGED, 3389219089Spjd mtime, ctime, B_TRUE); 3390219089Spjd } 3391219089Spjd } 3392185029Spjd /* 3393185029Spjd * Do this after setting timestamps to prevent timestamp 3394185029Spjd * update from toggling bit 3395185029Spjd */ 3396168404Spjd 3397185029Spjd if (xoap && (mask & AT_XVATTR)) { 3398209962Smm 3399209962Smm /* 3400209962Smm * restore trimmed off masks 3401209962Smm * so that return masks can be set for caller. 3402209962Smm */ 3403209962Smm 3404209962Smm if (XVA_ISSET_REQ(&tmpxvattr, XAT_APPENDONLY)) { 3405209962Smm XVA_SET_REQ(xvap, XAT_APPENDONLY); 3406209962Smm } 3407209962Smm if (XVA_ISSET_REQ(&tmpxvattr, XAT_NOUNLINK)) { 3408209962Smm XVA_SET_REQ(xvap, XAT_NOUNLINK); 3409209962Smm } 3410209962Smm if (XVA_ISSET_REQ(&tmpxvattr, XAT_IMMUTABLE)) { 3411209962Smm XVA_SET_REQ(xvap, XAT_IMMUTABLE); 3412209962Smm } 3413209962Smm if (XVA_ISSET_REQ(&tmpxvattr, XAT_NODUMP)) { 3414209962Smm XVA_SET_REQ(xvap, XAT_NODUMP); 3415209962Smm } 3416209962Smm if (XVA_ISSET_REQ(&tmpxvattr, XAT_AV_MODIFIED)) { 3417209962Smm XVA_SET_REQ(xvap, XAT_AV_MODIFIED); 3418209962Smm } 3419209962Smm if (XVA_ISSET_REQ(&tmpxvattr, XAT_AV_QUARANTINED)) { 3420209962Smm XVA_SET_REQ(xvap, XAT_AV_QUARANTINED); 3421209962Smm } 3422209962Smm 3423219089Spjd if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) 3424185029Spjd ASSERT(vp->v_type == VREG); 3425185029Spjd 3426219089Spjd zfs_xvattr_set(zp, xvap, tx); 3427185029Spjd } 3428185029Spjd 3429209962Smm if (fuid_dirtied) 3430209962Smm zfs_fuid_sync(zfsvfs, tx); 3431209962Smm 3432168404Spjd if (mask != 0) 3433185029Spjd zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask, fuidp); 3434168404Spjd 3435168404Spjd mutex_exit(&zp->z_lock); 3436219089Spjd if (mask & (AT_UID|AT_GID|AT_MODE)) 3437219089Spjd mutex_exit(&zp->z_acl_lock); 3438168404Spjd 3439219089Spjd if (attrzp) { 3440219089Spjd if (mask & (AT_UID|AT_GID|AT_MODE)) 3441219089Spjd mutex_exit(&attrzp->z_acl_lock); 3442219089Spjd mutex_exit(&attrzp->z_lock); 3443219089Spjd } 3444209962Smmout: 3445219089Spjd if (err == 0 && attrzp) { 3446219089Spjd err2 = sa_bulk_update(attrzp->z_sa_hdl, xattr_bulk, 3447219089Spjd xattr_count, tx); 3448219089Spjd ASSERT(err2 == 0); 3449219089Spjd } 3450219089Spjd 3451168404Spjd if (attrzp) 3452168404Spjd VN_RELE(ZTOV(attrzp)); 3453211932Smm if (aclp) 3454209962Smm zfs_acl_free(aclp); 3455168404Spjd 3456209962Smm if (fuidp) { 3457209962Smm zfs_fuid_info_free(fuidp); 3458209962Smm fuidp = NULL; 3459209962Smm } 3460209962Smm 3461219089Spjd if (err) { 3462209962Smm dmu_tx_abort(tx); 3463219089Spjd if (err == ERESTART) 3464219089Spjd goto top; 3465219089Spjd } else { 3466219089Spjd err2 = sa_bulk_update(zp->z_sa_hdl, bulk, count, tx); 3467209962Smm dmu_tx_commit(tx); 3468219089Spjd } 3469209962Smm 3470219089Spjdout2: 3471219089Spjd if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) 3472219089Spjd zil_commit(zilog, 0); 3473209962Smm 3474168404Spjd ZFS_EXIT(zfsvfs); 3475168404Spjd return (err); 3476168404Spjd} 3477168404Spjd 3478168404Spjdtypedef struct zfs_zlock { 3479168404Spjd krwlock_t *zl_rwlock; /* lock we acquired */ 3480168404Spjd znode_t *zl_znode; /* znode we held */ 3481168404Spjd struct zfs_zlock *zl_next; /* next in list */ 3482168404Spjd} zfs_zlock_t; 3483168404Spjd 3484168404Spjd/* 3485168404Spjd * Drop locks and release vnodes that were held by zfs_rename_lock(). 3486168404Spjd */ 3487168404Spjdstatic void 3488168404Spjdzfs_rename_unlock(zfs_zlock_t **zlpp) 3489168404Spjd{ 3490168404Spjd zfs_zlock_t *zl; 3491168404Spjd 3492168404Spjd while ((zl = *zlpp) != NULL) { 3493168404Spjd if (zl->zl_znode != NULL) 3494168404Spjd VN_RELE(ZTOV(zl->zl_znode)); 3495168404Spjd rw_exit(zl->zl_rwlock); 3496168404Spjd *zlpp = zl->zl_next; 3497168404Spjd kmem_free(zl, sizeof (*zl)); 3498168404Spjd } 3499168404Spjd} 3500168404Spjd 3501168404Spjd/* 3502168404Spjd * Search back through the directory tree, using the ".." entries. 3503168404Spjd * Lock each directory in the chain to prevent concurrent renames. 3504168404Spjd * Fail any attempt to move a directory into one of its own descendants. 3505168404Spjd * XXX - z_parent_lock can overlap with map or grow locks 3506168404Spjd */ 3507168404Spjdstatic int 3508168404Spjdzfs_rename_lock(znode_t *szp, znode_t *tdzp, znode_t *sdzp, zfs_zlock_t **zlpp) 3509168404Spjd{ 3510168404Spjd zfs_zlock_t *zl; 3511168404Spjd znode_t *zp = tdzp; 3512168404Spjd uint64_t rootid = zp->z_zfsvfs->z_root; 3513219089Spjd uint64_t oidp = zp->z_id; 3514168404Spjd krwlock_t *rwlp = &szp->z_parent_lock; 3515168404Spjd krw_t rw = RW_WRITER; 3516168404Spjd 3517168404Spjd /* 3518168404Spjd * First pass write-locks szp and compares to zp->z_id. 3519168404Spjd * Later passes read-lock zp and compare to zp->z_parent. 3520168404Spjd */ 3521168404Spjd do { 3522168404Spjd if (!rw_tryenter(rwlp, rw)) { 3523168404Spjd /* 3524168404Spjd * Another thread is renaming in this path. 3525168404Spjd * Note that if we are a WRITER, we don't have any 3526168404Spjd * parent_locks held yet. 3527168404Spjd */ 3528168404Spjd if (rw == RW_READER && zp->z_id > szp->z_id) { 3529168404Spjd /* 3530168404Spjd * Drop our locks and restart 3531168404Spjd */ 3532168404Spjd zfs_rename_unlock(&zl); 3533168404Spjd *zlpp = NULL; 3534168404Spjd zp = tdzp; 3535219089Spjd oidp = zp->z_id; 3536168404Spjd rwlp = &szp->z_parent_lock; 3537168404Spjd rw = RW_WRITER; 3538168404Spjd continue; 3539168404Spjd } else { 3540168404Spjd /* 3541168404Spjd * Wait for other thread to drop its locks 3542168404Spjd */ 3543168404Spjd rw_enter(rwlp, rw); 3544168404Spjd } 3545168404Spjd } 3546168404Spjd 3547168404Spjd zl = kmem_alloc(sizeof (*zl), KM_SLEEP); 3548168404Spjd zl->zl_rwlock = rwlp; 3549168404Spjd zl->zl_znode = NULL; 3550168404Spjd zl->zl_next = *zlpp; 3551168404Spjd *zlpp = zl; 3552168404Spjd 3553219089Spjd if (oidp == szp->z_id) /* We're a descendant of szp */ 3554168404Spjd return (EINVAL); 3555168404Spjd 3556219089Spjd if (oidp == rootid) /* We've hit the top */ 3557168404Spjd return (0); 3558168404Spjd 3559168404Spjd if (rw == RW_READER) { /* i.e. not the first pass */ 3560219089Spjd int error = zfs_zget(zp->z_zfsvfs, oidp, &zp); 3561168404Spjd if (error) 3562168404Spjd return (error); 3563168404Spjd zl->zl_znode = zp; 3564168404Spjd } 3565219089Spjd (void) sa_lookup(zp->z_sa_hdl, SA_ZPL_PARENT(zp->z_zfsvfs), 3566219089Spjd &oidp, sizeof (oidp)); 3567168404Spjd rwlp = &zp->z_parent_lock; 3568168404Spjd rw = RW_READER; 3569168404Spjd 3570168404Spjd } while (zp->z_id != sdzp->z_id); 3571168404Spjd 3572168404Spjd return (0); 3573168404Spjd} 3574168404Spjd 3575168404Spjd/* 3576168404Spjd * Move an entry from the provided source directory to the target 3577168404Spjd * directory. Change the entry name as indicated. 3578168404Spjd * 3579168404Spjd * IN: sdvp - Source directory containing the "old entry". 3580168404Spjd * snm - Old entry name. 3581168404Spjd * tdvp - Target directory to contain the "new entry". 3582168404Spjd * tnm - New entry name. 3583168404Spjd * cr - credentials of caller. 3584185029Spjd * ct - caller context 3585185029Spjd * flags - case flags 3586168404Spjd * 3587168404Spjd * RETURN: 0 if success 3588168404Spjd * error code if failure 3589168404Spjd * 3590168404Spjd * Timestamps: 3591168404Spjd * sdvp,tdvp - ctime|mtime updated 3592168404Spjd */ 3593185029Spjd/*ARGSUSED*/ 3594168404Spjdstatic int 3595185029Spjdzfs_rename(vnode_t *sdvp, char *snm, vnode_t *tdvp, char *tnm, cred_t *cr, 3596185029Spjd caller_context_t *ct, int flags) 3597168404Spjd{ 3598168404Spjd znode_t *tdzp, *szp, *tzp; 3599168404Spjd znode_t *sdzp = VTOZ(sdvp); 3600168404Spjd zfsvfs_t *zfsvfs = sdzp->z_zfsvfs; 3601185029Spjd zilog_t *zilog; 3602168962Spjd vnode_t *realvp; 3603168404Spjd zfs_dirlock_t *sdl, *tdl; 3604168404Spjd dmu_tx_t *tx; 3605168404Spjd zfs_zlock_t *zl; 3606185029Spjd int cmp, serr, terr; 3607185029Spjd int error = 0; 3608185029Spjd int zflg = 0; 3609168404Spjd 3610168404Spjd ZFS_ENTER(zfsvfs); 3611185029Spjd ZFS_VERIFY_ZP(sdzp); 3612185029Spjd zilog = zfsvfs->z_log; 3613168404Spjd 3614168962Spjd /* 3615168962Spjd * Make sure we have the real vp for the target directory. 3616168962Spjd */ 3617185029Spjd if (VOP_REALVP(tdvp, &realvp, ct) == 0) 3618168962Spjd tdvp = realvp; 3619168962Spjd 3620212694Smm if (tdvp->v_vfsp != sdvp->v_vfsp || zfsctl_is_node(tdvp)) { 3621168404Spjd ZFS_EXIT(zfsvfs); 3622168962Spjd return (EXDEV); 3623168404Spjd } 3624168404Spjd 3625168404Spjd tdzp = VTOZ(tdvp); 3626185029Spjd ZFS_VERIFY_ZP(tdzp); 3627185029Spjd if (zfsvfs->z_utf8 && u8_validate(tnm, 3628185029Spjd strlen(tnm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { 3629185029Spjd ZFS_EXIT(zfsvfs); 3630185029Spjd return (EILSEQ); 3631185029Spjd } 3632185029Spjd 3633185029Spjd if (flags & FIGNORECASE) 3634185029Spjd zflg |= ZCILOOK; 3635185029Spjd 3636168404Spjdtop: 3637168404Spjd szp = NULL; 3638168404Spjd tzp = NULL; 3639168404Spjd zl = NULL; 3640168404Spjd 3641168404Spjd /* 3642168404Spjd * This is to prevent the creation of links into attribute space 3643168404Spjd * by renaming a linked file into/outof an attribute directory. 3644168404Spjd * See the comment in zfs_link() for why this is considered bad. 3645168404Spjd */ 3646219089Spjd if ((tdzp->z_pflags & ZFS_XATTR) != (sdzp->z_pflags & ZFS_XATTR)) { 3647168962Spjd ZFS_EXIT(zfsvfs); 3648168962Spjd return (EINVAL); 3649168404Spjd } 3650168404Spjd 3651168404Spjd /* 3652168404Spjd * Lock source and target directory entries. To prevent deadlock, 3653168404Spjd * a lock ordering must be defined. We lock the directory with 3654168404Spjd * the smallest object id first, or if it's a tie, the one with 3655168404Spjd * the lexically first name. 3656168404Spjd */ 3657168404Spjd if (sdzp->z_id < tdzp->z_id) { 3658168962Spjd cmp = -1; 3659168962Spjd } else if (sdzp->z_id > tdzp->z_id) { 3660168962Spjd cmp = 1; 3661168962Spjd } else { 3662185029Spjd /* 3663185029Spjd * First compare the two name arguments without 3664185029Spjd * considering any case folding. 3665185029Spjd */ 3666185029Spjd int nofold = (zfsvfs->z_norm & ~U8_TEXTPREP_TOUPPER); 3667185029Spjd 3668185029Spjd cmp = u8_strcmp(snm, tnm, 0, nofold, U8_UNICODE_LATEST, &error); 3669185029Spjd ASSERT(error == 0 || !zfsvfs->z_utf8); 3670168962Spjd if (cmp == 0) { 3671168962Spjd /* 3672168962Spjd * POSIX: "If the old argument and the new argument 3673168962Spjd * both refer to links to the same existing file, 3674168962Spjd * the rename() function shall return successfully 3675168962Spjd * and perform no other action." 3676168962Spjd */ 3677168962Spjd ZFS_EXIT(zfsvfs); 3678168962Spjd return (0); 3679168962Spjd } 3680185029Spjd /* 3681185029Spjd * If the file system is case-folding, then we may 3682185029Spjd * have some more checking to do. A case-folding file 3683185029Spjd * system is either supporting mixed case sensitivity 3684185029Spjd * access or is completely case-insensitive. Note 3685185029Spjd * that the file system is always case preserving. 3686185029Spjd * 3687185029Spjd * In mixed sensitivity mode case sensitive behavior 3688185029Spjd * is the default. FIGNORECASE must be used to 3689185029Spjd * explicitly request case insensitive behavior. 3690185029Spjd * 3691185029Spjd * If the source and target names provided differ only 3692185029Spjd * by case (e.g., a request to rename 'tim' to 'Tim'), 3693185029Spjd * we will treat this as a special case in the 3694185029Spjd * case-insensitive mode: as long as the source name 3695185029Spjd * is an exact match, we will allow this to proceed as 3696185029Spjd * a name-change request. 3697185029Spjd */ 3698185029Spjd if ((zfsvfs->z_case == ZFS_CASE_INSENSITIVE || 3699185029Spjd (zfsvfs->z_case == ZFS_CASE_MIXED && 3700185029Spjd flags & FIGNORECASE)) && 3701185029Spjd u8_strcmp(snm, tnm, 0, zfsvfs->z_norm, U8_UNICODE_LATEST, 3702185029Spjd &error) == 0) { 3703185029Spjd /* 3704185029Spjd * case preserving rename request, require exact 3705185029Spjd * name matches 3706185029Spjd */ 3707185029Spjd zflg |= ZCIEXACT; 3708185029Spjd zflg &= ~ZCILOOK; 3709185029Spjd } 3710168962Spjd } 3711185029Spjd 3712208131Smm /* 3713208131Smm * If the source and destination directories are the same, we should 3714208131Smm * grab the z_name_lock of that directory only once. 3715208131Smm */ 3716208131Smm if (sdzp == tdzp) { 3717208131Smm zflg |= ZHAVELOCK; 3718208131Smm rw_enter(&sdzp->z_name_lock, RW_READER); 3719208131Smm } 3720208131Smm 3721168962Spjd if (cmp < 0) { 3722185029Spjd serr = zfs_dirent_lock(&sdl, sdzp, snm, &szp, 3723185029Spjd ZEXISTS | zflg, NULL, NULL); 3724185029Spjd terr = zfs_dirent_lock(&tdl, 3725185029Spjd tdzp, tnm, &tzp, ZRENAMING | zflg, NULL, NULL); 3726168962Spjd } else { 3727185029Spjd terr = zfs_dirent_lock(&tdl, 3728185029Spjd tdzp, tnm, &tzp, zflg, NULL, NULL); 3729185029Spjd serr = zfs_dirent_lock(&sdl, 3730185029Spjd sdzp, snm, &szp, ZEXISTS | ZRENAMING | zflg, 3731185029Spjd NULL, NULL); 3732168404Spjd } 3733168404Spjd 3734168962Spjd if (serr) { 3735168404Spjd /* 3736168404Spjd * Source entry invalid or not there. 3737168404Spjd */ 3738168962Spjd if (!terr) { 3739168404Spjd zfs_dirent_unlock(tdl); 3740168962Spjd if (tzp) 3741168962Spjd VN_RELE(ZTOV(tzp)); 3742168962Spjd } 3743208131Smm 3744208131Smm if (sdzp == tdzp) 3745208131Smm rw_exit(&sdzp->z_name_lock); 3746208131Smm 3747219089Spjd /* 3748219089Spjd * FreeBSD: In OpenSolaris they only check if rename source is 3749219089Spjd * ".." here, because "." is handled in their lookup. This is 3750219089Spjd * not the case for FreeBSD, so we check for "." explicitly. 3751219089Spjd */ 3752168404Spjd if (strcmp(snm, ".") == 0 || strcmp(snm, "..") == 0) 3753168404Spjd serr = EINVAL; 3754168962Spjd ZFS_EXIT(zfsvfs); 3755168962Spjd return (serr); 3756168404Spjd } 3757168404Spjd if (terr) { 3758168404Spjd zfs_dirent_unlock(sdl); 3759168962Spjd VN_RELE(ZTOV(szp)); 3760208131Smm 3761208131Smm if (sdzp == tdzp) 3762208131Smm rw_exit(&sdzp->z_name_lock); 3763208131Smm 3764168404Spjd if (strcmp(tnm, "..") == 0) 3765168404Spjd terr = EINVAL; 3766168962Spjd ZFS_EXIT(zfsvfs); 3767168962Spjd return (terr); 3768168404Spjd } 3769168404Spjd 3770168404Spjd /* 3771168404Spjd * Must have write access at the source to remove the old entry 3772168404Spjd * and write access at the target to create the new entry. 3773168404Spjd * Note that if target and source are the same, this can be 3774168404Spjd * done in a single check. 3775168404Spjd */ 3776168404Spjd 3777168404Spjd if (error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr)) 3778168404Spjd goto out; 3779168404Spjd 3780168962Spjd if (ZTOV(szp)->v_type == VDIR) { 3781168404Spjd /* 3782168404Spjd * Check to make sure rename is valid. 3783168404Spjd * Can't do a move like this: /usr/a/b to /usr/a/b/c/d 3784168404Spjd */ 3785168404Spjd if (error = zfs_rename_lock(szp, tdzp, sdzp, &zl)) 3786168404Spjd goto out; 3787168404Spjd } 3788168404Spjd 3789168404Spjd /* 3790168404Spjd * Does target exist? 3791168404Spjd */ 3792168404Spjd if (tzp) { 3793168404Spjd /* 3794168404Spjd * Source and target must be the same type. 3795168404Spjd */ 3796168962Spjd if (ZTOV(szp)->v_type == VDIR) { 3797168962Spjd if (ZTOV(tzp)->v_type != VDIR) { 3798168404Spjd error = ENOTDIR; 3799168404Spjd goto out; 3800168404Spjd } 3801168404Spjd } else { 3802168962Spjd if (ZTOV(tzp)->v_type == VDIR) { 3803168404Spjd error = EISDIR; 3804168404Spjd goto out; 3805168404Spjd } 3806168404Spjd } 3807168404Spjd /* 3808168404Spjd * POSIX dictates that when the source and target 3809168404Spjd * entries refer to the same file object, rename 3810168404Spjd * must do nothing and exit without error. 3811168404Spjd */ 3812168404Spjd if (szp->z_id == tzp->z_id) { 3813168404Spjd error = 0; 3814168404Spjd goto out; 3815168404Spjd } 3816168404Spjd } 3817168404Spjd 3818185029Spjd vnevent_rename_src(ZTOV(szp), sdvp, snm, ct); 3819168962Spjd if (tzp) 3820185029Spjd vnevent_rename_dest(ZTOV(tzp), tdvp, tnm, ct); 3821168962Spjd 3822185029Spjd /* 3823185029Spjd * notify the target directory if it is not the same 3824185029Spjd * as source directory. 3825185029Spjd */ 3826185029Spjd if (tdvp != sdvp) { 3827185029Spjd vnevent_rename_dest_dir(tdvp, ct); 3828185029Spjd } 3829185029Spjd 3830168404Spjd tx = dmu_tx_create(zfsvfs->z_os); 3831219089Spjd dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); 3832219089Spjd dmu_tx_hold_sa(tx, sdzp->z_sa_hdl, B_FALSE); 3833168404Spjd dmu_tx_hold_zap(tx, sdzp->z_id, FALSE, snm); 3834168404Spjd dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, tnm); 3835219089Spjd if (sdzp != tdzp) { 3836219089Spjd dmu_tx_hold_sa(tx, tdzp->z_sa_hdl, B_FALSE); 3837219089Spjd zfs_sa_upgrade_txholds(tx, tdzp); 3838219089Spjd } 3839219089Spjd if (tzp) { 3840219089Spjd dmu_tx_hold_sa(tx, tzp->z_sa_hdl, B_FALSE); 3841219089Spjd zfs_sa_upgrade_txholds(tx, tzp); 3842219089Spjd } 3843219089Spjd 3844219089Spjd zfs_sa_upgrade_txholds(tx, szp); 3845168404Spjd dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); 3846209962Smm error = dmu_tx_assign(tx, TXG_NOWAIT); 3847168404Spjd if (error) { 3848168404Spjd if (zl != NULL) 3849168404Spjd zfs_rename_unlock(&zl); 3850168404Spjd zfs_dirent_unlock(sdl); 3851168404Spjd zfs_dirent_unlock(tdl); 3852208131Smm 3853208131Smm if (sdzp == tdzp) 3854208131Smm rw_exit(&sdzp->z_name_lock); 3855208131Smm 3856168962Spjd VN_RELE(ZTOV(szp)); 3857168962Spjd if (tzp) 3858168962Spjd VN_RELE(ZTOV(tzp)); 3859209962Smm if (error == ERESTART) { 3860168404Spjd dmu_tx_wait(tx); 3861168404Spjd dmu_tx_abort(tx); 3862168404Spjd goto top; 3863168404Spjd } 3864168404Spjd dmu_tx_abort(tx); 3865168962Spjd ZFS_EXIT(zfsvfs); 3866168962Spjd return (error); 3867168404Spjd } 3868168404Spjd 3869168404Spjd if (tzp) /* Attempt to remove the existing target */ 3870185029Spjd error = zfs_link_destroy(tdl, tzp, tx, zflg, NULL); 3871168404Spjd 3872168404Spjd if (error == 0) { 3873168404Spjd error = zfs_link_create(tdl, szp, tx, ZRENAMING); 3874168404Spjd if (error == 0) { 3875219089Spjd szp->z_pflags |= ZFS_AV_MODIFIED; 3876185029Spjd 3877219089Spjd error = sa_update(szp->z_sa_hdl, SA_ZPL_FLAGS(zfsvfs), 3878219089Spjd (void *)&szp->z_pflags, sizeof (uint64_t), tx); 3879219089Spjd ASSERT3U(error, ==, 0); 3880219089Spjd 3881168404Spjd error = zfs_link_destroy(sdl, szp, tx, ZRENAMING, NULL); 3882219089Spjd if (error == 0) { 3883219089Spjd zfs_log_rename(zilog, tx, TX_RENAME | 3884219089Spjd (flags & FIGNORECASE ? TX_CI : 0), sdzp, 3885219089Spjd sdl->dl_name, tdzp, tdl->dl_name, szp); 3886185029Spjd 3887219089Spjd /* 3888219089Spjd * Update path information for the target vnode 3889219089Spjd */ 3890219089Spjd vn_renamepath(tdvp, ZTOV(szp), tnm, 3891219089Spjd strlen(tnm)); 3892219089Spjd } else { 3893219089Spjd /* 3894219089Spjd * At this point, we have successfully created 3895219089Spjd * the target name, but have failed to remove 3896219089Spjd * the source name. Since the create was done 3897219089Spjd * with the ZRENAMING flag, there are 3898219089Spjd * complications; for one, the link count is 3899219089Spjd * wrong. The easiest way to deal with this 3900219089Spjd * is to remove the newly created target, and 3901219089Spjd * return the original error. This must 3902219089Spjd * succeed; fortunately, it is very unlikely to 3903219089Spjd * fail, since we just created it. 3904219089Spjd */ 3905219089Spjd VERIFY3U(zfs_link_destroy(tdl, szp, tx, 3906219089Spjd ZRENAMING, NULL), ==, 0); 3907219089Spjd } 3908168404Spjd } 3909168404Spjd#ifdef FREEBSD_NAMECACHE 3910168404Spjd if (error == 0) { 3911168404Spjd cache_purge(sdvp); 3912168404Spjd cache_purge(tdvp); 3913168404Spjd } 3914168404Spjd#endif 3915168404Spjd } 3916168404Spjd 3917168404Spjd dmu_tx_commit(tx); 3918168404Spjdout: 3919168404Spjd if (zl != NULL) 3920168404Spjd zfs_rename_unlock(&zl); 3921168404Spjd 3922168404Spjd zfs_dirent_unlock(sdl); 3923168404Spjd zfs_dirent_unlock(tdl); 3924168404Spjd 3925208131Smm if (sdzp == tdzp) 3926208131Smm rw_exit(&sdzp->z_name_lock); 3927208131Smm 3928219089Spjd 3929168962Spjd VN_RELE(ZTOV(szp)); 3930168404Spjd if (tzp) 3931168962Spjd VN_RELE(ZTOV(tzp)); 3932168404Spjd 3933219089Spjd if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) 3934219089Spjd zil_commit(zilog, 0); 3935219089Spjd 3936168404Spjd ZFS_EXIT(zfsvfs); 3937168404Spjd 3938168404Spjd return (error); 3939168404Spjd} 3940168404Spjd 3941168404Spjd/* 3942168404Spjd * Insert the indicated symbolic reference entry into the directory. 3943168404Spjd * 3944168404Spjd * IN: dvp - Directory to contain new symbolic link. 3945168404Spjd * link - Name for new symlink entry. 3946168404Spjd * vap - Attributes of new entry. 3947168404Spjd * target - Target path of new symlink. 3948168404Spjd * cr - credentials of caller. 3949185029Spjd * ct - caller context 3950185029Spjd * flags - case flags 3951168404Spjd * 3952168404Spjd * RETURN: 0 if success 3953168404Spjd * error code if failure 3954168404Spjd * 3955168404Spjd * Timestamps: 3956168404Spjd * dvp - ctime|mtime updated 3957168404Spjd */ 3958185029Spjd/*ARGSUSED*/ 3959168404Spjdstatic int 3960185029Spjdzfs_symlink(vnode_t *dvp, vnode_t **vpp, char *name, vattr_t *vap, char *link, 3961185029Spjd cred_t *cr, kthread_t *td) 3962168404Spjd{ 3963168404Spjd znode_t *zp, *dzp = VTOZ(dvp); 3964168404Spjd zfs_dirlock_t *dl; 3965168404Spjd dmu_tx_t *tx; 3966168404Spjd zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 3967185029Spjd zilog_t *zilog; 3968219089Spjd uint64_t len = strlen(link); 3969168404Spjd int error; 3970185029Spjd int zflg = ZNEW; 3971209962Smm zfs_acl_ids_t acl_ids; 3972209962Smm boolean_t fuid_dirtied; 3973219089Spjd uint64_t txtype = TX_SYMLINK; 3974185029Spjd int flags = 0; 3975168404Spjd 3976168962Spjd ASSERT(vap->va_type == VLNK); 3977168404Spjd 3978168404Spjd ZFS_ENTER(zfsvfs); 3979185029Spjd ZFS_VERIFY_ZP(dzp); 3980185029Spjd zilog = zfsvfs->z_log; 3981185029Spjd 3982185029Spjd if (zfsvfs->z_utf8 && u8_validate(name, strlen(name), 3983185029Spjd NULL, U8_VALIDATE_ENTIRE, &error) < 0) { 3984185029Spjd ZFS_EXIT(zfsvfs); 3985185029Spjd return (EILSEQ); 3986185029Spjd } 3987185029Spjd if (flags & FIGNORECASE) 3988185029Spjd zflg |= ZCILOOK; 3989168404Spjd 3990168404Spjd if (len > MAXPATHLEN) { 3991168404Spjd ZFS_EXIT(zfsvfs); 3992168404Spjd return (ENAMETOOLONG); 3993168404Spjd } 3994168404Spjd 3995219089Spjd if ((error = zfs_acl_ids_create(dzp, 0, 3996219089Spjd vap, cr, NULL, &acl_ids)) != 0) { 3997219089Spjd ZFS_EXIT(zfsvfs); 3998219089Spjd return (error); 3999219089Spjd } 4000219089Spjdtop: 4001168404Spjd /* 4002168404Spjd * Attempt to lock directory; fail if entry already exists. 4003168404Spjd */ 4004185029Spjd error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, NULL); 4005185029Spjd if (error) { 4006219089Spjd zfs_acl_ids_free(&acl_ids); 4007168404Spjd ZFS_EXIT(zfsvfs); 4008168404Spjd return (error); 4009168404Spjd } 4010168404Spjd 4011219089Spjd if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) { 4012219089Spjd zfs_acl_ids_free(&acl_ids); 4013219089Spjd zfs_dirent_unlock(dl); 4014219089Spjd ZFS_EXIT(zfsvfs); 4015219089Spjd return (error); 4016219089Spjd } 4017219089Spjd 4018209962Smm if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) { 4019209962Smm zfs_acl_ids_free(&acl_ids); 4020209962Smm zfs_dirent_unlock(dl); 4021209962Smm ZFS_EXIT(zfsvfs); 4022209962Smm return (EDQUOT); 4023209962Smm } 4024168404Spjd tx = dmu_tx_create(zfsvfs->z_os); 4025209962Smm fuid_dirtied = zfsvfs->z_fuid_dirty; 4026168404Spjd dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len)); 4027168404Spjd dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); 4028219089Spjd dmu_tx_hold_sa_create(tx, acl_ids.z_aclp->z_acl_bytes + 4029219089Spjd ZFS_SA_BASE_ATTR_SIZE + len); 4030219089Spjd dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); 4031219089Spjd if (!zfsvfs->z_use_sa && acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { 4032219089Spjd dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, 4033219089Spjd acl_ids.z_aclp->z_acl_bytes); 4034219089Spjd } 4035209962Smm if (fuid_dirtied) 4036209962Smm zfs_fuid_txhold(zfsvfs, tx); 4037209962Smm error = dmu_tx_assign(tx, TXG_NOWAIT); 4038168404Spjd if (error) { 4039168404Spjd zfs_dirent_unlock(dl); 4040209962Smm if (error == ERESTART) { 4041168404Spjd dmu_tx_wait(tx); 4042168404Spjd dmu_tx_abort(tx); 4043168404Spjd goto top; 4044168404Spjd } 4045219089Spjd zfs_acl_ids_free(&acl_ids); 4046168404Spjd dmu_tx_abort(tx); 4047168404Spjd ZFS_EXIT(zfsvfs); 4048168404Spjd return (error); 4049168404Spjd } 4050168404Spjd 4051168404Spjd /* 4052168404Spjd * Create a new object for the symlink. 4053219089Spjd * for version 4 ZPL datsets the symlink will be an SA attribute 4054168404Spjd */ 4055219089Spjd zfs_mknode(dzp, vap, tx, cr, 0, &zp, &acl_ids); 4056168404Spjd 4057219089Spjd if (fuid_dirtied) 4058219089Spjd zfs_fuid_sync(zfsvfs, tx); 4059209962Smm 4060219089Spjd mutex_enter(&zp->z_lock); 4061219089Spjd if (zp->z_is_sa) 4062219089Spjd error = sa_update(zp->z_sa_hdl, SA_ZPL_SYMLINK(zfsvfs), 4063219089Spjd link, len, tx); 4064219089Spjd else 4065219089Spjd zfs_sa_symlink(zp, link, len, tx); 4066219089Spjd mutex_exit(&zp->z_lock); 4067168404Spjd 4068219089Spjd zp->z_size = len; 4069219089Spjd (void) sa_update(zp->z_sa_hdl, SA_ZPL_SIZE(zfsvfs), 4070219089Spjd &zp->z_size, sizeof (zp->z_size), tx); 4071168404Spjd /* 4072168404Spjd * Insert the new object into the directory. 4073168404Spjd */ 4074168404Spjd (void) zfs_link_create(dl, zp, tx, ZNEW); 4075168404Spjd 4076219089Spjd if (flags & FIGNORECASE) 4077219089Spjd txtype |= TX_CI; 4078219089Spjd zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link); 4079219089Spjd *vpp = ZTOV(zp); 4080219089Spjd 4081209962Smm zfs_acl_ids_free(&acl_ids); 4082209962Smm 4083168404Spjd dmu_tx_commit(tx); 4084168404Spjd 4085168404Spjd zfs_dirent_unlock(dl); 4086168404Spjd 4087219089Spjd if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) 4088219089Spjd zil_commit(zilog, 0); 4089219089Spjd 4090168404Spjd ZFS_EXIT(zfsvfs); 4091168404Spjd return (error); 4092168404Spjd} 4093168404Spjd 4094168404Spjd/* 4095168404Spjd * Return, in the buffer contained in the provided uio structure, 4096168404Spjd * the symbolic path referred to by vp. 4097168404Spjd * 4098168404Spjd * IN: vp - vnode of symbolic link. 4099168404Spjd * uoip - structure to contain the link path. 4100168404Spjd * cr - credentials of caller. 4101185029Spjd * ct - caller context 4102168404Spjd * 4103168404Spjd * OUT: uio - structure to contain the link path. 4104168404Spjd * 4105168404Spjd * RETURN: 0 if success 4106168404Spjd * error code if failure 4107168404Spjd * 4108168404Spjd * Timestamps: 4109168404Spjd * vp - atime updated 4110168404Spjd */ 4111168404Spjd/* ARGSUSED */ 4112168404Spjdstatic int 4113185029Spjdzfs_readlink(vnode_t *vp, uio_t *uio, cred_t *cr, caller_context_t *ct) 4114168404Spjd{ 4115168404Spjd znode_t *zp = VTOZ(vp); 4116168404Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 4117168404Spjd int error; 4118168404Spjd 4119168404Spjd ZFS_ENTER(zfsvfs); 4120185029Spjd ZFS_VERIFY_ZP(zp); 4121168404Spjd 4122219089Spjd mutex_enter(&zp->z_lock); 4123219089Spjd if (zp->z_is_sa) 4124219089Spjd error = sa_lookup_uio(zp->z_sa_hdl, 4125219089Spjd SA_ZPL_SYMLINK(zfsvfs), uio); 4126219089Spjd else 4127219089Spjd error = zfs_sa_readlink(zp, uio); 4128219089Spjd mutex_exit(&zp->z_lock); 4129168404Spjd 4130168404Spjd ZFS_ACCESSTIME_STAMP(zfsvfs, zp); 4131219089Spjd 4132168404Spjd ZFS_EXIT(zfsvfs); 4133168404Spjd return (error); 4134168404Spjd} 4135168404Spjd 4136168404Spjd/* 4137168404Spjd * Insert a new entry into directory tdvp referencing svp. 4138168404Spjd * 4139168404Spjd * IN: tdvp - Directory to contain new entry. 4140168404Spjd * svp - vnode of new entry. 4141168404Spjd * name - name of new entry. 4142168404Spjd * cr - credentials of caller. 4143185029Spjd * ct - caller context 4144168404Spjd * 4145168404Spjd * RETURN: 0 if success 4146168404Spjd * error code if failure 4147168404Spjd * 4148168404Spjd * Timestamps: 4149168404Spjd * tdvp - ctime|mtime updated 4150168404Spjd * svp - ctime updated 4151168404Spjd */ 4152168404Spjd/* ARGSUSED */ 4153168404Spjdstatic int 4154185029Spjdzfs_link(vnode_t *tdvp, vnode_t *svp, char *name, cred_t *cr, 4155185029Spjd caller_context_t *ct, int flags) 4156168404Spjd{ 4157168404Spjd znode_t *dzp = VTOZ(tdvp); 4158168404Spjd znode_t *tzp, *szp; 4159168404Spjd zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 4160185029Spjd zilog_t *zilog; 4161168404Spjd zfs_dirlock_t *dl; 4162168404Spjd dmu_tx_t *tx; 4163168962Spjd vnode_t *realvp; 4164168404Spjd int error; 4165185029Spjd int zf = ZNEW; 4166212694Smm uint64_t parent; 4167185029Spjd uid_t owner; 4168168404Spjd 4169168404Spjd ASSERT(tdvp->v_type == VDIR); 4170168404Spjd 4171168404Spjd ZFS_ENTER(zfsvfs); 4172185029Spjd ZFS_VERIFY_ZP(dzp); 4173185029Spjd zilog = zfsvfs->z_log; 4174168404Spjd 4175185029Spjd if (VOP_REALVP(svp, &realvp, ct) == 0) 4176168962Spjd svp = realvp; 4177168962Spjd 4178212694Smm /* 4179212694Smm * POSIX dictates that we return EPERM here. 4180212694Smm * Better choices include ENOTSUP or EISDIR. 4181212694Smm */ 4182212694Smm if (svp->v_type == VDIR) { 4183168404Spjd ZFS_EXIT(zfsvfs); 4184212694Smm return (EPERM); 4185212694Smm } 4186212694Smm 4187212694Smm if (svp->v_vfsp != tdvp->v_vfsp || zfsctl_is_node(svp)) { 4188212694Smm ZFS_EXIT(zfsvfs); 4189168404Spjd return (EXDEV); 4190168404Spjd } 4191212694Smm 4192185029Spjd szp = VTOZ(svp); 4193185029Spjd ZFS_VERIFY_ZP(szp); 4194168404Spjd 4195212694Smm /* Prevent links to .zfs/shares files */ 4196212694Smm 4197219089Spjd if ((error = sa_lookup(szp->z_sa_hdl, SA_ZPL_PARENT(zfsvfs), 4198219089Spjd &parent, sizeof (uint64_t))) != 0) { 4199212694Smm ZFS_EXIT(zfsvfs); 4200219089Spjd return (error); 4201219089Spjd } 4202219089Spjd if (parent == zfsvfs->z_shares_dir) { 4203219089Spjd ZFS_EXIT(zfsvfs); 4204212694Smm return (EPERM); 4205212694Smm } 4206212694Smm 4207185029Spjd if (zfsvfs->z_utf8 && u8_validate(name, 4208185029Spjd strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { 4209185029Spjd ZFS_EXIT(zfsvfs); 4210185029Spjd return (EILSEQ); 4211185029Spjd } 4212185029Spjd if (flags & FIGNORECASE) 4213185029Spjd zf |= ZCILOOK; 4214185029Spjd 4215168404Spjd /* 4216168404Spjd * We do not support links between attributes and non-attributes 4217168404Spjd * because of the potential security risk of creating links 4218168404Spjd * into "normal" file space in order to circumvent restrictions 4219168404Spjd * imposed in attribute space. 4220168404Spjd */ 4221219089Spjd if ((szp->z_pflags & ZFS_XATTR) != (dzp->z_pflags & ZFS_XATTR)) { 4222168404Spjd ZFS_EXIT(zfsvfs); 4223168404Spjd return (EINVAL); 4224168404Spjd } 4225168404Spjd 4226168404Spjd 4227219089Spjd owner = zfs_fuid_map_id(zfsvfs, szp->z_uid, cr, ZFS_OWNER); 4228219089Spjd if (owner != crgetuid(cr) && secpolicy_basic_link(svp, cr) != 0) { 4229168404Spjd ZFS_EXIT(zfsvfs); 4230168404Spjd return (EPERM); 4231168404Spjd } 4232168404Spjd 4233185029Spjd if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) { 4234168404Spjd ZFS_EXIT(zfsvfs); 4235168404Spjd return (error); 4236168404Spjd } 4237168404Spjd 4238212694Smmtop: 4239168404Spjd /* 4240168404Spjd * Attempt to lock directory; fail if entry already exists. 4241168404Spjd */ 4242185029Spjd error = zfs_dirent_lock(&dl, dzp, name, &tzp, zf, NULL, NULL); 4243185029Spjd if (error) { 4244168404Spjd ZFS_EXIT(zfsvfs); 4245168404Spjd return (error); 4246168404Spjd } 4247168404Spjd 4248168404Spjd tx = dmu_tx_create(zfsvfs->z_os); 4249219089Spjd dmu_tx_hold_sa(tx, szp->z_sa_hdl, B_FALSE); 4250168404Spjd dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); 4251219089Spjd zfs_sa_upgrade_txholds(tx, szp); 4252219089Spjd zfs_sa_upgrade_txholds(tx, dzp); 4253209962Smm error = dmu_tx_assign(tx, TXG_NOWAIT); 4254168404Spjd if (error) { 4255168404Spjd zfs_dirent_unlock(dl); 4256209962Smm if (error == ERESTART) { 4257168404Spjd dmu_tx_wait(tx); 4258168404Spjd dmu_tx_abort(tx); 4259168404Spjd goto top; 4260168404Spjd } 4261168404Spjd dmu_tx_abort(tx); 4262168404Spjd ZFS_EXIT(zfsvfs); 4263168404Spjd return (error); 4264168404Spjd } 4265168404Spjd 4266168404Spjd error = zfs_link_create(dl, szp, tx, 0); 4267168404Spjd 4268185029Spjd if (error == 0) { 4269185029Spjd uint64_t txtype = TX_LINK; 4270185029Spjd if (flags & FIGNORECASE) 4271185029Spjd txtype |= TX_CI; 4272185029Spjd zfs_log_link(zilog, tx, txtype, dzp, szp, name); 4273185029Spjd } 4274168404Spjd 4275168404Spjd dmu_tx_commit(tx); 4276168404Spjd 4277168404Spjd zfs_dirent_unlock(dl); 4278168404Spjd 4279185029Spjd if (error == 0) { 4280185029Spjd vnevent_link(svp, ct); 4281185029Spjd } 4282185029Spjd 4283219089Spjd if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) 4284219089Spjd zil_commit(zilog, 0); 4285219089Spjd 4286168404Spjd ZFS_EXIT(zfsvfs); 4287168404Spjd return (error); 4288168404Spjd} 4289168404Spjd 4290219089Spjd#ifdef sun 4291219089Spjd/* 4292219089Spjd * zfs_null_putapage() is used when the file system has been force 4293219089Spjd * unmounted. It just drops the pages. 4294219089Spjd */ 4295219089Spjd/* ARGSUSED */ 4296219089Spjdstatic int 4297219089Spjdzfs_null_putapage(vnode_t *vp, page_t *pp, u_offset_t *offp, 4298219089Spjd size_t *lenp, int flags, cred_t *cr) 4299219089Spjd{ 4300219089Spjd pvn_write_done(pp, B_INVAL|B_FORCE|B_ERROR); 4301219089Spjd return (0); 4302219089Spjd} 4303219089Spjd 4304219089Spjd/* 4305219089Spjd * Push a page out to disk, klustering if possible. 4306219089Spjd * 4307219089Spjd * IN: vp - file to push page to. 4308219089Spjd * pp - page to push. 4309219089Spjd * flags - additional flags. 4310219089Spjd * cr - credentials of caller. 4311219089Spjd * 4312219089Spjd * OUT: offp - start of range pushed. 4313219089Spjd * lenp - len of range pushed. 4314219089Spjd * 4315219089Spjd * RETURN: 0 if success 4316219089Spjd * error code if failure 4317219089Spjd * 4318219089Spjd * NOTE: callers must have locked the page to be pushed. On 4319219089Spjd * exit, the page (and all other pages in the kluster) must be 4320219089Spjd * unlocked. 4321219089Spjd */ 4322219089Spjd/* ARGSUSED */ 4323219089Spjdstatic int 4324219089Spjdzfs_putapage(vnode_t *vp, page_t *pp, u_offset_t *offp, 4325219089Spjd size_t *lenp, int flags, cred_t *cr) 4326219089Spjd{ 4327219089Spjd znode_t *zp = VTOZ(vp); 4328219089Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 4329219089Spjd dmu_tx_t *tx; 4330219089Spjd u_offset_t off, koff; 4331219089Spjd size_t len, klen; 4332219089Spjd int err; 4333219089Spjd 4334219089Spjd off = pp->p_offset; 4335219089Spjd len = PAGESIZE; 4336219089Spjd /* 4337219089Spjd * If our blocksize is bigger than the page size, try to kluster 4338219089Spjd * multiple pages so that we write a full block (thus avoiding 4339219089Spjd * a read-modify-write). 4340219089Spjd */ 4341219089Spjd if (off < zp->z_size && zp->z_blksz > PAGESIZE) { 4342219089Spjd klen = P2ROUNDUP((ulong_t)zp->z_blksz, PAGESIZE); 4343219089Spjd koff = ISP2(klen) ? P2ALIGN(off, (u_offset_t)klen) : 0; 4344219089Spjd ASSERT(koff <= zp->z_size); 4345219089Spjd if (koff + klen > zp->z_size) 4346219089Spjd klen = P2ROUNDUP(zp->z_size - koff, (uint64_t)PAGESIZE); 4347219089Spjd pp = pvn_write_kluster(vp, pp, &off, &len, koff, klen, flags); 4348219089Spjd } 4349219089Spjd ASSERT3U(btop(len), ==, btopr(len)); 4350219089Spjd 4351219089Spjd /* 4352219089Spjd * Can't push pages past end-of-file. 4353219089Spjd */ 4354219089Spjd if (off >= zp->z_size) { 4355219089Spjd /* ignore all pages */ 4356219089Spjd err = 0; 4357219089Spjd goto out; 4358219089Spjd } else if (off + len > zp->z_size) { 4359219089Spjd int npages = btopr(zp->z_size - off); 4360219089Spjd page_t *trunc; 4361219089Spjd 4362219089Spjd page_list_break(&pp, &trunc, npages); 4363219089Spjd /* ignore pages past end of file */ 4364219089Spjd if (trunc) 4365219089Spjd pvn_write_done(trunc, flags); 4366219089Spjd len = zp->z_size - off; 4367219089Spjd } 4368219089Spjd 4369219089Spjd if (zfs_owner_overquota(zfsvfs, zp, B_FALSE) || 4370219089Spjd zfs_owner_overquota(zfsvfs, zp, B_TRUE)) { 4371219089Spjd err = EDQUOT; 4372219089Spjd goto out; 4373219089Spjd } 4374219089Spjdtop: 4375219089Spjd tx = dmu_tx_create(zfsvfs->z_os); 4376219089Spjd dmu_tx_hold_write(tx, zp->z_id, off, len); 4377219089Spjd 4378219089Spjd dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); 4379219089Spjd zfs_sa_upgrade_txholds(tx, zp); 4380219089Spjd err = dmu_tx_assign(tx, TXG_NOWAIT); 4381219089Spjd if (err != 0) { 4382219089Spjd if (err == ERESTART) { 4383219089Spjd dmu_tx_wait(tx); 4384219089Spjd dmu_tx_abort(tx); 4385219089Spjd goto top; 4386219089Spjd } 4387219089Spjd dmu_tx_abort(tx); 4388219089Spjd goto out; 4389219089Spjd } 4390219089Spjd 4391219089Spjd if (zp->z_blksz <= PAGESIZE) { 4392219089Spjd caddr_t va = zfs_map_page(pp, S_READ); 4393219089Spjd ASSERT3U(len, <=, PAGESIZE); 4394219089Spjd dmu_write(zfsvfs->z_os, zp->z_id, off, len, va, tx); 4395219089Spjd zfs_unmap_page(pp, va); 4396219089Spjd } else { 4397219089Spjd err = dmu_write_pages(zfsvfs->z_os, zp->z_id, off, len, pp, tx); 4398219089Spjd } 4399219089Spjd 4400219089Spjd if (err == 0) { 4401219089Spjd uint64_t mtime[2], ctime[2]; 4402219089Spjd sa_bulk_attr_t bulk[3]; 4403219089Spjd int count = 0; 4404219089Spjd 4405219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_MTIME(zfsvfs), NULL, 4406219089Spjd &mtime, 16); 4407219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_CTIME(zfsvfs), NULL, 4408219089Spjd &ctime, 16); 4409219089Spjd SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, 4410219089Spjd &zp->z_pflags, 8); 4411219089Spjd zfs_tstamp_update_setup(zp, CONTENT_MODIFIED, mtime, ctime, 4412219089Spjd B_TRUE); 4413219089Spjd zfs_log_write(zfsvfs->z_log, tx, TX_WRITE, zp, off, len, 0); 4414219089Spjd } 4415219089Spjd dmu_tx_commit(tx); 4416219089Spjd 4417219089Spjdout: 4418219089Spjd pvn_write_done(pp, (err ? B_ERROR : 0) | flags); 4419219089Spjd if (offp) 4420219089Spjd *offp = off; 4421219089Spjd if (lenp) 4422219089Spjd *lenp = len; 4423219089Spjd 4424219089Spjd return (err); 4425219089Spjd} 4426219089Spjd 4427219089Spjd/* 4428219089Spjd * Copy the portion of the file indicated from pages into the file. 4429219089Spjd * The pages are stored in a page list attached to the files vnode. 4430219089Spjd * 4431219089Spjd * IN: vp - vnode of file to push page data to. 4432219089Spjd * off - position in file to put data. 4433219089Spjd * len - amount of data to write. 4434219089Spjd * flags - flags to control the operation. 4435219089Spjd * cr - credentials of caller. 4436219089Spjd * ct - caller context. 4437219089Spjd * 4438219089Spjd * RETURN: 0 if success 4439219089Spjd * error code if failure 4440219089Spjd * 4441219089Spjd * Timestamps: 4442219089Spjd * vp - ctime|mtime updated 4443219089Spjd */ 4444185029Spjd/*ARGSUSED*/ 4445219089Spjdstatic int 4446219089Spjdzfs_putpage(vnode_t *vp, offset_t off, size_t len, int flags, cred_t *cr, 4447219089Spjd caller_context_t *ct) 4448219089Spjd{ 4449219089Spjd znode_t *zp = VTOZ(vp); 4450219089Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 4451219089Spjd page_t *pp; 4452219089Spjd size_t io_len; 4453219089Spjd u_offset_t io_off; 4454219089Spjd uint_t blksz; 4455219089Spjd rl_t *rl; 4456219089Spjd int error = 0; 4457219089Spjd 4458219089Spjd ZFS_ENTER(zfsvfs); 4459219089Spjd ZFS_VERIFY_ZP(zp); 4460219089Spjd 4461219089Spjd /* 4462219089Spjd * Align this request to the file block size in case we kluster. 4463219089Spjd * XXX - this can result in pretty aggresive locking, which can 4464219089Spjd * impact simultanious read/write access. One option might be 4465219089Spjd * to break up long requests (len == 0) into block-by-block 4466219089Spjd * operations to get narrower locking. 4467219089Spjd */ 4468219089Spjd blksz = zp->z_blksz; 4469219089Spjd if (ISP2(blksz)) 4470219089Spjd io_off = P2ALIGN_TYPED(off, blksz, u_offset_t); 4471219089Spjd else 4472219089Spjd io_off = 0; 4473219089Spjd if (len > 0 && ISP2(blksz)) 4474219089Spjd io_len = P2ROUNDUP_TYPED(len + (off - io_off), blksz, size_t); 4475219089Spjd else 4476219089Spjd io_len = 0; 4477219089Spjd 4478219089Spjd if (io_len == 0) { 4479219089Spjd /* 4480219089Spjd * Search the entire vp list for pages >= io_off. 4481219089Spjd */ 4482219089Spjd rl = zfs_range_lock(zp, io_off, UINT64_MAX, RL_WRITER); 4483219089Spjd error = pvn_vplist_dirty(vp, io_off, zfs_putapage, flags, cr); 4484219089Spjd goto out; 4485219089Spjd } 4486219089Spjd rl = zfs_range_lock(zp, io_off, io_len, RL_WRITER); 4487219089Spjd 4488219089Spjd if (off > zp->z_size) { 4489219089Spjd /* past end of file */ 4490219089Spjd zfs_range_unlock(rl); 4491219089Spjd ZFS_EXIT(zfsvfs); 4492219089Spjd return (0); 4493219089Spjd } 4494219089Spjd 4495219089Spjd len = MIN(io_len, P2ROUNDUP(zp->z_size, PAGESIZE) - io_off); 4496219089Spjd 4497219089Spjd for (off = io_off; io_off < off + len; io_off += io_len) { 4498219089Spjd if ((flags & B_INVAL) || ((flags & B_ASYNC) == 0)) { 4499219089Spjd pp = page_lookup(vp, io_off, 4500219089Spjd (flags & (B_INVAL | B_FREE)) ? SE_EXCL : SE_SHARED); 4501219089Spjd } else { 4502219089Spjd pp = page_lookup_nowait(vp, io_off, 4503219089Spjd (flags & B_FREE) ? SE_EXCL : SE_SHARED); 4504219089Spjd } 4505219089Spjd 4506219089Spjd if (pp != NULL && pvn_getdirty(pp, flags)) { 4507219089Spjd int err; 4508219089Spjd 4509219089Spjd /* 4510219089Spjd * Found a dirty page to push 4511219089Spjd */ 4512219089Spjd err = zfs_putapage(vp, pp, &io_off, &io_len, flags, cr); 4513219089Spjd if (err) 4514219089Spjd error = err; 4515219089Spjd } else { 4516219089Spjd io_len = PAGESIZE; 4517219089Spjd } 4518219089Spjd } 4519219089Spjdout: 4520219089Spjd zfs_range_unlock(rl); 4521219089Spjd if ((flags & B_ASYNC) == 0 || zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) 4522219089Spjd zil_commit(zfsvfs->z_log, zp->z_id); 4523219089Spjd ZFS_EXIT(zfsvfs); 4524219089Spjd return (error); 4525219089Spjd} 4526219089Spjd#endif /* sun */ 4527219089Spjd 4528219089Spjd/*ARGSUSED*/ 4529168962Spjdvoid 4530185029Spjdzfs_inactive(vnode_t *vp, cred_t *cr, caller_context_t *ct) 4531168404Spjd{ 4532168962Spjd znode_t *zp = VTOZ(vp); 4533168962Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 4534168962Spjd int error; 4535168404Spjd 4536185029Spjd rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER); 4537219089Spjd if (zp->z_sa_hdl == NULL) { 4538185029Spjd /* 4539185029Spjd * The fs has been unmounted, or we did a 4540185029Spjd * suspend/resume and this file no longer exists. 4541185029Spjd */ 4542168404Spjd VI_LOCK(vp); 4543219089Spjd ASSERT(vp->v_count <= 1); 4544219089Spjd vp->v_count = 0; 4545196299Spjd VI_UNLOCK(vp); 4546196299Spjd vrecycle(vp, curthread); 4547185029Spjd rw_exit(&zfsvfs->z_teardown_inactive_lock); 4548168962Spjd return; 4549168404Spjd } 4550168404Spjd 4551168404Spjd if (zp->z_atime_dirty && zp->z_unlinked == 0) { 4552168404Spjd dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os); 4553168404Spjd 4554219089Spjd dmu_tx_hold_sa(tx, zp->z_sa_hdl, B_FALSE); 4555219089Spjd zfs_sa_upgrade_txholds(tx, zp); 4556168404Spjd error = dmu_tx_assign(tx, TXG_WAIT); 4557168404Spjd if (error) { 4558168404Spjd dmu_tx_abort(tx); 4559168404Spjd } else { 4560168404Spjd mutex_enter(&zp->z_lock); 4561219089Spjd (void) sa_update(zp->z_sa_hdl, SA_ZPL_ATIME(zfsvfs), 4562219089Spjd (void *)&zp->z_atime, sizeof (zp->z_atime), tx); 4563168404Spjd zp->z_atime_dirty = 0; 4564168404Spjd mutex_exit(&zp->z_lock); 4565168404Spjd dmu_tx_commit(tx); 4566168404Spjd } 4567168404Spjd } 4568168404Spjd 4569168404Spjd zfs_zinactive(zp); 4570185029Spjd rw_exit(&zfsvfs->z_teardown_inactive_lock); 4571168404Spjd} 4572168404Spjd 4573219089Spjd#ifdef sun 4574219089Spjd/* 4575219089Spjd * Bounds-check the seek operation. 4576219089Spjd * 4577219089Spjd * IN: vp - vnode seeking within 4578219089Spjd * ooff - old file offset 4579219089Spjd * noffp - pointer to new file offset 4580219089Spjd * ct - caller context 4581219089Spjd * 4582219089Spjd * RETURN: 0 if success 4583219089Spjd * EINVAL if new offset invalid 4584219089Spjd */ 4585219089Spjd/* ARGSUSED */ 4586219089Spjdstatic int 4587219089Spjdzfs_seek(vnode_t *vp, offset_t ooff, offset_t *noffp, 4588219089Spjd caller_context_t *ct) 4589219089Spjd{ 4590219089Spjd if (vp->v_type == VDIR) 4591219089Spjd return (0); 4592219089Spjd return ((*noffp < 0 || *noffp > MAXOFFSET_T) ? EINVAL : 0); 4593219089Spjd} 4594219089Spjd 4595219089Spjd/* 4596219089Spjd * Pre-filter the generic locking function to trap attempts to place 4597219089Spjd * a mandatory lock on a memory mapped file. 4598219089Spjd */ 4599219089Spjdstatic int 4600219089Spjdzfs_frlock(vnode_t *vp, int cmd, flock64_t *bfp, int flag, offset_t offset, 4601219089Spjd flk_callback_t *flk_cbp, cred_t *cr, caller_context_t *ct) 4602219089Spjd{ 4603219089Spjd znode_t *zp = VTOZ(vp); 4604219089Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 4605219089Spjd 4606219089Spjd ZFS_ENTER(zfsvfs); 4607219089Spjd ZFS_VERIFY_ZP(zp); 4608219089Spjd 4609219089Spjd /* 4610219089Spjd * We are following the UFS semantics with respect to mapcnt 4611219089Spjd * here: If we see that the file is mapped already, then we will 4612219089Spjd * return an error, but we don't worry about races between this 4613219089Spjd * function and zfs_map(). 4614219089Spjd */ 4615219089Spjd if (zp->z_mapcnt > 0 && MANDMODE(zp->z_mode)) { 4616219089Spjd ZFS_EXIT(zfsvfs); 4617219089Spjd return (EAGAIN); 4618219089Spjd } 4619219089Spjd ZFS_EXIT(zfsvfs); 4620219089Spjd return (fs_frlock(vp, cmd, bfp, flag, offset, flk_cbp, cr, ct)); 4621219089Spjd} 4622219089Spjd 4623219089Spjd/* 4624219089Spjd * If we can't find a page in the cache, we will create a new page 4625219089Spjd * and fill it with file data. For efficiency, we may try to fill 4626219089Spjd * multiple pages at once (klustering) to fill up the supplied page 4627219089Spjd * list. Note that the pages to be filled are held with an exclusive 4628219089Spjd * lock to prevent access by other threads while they are being filled. 4629219089Spjd */ 4630219089Spjdstatic int 4631219089Spjdzfs_fillpage(vnode_t *vp, u_offset_t off, struct seg *seg, 4632219089Spjd caddr_t addr, page_t *pl[], size_t plsz, enum seg_rw rw) 4633219089Spjd{ 4634219089Spjd znode_t *zp = VTOZ(vp); 4635219089Spjd page_t *pp, *cur_pp; 4636219089Spjd objset_t *os = zp->z_zfsvfs->z_os; 4637219089Spjd u_offset_t io_off, total; 4638219089Spjd size_t io_len; 4639219089Spjd int err; 4640219089Spjd 4641219089Spjd if (plsz == PAGESIZE || zp->z_blksz <= PAGESIZE) { 4642219089Spjd /* 4643219089Spjd * We only have a single page, don't bother klustering 4644219089Spjd */ 4645219089Spjd io_off = off; 4646219089Spjd io_len = PAGESIZE; 4647219089Spjd pp = page_create_va(vp, io_off, io_len, 4648219089Spjd PG_EXCL | PG_WAIT, seg, addr); 4649219089Spjd } else { 4650219089Spjd /* 4651219089Spjd * Try to find enough pages to fill the page list 4652219089Spjd */ 4653219089Spjd pp = pvn_read_kluster(vp, off, seg, addr, &io_off, 4654219089Spjd &io_len, off, plsz, 0); 4655219089Spjd } 4656219089Spjd if (pp == NULL) { 4657219089Spjd /* 4658219089Spjd * The page already exists, nothing to do here. 4659219089Spjd */ 4660219089Spjd *pl = NULL; 4661219089Spjd return (0); 4662219089Spjd } 4663219089Spjd 4664219089Spjd /* 4665219089Spjd * Fill the pages in the kluster. 4666219089Spjd */ 4667219089Spjd cur_pp = pp; 4668219089Spjd for (total = io_off + io_len; io_off < total; io_off += PAGESIZE) { 4669219089Spjd caddr_t va; 4670219089Spjd 4671219089Spjd ASSERT3U(io_off, ==, cur_pp->p_offset); 4672219089Spjd va = zfs_map_page(cur_pp, S_WRITE); 4673219089Spjd err = dmu_read(os, zp->z_id, io_off, PAGESIZE, va, 4674219089Spjd DMU_READ_PREFETCH); 4675219089Spjd zfs_unmap_page(cur_pp, va); 4676219089Spjd if (err) { 4677219089Spjd /* On error, toss the entire kluster */ 4678219089Spjd pvn_read_done(pp, B_ERROR); 4679219089Spjd /* convert checksum errors into IO errors */ 4680219089Spjd if (err == ECKSUM) 4681219089Spjd err = EIO; 4682219089Spjd return (err); 4683219089Spjd } 4684219089Spjd cur_pp = cur_pp->p_next; 4685219089Spjd } 4686219089Spjd 4687219089Spjd /* 4688219089Spjd * Fill in the page list array from the kluster starting 4689219089Spjd * from the desired offset `off'. 4690219089Spjd * NOTE: the page list will always be null terminated. 4691219089Spjd */ 4692219089Spjd pvn_plist_init(pp, pl, plsz, off, io_len, rw); 4693219089Spjd ASSERT(pl == NULL || (*pl)->p_offset == off); 4694219089Spjd 4695219089Spjd return (0); 4696219089Spjd} 4697219089Spjd 4698219089Spjd/* 4699219089Spjd * Return pointers to the pages for the file region [off, off + len] 4700219089Spjd * in the pl array. If plsz is greater than len, this function may 4701219089Spjd * also return page pointers from after the specified region 4702219089Spjd * (i.e. the region [off, off + plsz]). These additional pages are 4703219089Spjd * only returned if they are already in the cache, or were created as 4704219089Spjd * part of a klustered read. 4705219089Spjd * 4706219089Spjd * IN: vp - vnode of file to get data from. 4707219089Spjd * off - position in file to get data from. 4708219089Spjd * len - amount of data to retrieve. 4709219089Spjd * plsz - length of provided page list. 4710219089Spjd * seg - segment to obtain pages for. 4711219089Spjd * addr - virtual address of fault. 4712219089Spjd * rw - mode of created pages. 4713219089Spjd * cr - credentials of caller. 4714219089Spjd * ct - caller context. 4715219089Spjd * 4716219089Spjd * OUT: protp - protection mode of created pages. 4717219089Spjd * pl - list of pages created. 4718219089Spjd * 4719219089Spjd * RETURN: 0 if success 4720219089Spjd * error code if failure 4721219089Spjd * 4722219089Spjd * Timestamps: 4723219089Spjd * vp - atime updated 4724219089Spjd */ 4725219089Spjd/* ARGSUSED */ 4726219089Spjdstatic int 4727219089Spjdzfs_getpage(vnode_t *vp, offset_t off, size_t len, uint_t *protp, 4728219089Spjd page_t *pl[], size_t plsz, struct seg *seg, caddr_t addr, 4729219089Spjd enum seg_rw rw, cred_t *cr, caller_context_t *ct) 4730219089Spjd{ 4731219089Spjd znode_t *zp = VTOZ(vp); 4732219089Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 4733219089Spjd page_t **pl0 = pl; 4734219089Spjd int err = 0; 4735219089Spjd 4736219089Spjd /* we do our own caching, faultahead is unnecessary */ 4737219089Spjd if (pl == NULL) 4738219089Spjd return (0); 4739219089Spjd else if (len > plsz) 4740219089Spjd len = plsz; 4741219089Spjd else 4742219089Spjd len = P2ROUNDUP(len, PAGESIZE); 4743219089Spjd ASSERT(plsz >= len); 4744219089Spjd 4745219089Spjd ZFS_ENTER(zfsvfs); 4746219089Spjd ZFS_VERIFY_ZP(zp); 4747219089Spjd 4748219089Spjd if (protp) 4749219089Spjd *protp = PROT_ALL; 4750219089Spjd 4751219089Spjd /* 4752219089Spjd * Loop through the requested range [off, off + len) looking 4753219089Spjd * for pages. If we don't find a page, we will need to create 4754219089Spjd * a new page and fill it with data from the file. 4755219089Spjd */ 4756219089Spjd while (len > 0) { 4757219089Spjd if (*pl = page_lookup(vp, off, SE_SHARED)) 4758219089Spjd *(pl+1) = NULL; 4759219089Spjd else if (err = zfs_fillpage(vp, off, seg, addr, pl, plsz, rw)) 4760219089Spjd goto out; 4761219089Spjd while (*pl) { 4762219089Spjd ASSERT3U((*pl)->p_offset, ==, off); 4763219089Spjd off += PAGESIZE; 4764219089Spjd addr += PAGESIZE; 4765219089Spjd if (len > 0) { 4766219089Spjd ASSERT3U(len, >=, PAGESIZE); 4767219089Spjd len -= PAGESIZE; 4768219089Spjd } 4769219089Spjd ASSERT3U(plsz, >=, PAGESIZE); 4770219089Spjd plsz -= PAGESIZE; 4771219089Spjd pl++; 4772219089Spjd } 4773219089Spjd } 4774219089Spjd 4775219089Spjd /* 4776219089Spjd * Fill out the page array with any pages already in the cache. 4777219089Spjd */ 4778219089Spjd while (plsz > 0 && 4779219089Spjd (*pl++ = page_lookup_nowait(vp, off, SE_SHARED))) { 4780219089Spjd off += PAGESIZE; 4781219089Spjd plsz -= PAGESIZE; 4782219089Spjd } 4783219089Spjdout: 4784219089Spjd if (err) { 4785219089Spjd /* 4786219089Spjd * Release any pages we have previously locked. 4787219089Spjd */ 4788219089Spjd while (pl > pl0) 4789219089Spjd page_unlock(*--pl); 4790219089Spjd } else { 4791219089Spjd ZFS_ACCESSTIME_STAMP(zfsvfs, zp); 4792219089Spjd } 4793219089Spjd 4794219089Spjd *pl = NULL; 4795219089Spjd 4796219089Spjd ZFS_EXIT(zfsvfs); 4797219089Spjd return (err); 4798219089Spjd} 4799219089Spjd 4800219089Spjd/* 4801219089Spjd * Request a memory map for a section of a file. This code interacts 4802219089Spjd * with common code and the VM system as follows: 4803219089Spjd * 4804219089Spjd * common code calls mmap(), which ends up in smmap_common() 4805219089Spjd * 4806219089Spjd * this calls VOP_MAP(), which takes you into (say) zfs 4807219089Spjd * 4808219089Spjd * zfs_map() calls as_map(), passing segvn_create() as the callback 4809219089Spjd * 4810219089Spjd * segvn_create() creates the new segment and calls VOP_ADDMAP() 4811219089Spjd * 4812219089Spjd * zfs_addmap() updates z_mapcnt 4813219089Spjd */ 4814219089Spjd/*ARGSUSED*/ 4815219089Spjdstatic int 4816219089Spjdzfs_map(vnode_t *vp, offset_t off, struct as *as, caddr_t *addrp, 4817219089Spjd size_t len, uchar_t prot, uchar_t maxprot, uint_t flags, cred_t *cr, 4818219089Spjd caller_context_t *ct) 4819219089Spjd{ 4820219089Spjd znode_t *zp = VTOZ(vp); 4821219089Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 4822219089Spjd segvn_crargs_t vn_a; 4823219089Spjd int error; 4824219089Spjd 4825219089Spjd ZFS_ENTER(zfsvfs); 4826219089Spjd ZFS_VERIFY_ZP(zp); 4827219089Spjd 4828219089Spjd if ((prot & PROT_WRITE) && (zp->z_pflags & 4829219089Spjd (ZFS_IMMUTABLE | ZFS_READONLY | ZFS_APPENDONLY))) { 4830219089Spjd ZFS_EXIT(zfsvfs); 4831219089Spjd return (EPERM); 4832219089Spjd } 4833219089Spjd 4834219089Spjd if ((prot & (PROT_READ | PROT_EXEC)) && 4835219089Spjd (zp->z_pflags & ZFS_AV_QUARANTINED)) { 4836219089Spjd ZFS_EXIT(zfsvfs); 4837219089Spjd return (EACCES); 4838219089Spjd } 4839219089Spjd 4840219089Spjd if (vp->v_flag & VNOMAP) { 4841219089Spjd ZFS_EXIT(zfsvfs); 4842219089Spjd return (ENOSYS); 4843219089Spjd } 4844219089Spjd 4845219089Spjd if (off < 0 || len > MAXOFFSET_T - off) { 4846219089Spjd ZFS_EXIT(zfsvfs); 4847219089Spjd return (ENXIO); 4848219089Spjd } 4849219089Spjd 4850219089Spjd if (vp->v_type != VREG) { 4851219089Spjd ZFS_EXIT(zfsvfs); 4852219089Spjd return (ENODEV); 4853219089Spjd } 4854219089Spjd 4855219089Spjd /* 4856219089Spjd * If file is locked, disallow mapping. 4857219089Spjd */ 4858219089Spjd if (MANDMODE(zp->z_mode) && vn_has_flocks(vp)) { 4859219089Spjd ZFS_EXIT(zfsvfs); 4860219089Spjd return (EAGAIN); 4861219089Spjd } 4862219089Spjd 4863219089Spjd as_rangelock(as); 4864219089Spjd error = choose_addr(as, addrp, len, off, ADDR_VACALIGN, flags); 4865219089Spjd if (error != 0) { 4866219089Spjd as_rangeunlock(as); 4867219089Spjd ZFS_EXIT(zfsvfs); 4868219089Spjd return (error); 4869219089Spjd } 4870219089Spjd 4871219089Spjd vn_a.vp = vp; 4872219089Spjd vn_a.offset = (u_offset_t)off; 4873219089Spjd vn_a.type = flags & MAP_TYPE; 4874219089Spjd vn_a.prot = prot; 4875219089Spjd vn_a.maxprot = maxprot; 4876219089Spjd vn_a.cred = cr; 4877219089Spjd vn_a.amp = NULL; 4878219089Spjd vn_a.flags = flags & ~MAP_TYPE; 4879219089Spjd vn_a.szc = 0; 4880219089Spjd vn_a.lgrp_mem_policy_flags = 0; 4881219089Spjd 4882219089Spjd error = as_map(as, *addrp, len, segvn_create, &vn_a); 4883219089Spjd 4884219089Spjd as_rangeunlock(as); 4885219089Spjd ZFS_EXIT(zfsvfs); 4886219089Spjd return (error); 4887219089Spjd} 4888219089Spjd 4889219089Spjd/* ARGSUSED */ 4890219089Spjdstatic int 4891219089Spjdzfs_addmap(vnode_t *vp, offset_t off, struct as *as, caddr_t addr, 4892219089Spjd size_t len, uchar_t prot, uchar_t maxprot, uint_t flags, cred_t *cr, 4893219089Spjd caller_context_t *ct) 4894219089Spjd{ 4895219089Spjd uint64_t pages = btopr(len); 4896219089Spjd 4897219089Spjd atomic_add_64(&VTOZ(vp)->z_mapcnt, pages); 4898219089Spjd return (0); 4899219089Spjd} 4900219089Spjd 4901219089Spjd/* 4902219089Spjd * The reason we push dirty pages as part of zfs_delmap() is so that we get a 4903219089Spjd * more accurate mtime for the associated file. Since we don't have a way of 4904219089Spjd * detecting when the data was actually modified, we have to resort to 4905219089Spjd * heuristics. If an explicit msync() is done, then we mark the mtime when the 4906219089Spjd * last page is pushed. The problem occurs when the msync() call is omitted, 4907219089Spjd * which by far the most common case: 4908219089Spjd * 4909219089Spjd * open() 4910219089Spjd * mmap() 4911219089Spjd * <modify memory> 4912219089Spjd * munmap() 4913219089Spjd * close() 4914219089Spjd * <time lapse> 4915219089Spjd * putpage() via fsflush 4916219089Spjd * 4917219089Spjd * If we wait until fsflush to come along, we can have a modification time that 4918219089Spjd * is some arbitrary point in the future. In order to prevent this in the 4919219089Spjd * common case, we flush pages whenever a (MAP_SHARED, PROT_WRITE) mapping is 4920219089Spjd * torn down. 4921219089Spjd */ 4922219089Spjd/* ARGSUSED */ 4923219089Spjdstatic int 4924219089Spjdzfs_delmap(vnode_t *vp, offset_t off, struct as *as, caddr_t addr, 4925219089Spjd size_t len, uint_t prot, uint_t maxprot, uint_t flags, cred_t *cr, 4926219089Spjd caller_context_t *ct) 4927219089Spjd{ 4928219089Spjd uint64_t pages = btopr(len); 4929219089Spjd 4930219089Spjd ASSERT3U(VTOZ(vp)->z_mapcnt, >=, pages); 4931219089Spjd atomic_add_64(&VTOZ(vp)->z_mapcnt, -pages); 4932219089Spjd 4933219089Spjd if ((flags & MAP_SHARED) && (prot & PROT_WRITE) && 4934219089Spjd vn_has_cached_data(vp)) 4935219089Spjd (void) VOP_PUTPAGE(vp, off, len, B_ASYNC, cr, ct); 4936219089Spjd 4937219089Spjd return (0); 4938219089Spjd} 4939219089Spjd 4940219089Spjd/* 4941219089Spjd * Free or allocate space in a file. Currently, this function only 4942219089Spjd * supports the `F_FREESP' command. However, this command is somewhat 4943219089Spjd * misnamed, as its functionality includes the ability to allocate as 4944219089Spjd * well as free space. 4945219089Spjd * 4946219089Spjd * IN: vp - vnode of file to free data in. 4947219089Spjd * cmd - action to take (only F_FREESP supported). 4948219089Spjd * bfp - section of file to free/alloc. 4949219089Spjd * flag - current file open mode flags. 4950219089Spjd * offset - current file offset. 4951219089Spjd * cr - credentials of caller [UNUSED]. 4952219089Spjd * ct - caller context. 4953219089Spjd * 4954219089Spjd * RETURN: 0 if success 4955219089Spjd * error code if failure 4956219089Spjd * 4957219089Spjd * Timestamps: 4958219089Spjd * vp - ctime|mtime updated 4959219089Spjd */ 4960219089Spjd/* ARGSUSED */ 4961219089Spjdstatic int 4962219089Spjdzfs_space(vnode_t *vp, int cmd, flock64_t *bfp, int flag, 4963219089Spjd offset_t offset, cred_t *cr, caller_context_t *ct) 4964219089Spjd{ 4965219089Spjd znode_t *zp = VTOZ(vp); 4966219089Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 4967219089Spjd uint64_t off, len; 4968219089Spjd int error; 4969219089Spjd 4970219089Spjd ZFS_ENTER(zfsvfs); 4971219089Spjd ZFS_VERIFY_ZP(zp); 4972219089Spjd 4973219089Spjd if (cmd != F_FREESP) { 4974219089Spjd ZFS_EXIT(zfsvfs); 4975219089Spjd return (EINVAL); 4976219089Spjd } 4977219089Spjd 4978219089Spjd if (error = convoff(vp, bfp, 0, offset)) { 4979219089Spjd ZFS_EXIT(zfsvfs); 4980219089Spjd return (error); 4981219089Spjd } 4982219089Spjd 4983219089Spjd if (bfp->l_len < 0) { 4984219089Spjd ZFS_EXIT(zfsvfs); 4985219089Spjd return (EINVAL); 4986219089Spjd } 4987219089Spjd 4988219089Spjd off = bfp->l_start; 4989219089Spjd len = bfp->l_len; /* 0 means from off to end of file */ 4990219089Spjd 4991219089Spjd error = zfs_freesp(zp, off, len, flag, TRUE); 4992219089Spjd 4993219089Spjd ZFS_EXIT(zfsvfs); 4994219089Spjd return (error); 4995219089Spjd} 4996219089Spjd#endif /* sun */ 4997219089Spjd 4998168404SpjdCTASSERT(sizeof(struct zfid_short) <= sizeof(struct fid)); 4999168404SpjdCTASSERT(sizeof(struct zfid_long) <= sizeof(struct fid)); 5000168404Spjd 5001185029Spjd/*ARGSUSED*/ 5002168404Spjdstatic int 5003185029Spjdzfs_fid(vnode_t *vp, fid_t *fidp, caller_context_t *ct) 5004168404Spjd{ 5005168404Spjd znode_t *zp = VTOZ(vp); 5006168404Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 5007185029Spjd uint32_t gen; 5008219089Spjd uint64_t gen64; 5009168404Spjd uint64_t object = zp->z_id; 5010168404Spjd zfid_short_t *zfid; 5011219089Spjd int size, i, error; 5012168404Spjd 5013168404Spjd ZFS_ENTER(zfsvfs); 5014185029Spjd ZFS_VERIFY_ZP(zp); 5015168404Spjd 5016219089Spjd if ((error = sa_lookup(zp->z_sa_hdl, SA_ZPL_GEN(zfsvfs), 5017219089Spjd &gen64, sizeof (uint64_t))) != 0) { 5018219089Spjd ZFS_EXIT(zfsvfs); 5019219089Spjd return (error); 5020219089Spjd } 5021219089Spjd 5022219089Spjd gen = (uint32_t)gen64; 5023219089Spjd 5024168404Spjd size = (zfsvfs->z_parent != zfsvfs) ? LONG_FID_LEN : SHORT_FID_LEN; 5025168404Spjd fidp->fid_len = size; 5026168404Spjd 5027168404Spjd zfid = (zfid_short_t *)fidp; 5028168404Spjd 5029168404Spjd zfid->zf_len = size; 5030168404Spjd 5031168404Spjd for (i = 0; i < sizeof (zfid->zf_object); i++) 5032168404Spjd zfid->zf_object[i] = (uint8_t)(object >> (8 * i)); 5033168404Spjd 5034168404Spjd /* Must have a non-zero generation number to distinguish from .zfs */ 5035168404Spjd if (gen == 0) 5036168404Spjd gen = 1; 5037168404Spjd for (i = 0; i < sizeof (zfid->zf_gen); i++) 5038168404Spjd zfid->zf_gen[i] = (uint8_t)(gen >> (8 * i)); 5039168404Spjd 5040168404Spjd if (size == LONG_FID_LEN) { 5041168404Spjd uint64_t objsetid = dmu_objset_id(zfsvfs->z_os); 5042169023Spjd zfid_long_t *zlfid; 5043168404Spjd 5044168404Spjd zlfid = (zfid_long_t *)fidp; 5045168404Spjd 5046168404Spjd for (i = 0; i < sizeof (zlfid->zf_setid); i++) 5047168404Spjd zlfid->zf_setid[i] = (uint8_t)(objsetid >> (8 * i)); 5048168404Spjd 5049168404Spjd /* XXX - this should be the generation number for the objset */ 5050168404Spjd for (i = 0; i < sizeof (zlfid->zf_setgen); i++) 5051168404Spjd zlfid->zf_setgen[i] = 0; 5052168404Spjd } 5053168404Spjd 5054168404Spjd ZFS_EXIT(zfsvfs); 5055168404Spjd return (0); 5056168404Spjd} 5057168404Spjd 5058168404Spjdstatic int 5059185029Spjdzfs_pathconf(vnode_t *vp, int cmd, ulong_t *valp, cred_t *cr, 5060185029Spjd caller_context_t *ct) 5061168404Spjd{ 5062168404Spjd znode_t *zp, *xzp; 5063168404Spjd zfsvfs_t *zfsvfs; 5064168404Spjd zfs_dirlock_t *dl; 5065168404Spjd int error; 5066168404Spjd 5067168404Spjd switch (cmd) { 5068168404Spjd case _PC_LINK_MAX: 5069168404Spjd *valp = INT_MAX; 5070168404Spjd return (0); 5071168404Spjd 5072168404Spjd case _PC_FILESIZEBITS: 5073168404Spjd *valp = 64; 5074168404Spjd return (0); 5075219089Spjd#ifdef sun 5076168404Spjd case _PC_XATTR_EXISTS: 5077168404Spjd zp = VTOZ(vp); 5078168404Spjd zfsvfs = zp->z_zfsvfs; 5079168404Spjd ZFS_ENTER(zfsvfs); 5080185029Spjd ZFS_VERIFY_ZP(zp); 5081168404Spjd *valp = 0; 5082168404Spjd error = zfs_dirent_lock(&dl, zp, "", &xzp, 5083185029Spjd ZXATTR | ZEXISTS | ZSHARED, NULL, NULL); 5084168404Spjd if (error == 0) { 5085168404Spjd zfs_dirent_unlock(dl); 5086168404Spjd if (!zfs_dirempty(xzp)) 5087168404Spjd *valp = 1; 5088168404Spjd VN_RELE(ZTOV(xzp)); 5089168404Spjd } else if (error == ENOENT) { 5090168404Spjd /* 5091168404Spjd * If there aren't extended attributes, it's the 5092168404Spjd * same as having zero of them. 5093168404Spjd */ 5094168404Spjd error = 0; 5095168404Spjd } 5096168404Spjd ZFS_EXIT(zfsvfs); 5097168404Spjd return (error); 5098168404Spjd 5099219089Spjd case _PC_SATTR_ENABLED: 5100219089Spjd case _PC_SATTR_EXISTS: 5101219089Spjd *valp = vfs_has_feature(vp->v_vfsp, VFSFT_SYSATTR_VIEWS) && 5102219089Spjd (vp->v_type == VREG || vp->v_type == VDIR); 5103219089Spjd return (0); 5104219089Spjd 5105219089Spjd case _PC_ACCESS_FILTERING: 5106219089Spjd *valp = vfs_has_feature(vp->v_vfsp, VFSFT_ACCESS_FILTER) && 5107219089Spjd vp->v_type == VDIR; 5108219089Spjd return (0); 5109219089Spjd 5110219089Spjd case _PC_ACL_ENABLED: 5111219089Spjd *valp = _ACL_ACE_ENABLED; 5112219089Spjd return (0); 5113219089Spjd#endif /* sun */ 5114219089Spjd case _PC_MIN_HOLE_SIZE: 5115219089Spjd *valp = (int)SPA_MINBLOCKSIZE; 5116219089Spjd return (0); 5117219089Spjd#ifdef sun 5118219089Spjd case _PC_TIMESTAMP_RESOLUTION: 5119219089Spjd /* nanosecond timestamp resolution */ 5120219089Spjd *valp = 1L; 5121219089Spjd return (0); 5122219089Spjd#endif /* sun */ 5123168404Spjd case _PC_ACL_EXTENDED: 5124196949Strasz *valp = 0; 5125168404Spjd return (0); 5126168404Spjd 5127196949Strasz case _PC_ACL_NFS4: 5128196949Strasz *valp = 1; 5129196949Strasz return (0); 5130196949Strasz 5131196949Strasz case _PC_ACL_PATH_MAX: 5132196949Strasz *valp = ACL_MAX_ENTRIES; 5133196949Strasz return (0); 5134196949Strasz 5135168404Spjd default: 5136168962Spjd return (EOPNOTSUPP); 5137168404Spjd } 5138168404Spjd} 5139168404Spjd 5140168404Spjd/*ARGSUSED*/ 5141168404Spjdstatic int 5142185029Spjdzfs_getsecattr(vnode_t *vp, vsecattr_t *vsecp, int flag, cred_t *cr, 5143185029Spjd caller_context_t *ct) 5144168404Spjd{ 5145168404Spjd znode_t *zp = VTOZ(vp); 5146168404Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 5147168404Spjd int error; 5148185029Spjd boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; 5149168404Spjd 5150168404Spjd ZFS_ENTER(zfsvfs); 5151185029Spjd ZFS_VERIFY_ZP(zp); 5152185029Spjd error = zfs_getacl(zp, vsecp, skipaclchk, cr); 5153168404Spjd ZFS_EXIT(zfsvfs); 5154168404Spjd 5155168404Spjd return (error); 5156168404Spjd} 5157168404Spjd 5158168404Spjd/*ARGSUSED*/ 5159168404Spjdstatic int 5160185029Spjdzfs_setsecattr(vnode_t *vp, vsecattr_t *vsecp, int flag, cred_t *cr, 5161185029Spjd caller_context_t *ct) 5162168404Spjd{ 5163168404Spjd znode_t *zp = VTOZ(vp); 5164168404Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 5165168404Spjd int error; 5166185029Spjd boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; 5167219089Spjd zilog_t *zilog = zfsvfs->z_log; 5168168404Spjd 5169168404Spjd ZFS_ENTER(zfsvfs); 5170185029Spjd ZFS_VERIFY_ZP(zp); 5171219089Spjd 5172185029Spjd error = zfs_setacl(zp, vsecp, skipaclchk, cr); 5173219089Spjd 5174219089Spjd if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS) 5175219089Spjd zil_commit(zilog, 0); 5176219089Spjd 5177168404Spjd ZFS_EXIT(zfsvfs); 5178168404Spjd return (error); 5179168404Spjd} 5180168404Spjd 5181219089Spjd#ifdef sun 5182219089Spjd/* 5183219089Spjd * Tunable, both must be a power of 2. 5184219089Spjd * 5185219089Spjd * zcr_blksz_min: the smallest read we may consider to loan out an arcbuf 5186219089Spjd * zcr_blksz_max: if set to less than the file block size, allow loaning out of 5187219089Spjd * an arcbuf for a partial block read 5188219089Spjd */ 5189219089Spjdint zcr_blksz_min = (1 << 10); /* 1K */ 5190219089Spjdint zcr_blksz_max = (1 << 17); /* 128K */ 5191219089Spjd 5192219089Spjd/*ARGSUSED*/ 5193168962Spjdstatic int 5194219089Spjdzfs_reqzcbuf(vnode_t *vp, enum uio_rw ioflag, xuio_t *xuio, cred_t *cr, 5195219089Spjd caller_context_t *ct) 5196219089Spjd{ 5197219089Spjd znode_t *zp = VTOZ(vp); 5198219089Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 5199219089Spjd int max_blksz = zfsvfs->z_max_blksz; 5200219089Spjd uio_t *uio = &xuio->xu_uio; 5201219089Spjd ssize_t size = uio->uio_resid; 5202219089Spjd offset_t offset = uio->uio_loffset; 5203219089Spjd int blksz; 5204219089Spjd int fullblk, i; 5205219089Spjd arc_buf_t *abuf; 5206219089Spjd ssize_t maxsize; 5207219089Spjd int preamble, postamble; 5208219089Spjd 5209219089Spjd if (xuio->xu_type != UIOTYPE_ZEROCOPY) 5210219089Spjd return (EINVAL); 5211219089Spjd 5212219089Spjd ZFS_ENTER(zfsvfs); 5213219089Spjd ZFS_VERIFY_ZP(zp); 5214219089Spjd switch (ioflag) { 5215219089Spjd case UIO_WRITE: 5216219089Spjd /* 5217219089Spjd * Loan out an arc_buf for write if write size is bigger than 5218219089Spjd * max_blksz, and the file's block size is also max_blksz. 5219219089Spjd */ 5220219089Spjd blksz = max_blksz; 5221219089Spjd if (size < blksz || zp->z_blksz != blksz) { 5222219089Spjd ZFS_EXIT(zfsvfs); 5223219089Spjd return (EINVAL); 5224219089Spjd } 5225219089Spjd /* 5226219089Spjd * Caller requests buffers for write before knowing where the 5227219089Spjd * write offset might be (e.g. NFS TCP write). 5228219089Spjd */ 5229219089Spjd if (offset == -1) { 5230219089Spjd preamble = 0; 5231219089Spjd } else { 5232219089Spjd preamble = P2PHASE(offset, blksz); 5233219089Spjd if (preamble) { 5234219089Spjd preamble = blksz - preamble; 5235219089Spjd size -= preamble; 5236219089Spjd } 5237219089Spjd } 5238219089Spjd 5239219089Spjd postamble = P2PHASE(size, blksz); 5240219089Spjd size -= postamble; 5241219089Spjd 5242219089Spjd fullblk = size / blksz; 5243219089Spjd (void) dmu_xuio_init(xuio, 5244219089Spjd (preamble != 0) + fullblk + (postamble != 0)); 5245219089Spjd DTRACE_PROBE3(zfs_reqzcbuf_align, int, preamble, 5246219089Spjd int, postamble, int, 5247219089Spjd (preamble != 0) + fullblk + (postamble != 0)); 5248219089Spjd 5249219089Spjd /* 5250219089Spjd * Have to fix iov base/len for partial buffers. They 5251219089Spjd * currently represent full arc_buf's. 5252219089Spjd */ 5253219089Spjd if (preamble) { 5254219089Spjd /* data begins in the middle of the arc_buf */ 5255219089Spjd abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), 5256219089Spjd blksz); 5257219089Spjd ASSERT(abuf); 5258219089Spjd (void) dmu_xuio_add(xuio, abuf, 5259219089Spjd blksz - preamble, preamble); 5260219089Spjd } 5261219089Spjd 5262219089Spjd for (i = 0; i < fullblk; i++) { 5263219089Spjd abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), 5264219089Spjd blksz); 5265219089Spjd ASSERT(abuf); 5266219089Spjd (void) dmu_xuio_add(xuio, abuf, 0, blksz); 5267219089Spjd } 5268219089Spjd 5269219089Spjd if (postamble) { 5270219089Spjd /* data ends in the middle of the arc_buf */ 5271219089Spjd abuf = dmu_request_arcbuf(sa_get_db(zp->z_sa_hdl), 5272219089Spjd blksz); 5273219089Spjd ASSERT(abuf); 5274219089Spjd (void) dmu_xuio_add(xuio, abuf, 0, postamble); 5275219089Spjd } 5276219089Spjd break; 5277219089Spjd case UIO_READ: 5278219089Spjd /* 5279219089Spjd * Loan out an arc_buf for read if the read size is larger than 5280219089Spjd * the current file block size. Block alignment is not 5281219089Spjd * considered. Partial arc_buf will be loaned out for read. 5282219089Spjd */ 5283219089Spjd blksz = zp->z_blksz; 5284219089Spjd if (blksz < zcr_blksz_min) 5285219089Spjd blksz = zcr_blksz_min; 5286219089Spjd if (blksz > zcr_blksz_max) 5287219089Spjd blksz = zcr_blksz_max; 5288219089Spjd /* avoid potential complexity of dealing with it */ 5289219089Spjd if (blksz > max_blksz) { 5290219089Spjd ZFS_EXIT(zfsvfs); 5291219089Spjd return (EINVAL); 5292219089Spjd } 5293219089Spjd 5294219089Spjd maxsize = zp->z_size - uio->uio_loffset; 5295219089Spjd if (size > maxsize) 5296219089Spjd size = maxsize; 5297219089Spjd 5298219089Spjd if (size < blksz || vn_has_cached_data(vp)) { 5299219089Spjd ZFS_EXIT(zfsvfs); 5300219089Spjd return (EINVAL); 5301219089Spjd } 5302219089Spjd break; 5303219089Spjd default: 5304219089Spjd ZFS_EXIT(zfsvfs); 5305219089Spjd return (EINVAL); 5306219089Spjd } 5307219089Spjd 5308219089Spjd uio->uio_extflg = UIO_XUIO; 5309219089Spjd XUIO_XUZC_RW(xuio) = ioflag; 5310219089Spjd ZFS_EXIT(zfsvfs); 5311219089Spjd return (0); 5312219089Spjd} 5313219089Spjd 5314219089Spjd/*ARGSUSED*/ 5315219089Spjdstatic int 5316219089Spjdzfs_retzcbuf(vnode_t *vp, xuio_t *xuio, cred_t *cr, caller_context_t *ct) 5317219089Spjd{ 5318219089Spjd int i; 5319219089Spjd arc_buf_t *abuf; 5320219089Spjd int ioflag = XUIO_XUZC_RW(xuio); 5321219089Spjd 5322219089Spjd ASSERT(xuio->xu_type == UIOTYPE_ZEROCOPY); 5323219089Spjd 5324219089Spjd i = dmu_xuio_cnt(xuio); 5325219089Spjd while (i-- > 0) { 5326219089Spjd abuf = dmu_xuio_arcbuf(xuio, i); 5327219089Spjd /* 5328219089Spjd * if abuf == NULL, it must be a write buffer 5329219089Spjd * that has been returned in zfs_write(). 5330219089Spjd */ 5331219089Spjd if (abuf) 5332219089Spjd dmu_return_arcbuf(abuf); 5333219089Spjd ASSERT(abuf || ioflag == UIO_WRITE); 5334219089Spjd } 5335219089Spjd 5336219089Spjd dmu_xuio_fini(xuio); 5337219089Spjd return (0); 5338219089Spjd} 5339219089Spjd 5340219089Spjd/* 5341219089Spjd * Predeclare these here so that the compiler assumes that 5342219089Spjd * this is an "old style" function declaration that does 5343219089Spjd * not include arguments => we won't get type mismatch errors 5344219089Spjd * in the initializations that follow. 5345219089Spjd */ 5346219089Spjdstatic int zfs_inval(); 5347219089Spjdstatic int zfs_isdir(); 5348219089Spjd 5349219089Spjdstatic int 5350219089Spjdzfs_inval() 5351219089Spjd{ 5352219089Spjd return (EINVAL); 5353219089Spjd} 5354219089Spjd 5355219089Spjdstatic int 5356219089Spjdzfs_isdir() 5357219089Spjd{ 5358219089Spjd return (EISDIR); 5359219089Spjd} 5360219089Spjd/* 5361219089Spjd * Directory vnode operations template 5362219089Spjd */ 5363219089Spjdvnodeops_t *zfs_dvnodeops; 5364219089Spjdconst fs_operation_def_t zfs_dvnodeops_template[] = { 5365219089Spjd VOPNAME_OPEN, { .vop_open = zfs_open }, 5366219089Spjd VOPNAME_CLOSE, { .vop_close = zfs_close }, 5367219089Spjd VOPNAME_READ, { .error = zfs_isdir }, 5368219089Spjd VOPNAME_WRITE, { .error = zfs_isdir }, 5369219089Spjd VOPNAME_IOCTL, { .vop_ioctl = zfs_ioctl }, 5370219089Spjd VOPNAME_GETATTR, { .vop_getattr = zfs_getattr }, 5371219089Spjd VOPNAME_SETATTR, { .vop_setattr = zfs_setattr }, 5372219089Spjd VOPNAME_ACCESS, { .vop_access = zfs_access }, 5373219089Spjd VOPNAME_LOOKUP, { .vop_lookup = zfs_lookup }, 5374219089Spjd VOPNAME_CREATE, { .vop_create = zfs_create }, 5375219089Spjd VOPNAME_REMOVE, { .vop_remove = zfs_remove }, 5376219089Spjd VOPNAME_LINK, { .vop_link = zfs_link }, 5377219089Spjd VOPNAME_RENAME, { .vop_rename = zfs_rename }, 5378219089Spjd VOPNAME_MKDIR, { .vop_mkdir = zfs_mkdir }, 5379219089Spjd VOPNAME_RMDIR, { .vop_rmdir = zfs_rmdir }, 5380219089Spjd VOPNAME_READDIR, { .vop_readdir = zfs_readdir }, 5381219089Spjd VOPNAME_SYMLINK, { .vop_symlink = zfs_symlink }, 5382219089Spjd VOPNAME_FSYNC, { .vop_fsync = zfs_fsync }, 5383219089Spjd VOPNAME_INACTIVE, { .vop_inactive = zfs_inactive }, 5384219089Spjd VOPNAME_FID, { .vop_fid = zfs_fid }, 5385219089Spjd VOPNAME_SEEK, { .vop_seek = zfs_seek }, 5386219089Spjd VOPNAME_PATHCONF, { .vop_pathconf = zfs_pathconf }, 5387219089Spjd VOPNAME_GETSECATTR, { .vop_getsecattr = zfs_getsecattr }, 5388219089Spjd VOPNAME_SETSECATTR, { .vop_setsecattr = zfs_setsecattr }, 5389219089Spjd VOPNAME_VNEVENT, { .vop_vnevent = fs_vnevent_support }, 5390219089Spjd NULL, NULL 5391219089Spjd}; 5392219089Spjd 5393219089Spjd/* 5394219089Spjd * Regular file vnode operations template 5395219089Spjd */ 5396219089Spjdvnodeops_t *zfs_fvnodeops; 5397219089Spjdconst fs_operation_def_t zfs_fvnodeops_template[] = { 5398219089Spjd VOPNAME_OPEN, { .vop_open = zfs_open }, 5399219089Spjd VOPNAME_CLOSE, { .vop_close = zfs_close }, 5400219089Spjd VOPNAME_READ, { .vop_read = zfs_read }, 5401219089Spjd VOPNAME_WRITE, { .vop_write = zfs_write }, 5402219089Spjd VOPNAME_IOCTL, { .vop_ioctl = zfs_ioctl }, 5403219089Spjd VOPNAME_GETATTR, { .vop_getattr = zfs_getattr }, 5404219089Spjd VOPNAME_SETATTR, { .vop_setattr = zfs_setattr }, 5405219089Spjd VOPNAME_ACCESS, { .vop_access = zfs_access }, 5406219089Spjd VOPNAME_LOOKUP, { .vop_lookup = zfs_lookup }, 5407219089Spjd VOPNAME_RENAME, { .vop_rename = zfs_rename }, 5408219089Spjd VOPNAME_FSYNC, { .vop_fsync = zfs_fsync }, 5409219089Spjd VOPNAME_INACTIVE, { .vop_inactive = zfs_inactive }, 5410219089Spjd VOPNAME_FID, { .vop_fid = zfs_fid }, 5411219089Spjd VOPNAME_SEEK, { .vop_seek = zfs_seek }, 5412219089Spjd VOPNAME_FRLOCK, { .vop_frlock = zfs_frlock }, 5413219089Spjd VOPNAME_SPACE, { .vop_space = zfs_space }, 5414219089Spjd VOPNAME_GETPAGE, { .vop_getpage = zfs_getpage }, 5415219089Spjd VOPNAME_PUTPAGE, { .vop_putpage = zfs_putpage }, 5416219089Spjd VOPNAME_MAP, { .vop_map = zfs_map }, 5417219089Spjd VOPNAME_ADDMAP, { .vop_addmap = zfs_addmap }, 5418219089Spjd VOPNAME_DELMAP, { .vop_delmap = zfs_delmap }, 5419219089Spjd VOPNAME_PATHCONF, { .vop_pathconf = zfs_pathconf }, 5420219089Spjd VOPNAME_GETSECATTR, { .vop_getsecattr = zfs_getsecattr }, 5421219089Spjd VOPNAME_SETSECATTR, { .vop_setsecattr = zfs_setsecattr }, 5422219089Spjd VOPNAME_VNEVENT, { .vop_vnevent = fs_vnevent_support }, 5423219089Spjd VOPNAME_REQZCBUF, { .vop_reqzcbuf = zfs_reqzcbuf }, 5424219089Spjd VOPNAME_RETZCBUF, { .vop_retzcbuf = zfs_retzcbuf }, 5425219089Spjd NULL, NULL 5426219089Spjd}; 5427219089Spjd 5428219089Spjd/* 5429219089Spjd * Symbolic link vnode operations template 5430219089Spjd */ 5431219089Spjdvnodeops_t *zfs_symvnodeops; 5432219089Spjdconst fs_operation_def_t zfs_symvnodeops_template[] = { 5433219089Spjd VOPNAME_GETATTR, { .vop_getattr = zfs_getattr }, 5434219089Spjd VOPNAME_SETATTR, { .vop_setattr = zfs_setattr }, 5435219089Spjd VOPNAME_ACCESS, { .vop_access = zfs_access }, 5436219089Spjd VOPNAME_RENAME, { .vop_rename = zfs_rename }, 5437219089Spjd VOPNAME_READLINK, { .vop_readlink = zfs_readlink }, 5438219089Spjd VOPNAME_INACTIVE, { .vop_inactive = zfs_inactive }, 5439219089Spjd VOPNAME_FID, { .vop_fid = zfs_fid }, 5440219089Spjd VOPNAME_PATHCONF, { .vop_pathconf = zfs_pathconf }, 5441219089Spjd VOPNAME_VNEVENT, { .vop_vnevent = fs_vnevent_support }, 5442219089Spjd NULL, NULL 5443219089Spjd}; 5444219089Spjd 5445219089Spjd/* 5446219089Spjd * special share hidden files vnode operations template 5447219089Spjd */ 5448219089Spjdvnodeops_t *zfs_sharevnodeops; 5449219089Spjdconst fs_operation_def_t zfs_sharevnodeops_template[] = { 5450219089Spjd VOPNAME_GETATTR, { .vop_getattr = zfs_getattr }, 5451219089Spjd VOPNAME_ACCESS, { .vop_access = zfs_access }, 5452219089Spjd VOPNAME_INACTIVE, { .vop_inactive = zfs_inactive }, 5453219089Spjd VOPNAME_FID, { .vop_fid = zfs_fid }, 5454219089Spjd VOPNAME_PATHCONF, { .vop_pathconf = zfs_pathconf }, 5455219089Spjd VOPNAME_GETSECATTR, { .vop_getsecattr = zfs_getsecattr }, 5456219089Spjd VOPNAME_SETSECATTR, { .vop_setsecattr = zfs_setsecattr }, 5457219089Spjd VOPNAME_VNEVENT, { .vop_vnevent = fs_vnevent_support }, 5458219089Spjd NULL, NULL 5459219089Spjd}; 5460219089Spjd 5461219089Spjd/* 5462219089Spjd * Extended attribute directory vnode operations template 5463219089Spjd * This template is identical to the directory vnodes 5464219089Spjd * operation template except for restricted operations: 5465219089Spjd * VOP_MKDIR() 5466219089Spjd * VOP_SYMLINK() 5467219089Spjd * Note that there are other restrictions embedded in: 5468219089Spjd * zfs_create() - restrict type to VREG 5469219089Spjd * zfs_link() - no links into/out of attribute space 5470219089Spjd * zfs_rename() - no moves into/out of attribute space 5471219089Spjd */ 5472219089Spjdvnodeops_t *zfs_xdvnodeops; 5473219089Spjdconst fs_operation_def_t zfs_xdvnodeops_template[] = { 5474219089Spjd VOPNAME_OPEN, { .vop_open = zfs_open }, 5475219089Spjd VOPNAME_CLOSE, { .vop_close = zfs_close }, 5476219089Spjd VOPNAME_IOCTL, { .vop_ioctl = zfs_ioctl }, 5477219089Spjd VOPNAME_GETATTR, { .vop_getattr = zfs_getattr }, 5478219089Spjd VOPNAME_SETATTR, { .vop_setattr = zfs_setattr }, 5479219089Spjd VOPNAME_ACCESS, { .vop_access = zfs_access }, 5480219089Spjd VOPNAME_LOOKUP, { .vop_lookup = zfs_lookup }, 5481219089Spjd VOPNAME_CREATE, { .vop_create = zfs_create }, 5482219089Spjd VOPNAME_REMOVE, { .vop_remove = zfs_remove }, 5483219089Spjd VOPNAME_LINK, { .vop_link = zfs_link }, 5484219089Spjd VOPNAME_RENAME, { .vop_rename = zfs_rename }, 5485219089Spjd VOPNAME_MKDIR, { .error = zfs_inval }, 5486219089Spjd VOPNAME_RMDIR, { .vop_rmdir = zfs_rmdir }, 5487219089Spjd VOPNAME_READDIR, { .vop_readdir = zfs_readdir }, 5488219089Spjd VOPNAME_SYMLINK, { .error = zfs_inval }, 5489219089Spjd VOPNAME_FSYNC, { .vop_fsync = zfs_fsync }, 5490219089Spjd VOPNAME_INACTIVE, { .vop_inactive = zfs_inactive }, 5491219089Spjd VOPNAME_FID, { .vop_fid = zfs_fid }, 5492219089Spjd VOPNAME_SEEK, { .vop_seek = zfs_seek }, 5493219089Spjd VOPNAME_PATHCONF, { .vop_pathconf = zfs_pathconf }, 5494219089Spjd VOPNAME_GETSECATTR, { .vop_getsecattr = zfs_getsecattr }, 5495219089Spjd VOPNAME_SETSECATTR, { .vop_setsecattr = zfs_setsecattr }, 5496219089Spjd VOPNAME_VNEVENT, { .vop_vnevent = fs_vnevent_support }, 5497219089Spjd NULL, NULL 5498219089Spjd}; 5499219089Spjd 5500219089Spjd/* 5501219089Spjd * Error vnode operations template 5502219089Spjd */ 5503219089Spjdvnodeops_t *zfs_evnodeops; 5504219089Spjdconst fs_operation_def_t zfs_evnodeops_template[] = { 5505219089Spjd VOPNAME_INACTIVE, { .vop_inactive = zfs_inactive }, 5506219089Spjd VOPNAME_PATHCONF, { .vop_pathconf = zfs_pathconf }, 5507219089Spjd NULL, NULL 5508219089Spjd}; 5509219089Spjd#endif /* sun */ 5510219089Spjd 5511219089Spjdstatic int 5512213673Spjdioflags(int ioflags) 5513213673Spjd{ 5514213673Spjd int flags = 0; 5515213673Spjd 5516213673Spjd if (ioflags & IO_APPEND) 5517213673Spjd flags |= FAPPEND; 5518213673Spjd if (ioflags & IO_NDELAY) 5519213673Spjd flags |= FNONBLOCK; 5520213673Spjd if (ioflags & IO_SYNC) 5521213673Spjd flags |= (FSYNC | FDSYNC | FRSYNC); 5522213673Spjd 5523213673Spjd return (flags); 5524213673Spjd} 5525213673Spjd 5526213673Spjdstatic int 5527213937Savgzfs_getpages(struct vnode *vp, vm_page_t *m, int count, int reqpage) 5528213937Savg{ 5529213937Savg znode_t *zp = VTOZ(vp); 5530213937Savg zfsvfs_t *zfsvfs = zp->z_zfsvfs; 5531213937Savg objset_t *os = zp->z_zfsvfs->z_os; 5532213937Savg vm_page_t mreq; 5533213937Savg vm_object_t object; 5534213937Savg caddr_t va; 5535213937Savg struct sf_buf *sf; 5536213937Savg int i, error; 5537213937Savg int pcount, size; 5538213937Savg 5539213937Savg ZFS_ENTER(zfsvfs); 5540213937Savg ZFS_VERIFY_ZP(zp); 5541213937Savg 5542213937Savg pcount = round_page(count) / PAGE_SIZE; 5543213937Savg mreq = m[reqpage]; 5544213937Savg object = mreq->object; 5545213937Savg error = 0; 5546213937Savg 5547213937Savg KASSERT(vp->v_object == object, ("mismatching object")); 5548213937Savg 5549213937Savg VM_OBJECT_LOCK(object); 5550213937Savg 5551213937Savg for (i = 0; i < pcount; i++) { 5552213937Savg if (i != reqpage) { 5553213937Savg vm_page_lock(m[i]); 5554213937Savg vm_page_free(m[i]); 5555213937Savg vm_page_unlock(m[i]); 5556213937Savg } 5557213937Savg } 5558213937Savg 5559213937Savg if (mreq->valid) { 5560213937Savg if (mreq->valid != VM_PAGE_BITS_ALL) 5561213937Savg vm_page_zero_invalid(mreq, TRUE); 5562213937Savg VM_OBJECT_UNLOCK(object); 5563213937Savg ZFS_EXIT(zfsvfs); 5564213937Savg return (VM_PAGER_OK); 5565213937Savg } 5566213937Savg 5567213937Savg PCPU_INC(cnt.v_vnodein); 5568213937Savg PCPU_INC(cnt.v_vnodepgsin); 5569213937Savg 5570213937Savg if (IDX_TO_OFF(mreq->pindex) >= object->un_pager.vnp.vnp_size) { 5571213937Savg VM_OBJECT_UNLOCK(object); 5572213937Savg ZFS_EXIT(zfsvfs); 5573213937Savg return (VM_PAGER_BAD); 5574213937Savg } 5575213937Savg 5576213937Savg size = PAGE_SIZE; 5577213937Savg if (IDX_TO_OFF(mreq->pindex) + size > object->un_pager.vnp.vnp_size) 5578213937Savg size = object->un_pager.vnp.vnp_size - IDX_TO_OFF(mreq->pindex); 5579213937Savg 5580213937Savg VM_OBJECT_UNLOCK(object); 5581213937Savg va = zfs_map_page(mreq, &sf); 5582213937Savg error = dmu_read(os, zp->z_id, IDX_TO_OFF(mreq->pindex), 5583213937Savg size, va, DMU_READ_PREFETCH); 5584213937Savg if (size != PAGE_SIZE) 5585213937Savg bzero(va + size, PAGE_SIZE - size); 5586213937Savg zfs_unmap_page(sf); 5587213937Savg VM_OBJECT_LOCK(object); 5588213937Savg 5589213937Savg if (!error) 5590213937Savg mreq->valid = VM_PAGE_BITS_ALL; 5591213937Savg KASSERT(mreq->dirty == 0, ("zfs_getpages: page %p is dirty", mreq)); 5592213937Savg 5593213937Savg VM_OBJECT_UNLOCK(object); 5594213937Savg 5595213937Savg ZFS_ACCESSTIME_STAMP(zfsvfs, zp); 5596213937Savg ZFS_EXIT(zfsvfs); 5597213937Savg return (error ? VM_PAGER_ERROR : VM_PAGER_OK); 5598213937Savg} 5599213937Savg 5600213937Savgstatic int 5601213937Savgzfs_freebsd_getpages(ap) 5602213937Savg struct vop_getpages_args /* { 5603213937Savg struct vnode *a_vp; 5604213937Savg vm_page_t *a_m; 5605213937Savg int a_count; 5606213937Savg int a_reqpage; 5607213937Savg vm_ooffset_t a_offset; 5608213937Savg } */ *ap; 5609213937Savg{ 5610213937Savg 5611213937Savg return (zfs_getpages(ap->a_vp, ap->a_m, ap->a_count, ap->a_reqpage)); 5612213937Savg} 5613213937Savg 5614213937Savgstatic int 5615168962Spjdzfs_freebsd_open(ap) 5616168962Spjd struct vop_open_args /* { 5617168962Spjd struct vnode *a_vp; 5618168962Spjd int a_mode; 5619168962Spjd struct ucred *a_cred; 5620168962Spjd struct thread *a_td; 5621168962Spjd } */ *ap; 5622168962Spjd{ 5623168962Spjd vnode_t *vp = ap->a_vp; 5624168962Spjd znode_t *zp = VTOZ(vp); 5625168962Spjd int error; 5626168962Spjd 5627185029Spjd error = zfs_open(&vp, ap->a_mode, ap->a_cred, NULL); 5628168962Spjd if (error == 0) 5629219089Spjd vnode_create_vobject(vp, zp->z_size, ap->a_td); 5630168962Spjd return (error); 5631168962Spjd} 5632168962Spjd 5633168962Spjdstatic int 5634168962Spjdzfs_freebsd_close(ap) 5635168962Spjd struct vop_close_args /* { 5636168962Spjd struct vnode *a_vp; 5637168962Spjd int a_fflag; 5638168962Spjd struct ucred *a_cred; 5639168962Spjd struct thread *a_td; 5640168962Spjd } */ *ap; 5641168962Spjd{ 5642168962Spjd 5643185029Spjd return (zfs_close(ap->a_vp, ap->a_fflag, 0, 0, ap->a_cred, NULL)); 5644168962Spjd} 5645168962Spjd 5646168962Spjdstatic int 5647168962Spjdzfs_freebsd_ioctl(ap) 5648168962Spjd struct vop_ioctl_args /* { 5649168962Spjd struct vnode *a_vp; 5650168962Spjd u_long a_command; 5651168962Spjd caddr_t a_data; 5652168962Spjd int a_fflag; 5653168962Spjd struct ucred *cred; 5654168962Spjd struct thread *td; 5655168962Spjd } */ *ap; 5656168962Spjd{ 5657168962Spjd 5658168978Spjd return (zfs_ioctl(ap->a_vp, ap->a_command, (intptr_t)ap->a_data, 5659185029Spjd ap->a_fflag, ap->a_cred, NULL, NULL)); 5660168962Spjd} 5661168962Spjd 5662168962Spjdstatic int 5663168962Spjdzfs_freebsd_read(ap) 5664168962Spjd struct vop_read_args /* { 5665168962Spjd struct vnode *a_vp; 5666168962Spjd struct uio *a_uio; 5667168962Spjd int a_ioflag; 5668168962Spjd struct ucred *a_cred; 5669168962Spjd } */ *ap; 5670168962Spjd{ 5671168962Spjd 5672213673Spjd return (zfs_read(ap->a_vp, ap->a_uio, ioflags(ap->a_ioflag), 5673213673Spjd ap->a_cred, NULL)); 5674168962Spjd} 5675168962Spjd 5676168962Spjdstatic int 5677168962Spjdzfs_freebsd_write(ap) 5678168962Spjd struct vop_write_args /* { 5679168962Spjd struct vnode *a_vp; 5680168962Spjd struct uio *a_uio; 5681168962Spjd int a_ioflag; 5682168962Spjd struct ucred *a_cred; 5683168962Spjd } */ *ap; 5684168962Spjd{ 5685168962Spjd 5686207745Strasz if (vn_rlimit_fsize(ap->a_vp, ap->a_uio, ap->a_uio->uio_td)) 5687207745Strasz return (EFBIG); 5688207745Strasz 5689213673Spjd return (zfs_write(ap->a_vp, ap->a_uio, ioflags(ap->a_ioflag), 5690213673Spjd ap->a_cred, NULL)); 5691168962Spjd} 5692168962Spjd 5693168962Spjdstatic int 5694168962Spjdzfs_freebsd_access(ap) 5695168962Spjd struct vop_access_args /* { 5696168962Spjd struct vnode *a_vp; 5697192689Strasz accmode_t a_accmode; 5698168962Spjd struct ucred *a_cred; 5699168962Spjd struct thread *a_td; 5700168962Spjd } */ *ap; 5701168962Spjd{ 5702212002Sjh vnode_t *vp = ap->a_vp; 5703212002Sjh znode_t *zp = VTOZ(vp); 5704198703Spjd accmode_t accmode; 5705198703Spjd int error = 0; 5706168962Spjd 5707185172Spjd /* 5708198703Spjd * ZFS itself only knowns about VREAD, VWRITE, VEXEC and VAPPEND, 5709185172Spjd */ 5710198703Spjd accmode = ap->a_accmode & (VREAD|VWRITE|VEXEC|VAPPEND); 5711198703Spjd if (accmode != 0) 5712198703Spjd error = zfs_access(ap->a_vp, accmode, 0, ap->a_cred, NULL); 5713185172Spjd 5714198703Spjd /* 5715198703Spjd * VADMIN has to be handled by vaccess(). 5716198703Spjd */ 5717198703Spjd if (error == 0) { 5718198703Spjd accmode = ap->a_accmode & ~(VREAD|VWRITE|VEXEC|VAPPEND); 5719198703Spjd if (accmode != 0) { 5720219089Spjd error = vaccess(vp->v_type, zp->z_mode, zp->z_uid, 5721219089Spjd zp->z_gid, accmode, ap->a_cred, NULL); 5722198703Spjd } 5723185172Spjd } 5724185172Spjd 5725212002Sjh /* 5726212002Sjh * For VEXEC, ensure that at least one execute bit is set for 5727212002Sjh * non-directories. 5728212002Sjh */ 5729212002Sjh if (error == 0 && (ap->a_accmode & VEXEC) != 0 && vp->v_type != VDIR && 5730219089Spjd (zp->z_mode & (S_IXUSR | S_IXGRP | S_IXOTH)) == 0) { 5731212002Sjh error = EACCES; 5732219089Spjd } 5733212002Sjh 5734198703Spjd return (error); 5735168962Spjd} 5736168962Spjd 5737168962Spjdstatic int 5738168962Spjdzfs_freebsd_lookup(ap) 5739168962Spjd struct vop_lookup_args /* { 5740168962Spjd struct vnode *a_dvp; 5741168962Spjd struct vnode **a_vpp; 5742168962Spjd struct componentname *a_cnp; 5743168962Spjd } */ *ap; 5744168962Spjd{ 5745168962Spjd struct componentname *cnp = ap->a_cnp; 5746168962Spjd char nm[NAME_MAX + 1]; 5747168962Spjd 5748168962Spjd ASSERT(cnp->cn_namelen < sizeof(nm)); 5749168962Spjd strlcpy(nm, cnp->cn_nameptr, MIN(cnp->cn_namelen + 1, sizeof(nm))); 5750168962Spjd 5751168962Spjd return (zfs_lookup(ap->a_dvp, nm, ap->a_vpp, cnp, cnp->cn_nameiop, 5752185029Spjd cnp->cn_cred, cnp->cn_thread, 0)); 5753168962Spjd} 5754168962Spjd 5755168962Spjdstatic int 5756168962Spjdzfs_freebsd_create(ap) 5757168962Spjd struct vop_create_args /* { 5758168962Spjd struct vnode *a_dvp; 5759168962Spjd struct vnode **a_vpp; 5760168962Spjd struct componentname *a_cnp; 5761168962Spjd struct vattr *a_vap; 5762168962Spjd } */ *ap; 5763168962Spjd{ 5764168962Spjd struct componentname *cnp = ap->a_cnp; 5765168962Spjd vattr_t *vap = ap->a_vap; 5766168962Spjd int mode; 5767168962Spjd 5768168962Spjd ASSERT(cnp->cn_flags & SAVENAME); 5769168962Spjd 5770168962Spjd vattr_init_mask(vap); 5771168962Spjd mode = vap->va_mode & ALLPERMS; 5772168962Spjd 5773168962Spjd return (zfs_create(ap->a_dvp, cnp->cn_nameptr, vap, !EXCL, mode, 5774185029Spjd ap->a_vpp, cnp->cn_cred, cnp->cn_thread)); 5775168962Spjd} 5776168962Spjd 5777168962Spjdstatic int 5778168962Spjdzfs_freebsd_remove(ap) 5779168962Spjd struct vop_remove_args /* { 5780168962Spjd struct vnode *a_dvp; 5781168962Spjd struct vnode *a_vp; 5782168962Spjd struct componentname *a_cnp; 5783168962Spjd } */ *ap; 5784168962Spjd{ 5785168962Spjd 5786168962Spjd ASSERT(ap->a_cnp->cn_flags & SAVENAME); 5787168962Spjd 5788168962Spjd return (zfs_remove(ap->a_dvp, ap->a_cnp->cn_nameptr, 5789185029Spjd ap->a_cnp->cn_cred, NULL, 0)); 5790168962Spjd} 5791168962Spjd 5792168962Spjdstatic int 5793168962Spjdzfs_freebsd_mkdir(ap) 5794168962Spjd struct vop_mkdir_args /* { 5795168962Spjd struct vnode *a_dvp; 5796168962Spjd struct vnode **a_vpp; 5797168962Spjd struct componentname *a_cnp; 5798168962Spjd struct vattr *a_vap; 5799168962Spjd } */ *ap; 5800168962Spjd{ 5801168962Spjd vattr_t *vap = ap->a_vap; 5802168962Spjd 5803168962Spjd ASSERT(ap->a_cnp->cn_flags & SAVENAME); 5804168962Spjd 5805168962Spjd vattr_init_mask(vap); 5806168962Spjd 5807168962Spjd return (zfs_mkdir(ap->a_dvp, ap->a_cnp->cn_nameptr, vap, ap->a_vpp, 5808185029Spjd ap->a_cnp->cn_cred, NULL, 0, NULL)); 5809168962Spjd} 5810168962Spjd 5811168962Spjdstatic int 5812168962Spjdzfs_freebsd_rmdir(ap) 5813168962Spjd struct vop_rmdir_args /* { 5814168962Spjd struct vnode *a_dvp; 5815168962Spjd struct vnode *a_vp; 5816168962Spjd struct componentname *a_cnp; 5817168962Spjd } */ *ap; 5818168962Spjd{ 5819168962Spjd struct componentname *cnp = ap->a_cnp; 5820168962Spjd 5821168962Spjd ASSERT(cnp->cn_flags & SAVENAME); 5822168962Spjd 5823185029Spjd return (zfs_rmdir(ap->a_dvp, cnp->cn_nameptr, NULL, cnp->cn_cred, NULL, 0)); 5824168962Spjd} 5825168962Spjd 5826168962Spjdstatic int 5827168962Spjdzfs_freebsd_readdir(ap) 5828168962Spjd struct vop_readdir_args /* { 5829168962Spjd struct vnode *a_vp; 5830168962Spjd struct uio *a_uio; 5831168962Spjd struct ucred *a_cred; 5832168962Spjd int *a_eofflag; 5833168962Spjd int *a_ncookies; 5834168962Spjd u_long **a_cookies; 5835168962Spjd } */ *ap; 5836168962Spjd{ 5837168962Spjd 5838168962Spjd return (zfs_readdir(ap->a_vp, ap->a_uio, ap->a_cred, ap->a_eofflag, 5839168962Spjd ap->a_ncookies, ap->a_cookies)); 5840168962Spjd} 5841168962Spjd 5842168962Spjdstatic int 5843168962Spjdzfs_freebsd_fsync(ap) 5844168962Spjd struct vop_fsync_args /* { 5845168962Spjd struct vnode *a_vp; 5846168962Spjd int a_waitfor; 5847168962Spjd struct thread *a_td; 5848168962Spjd } */ *ap; 5849168962Spjd{ 5850168962Spjd 5851168962Spjd vop_stdfsync(ap); 5852185029Spjd return (zfs_fsync(ap->a_vp, 0, ap->a_td->td_ucred, NULL)); 5853168962Spjd} 5854168962Spjd 5855168962Spjdstatic int 5856168962Spjdzfs_freebsd_getattr(ap) 5857168962Spjd struct vop_getattr_args /* { 5858168962Spjd struct vnode *a_vp; 5859168962Spjd struct vattr *a_vap; 5860168962Spjd struct ucred *a_cred; 5861168962Spjd } */ *ap; 5862168962Spjd{ 5863185029Spjd vattr_t *vap = ap->a_vap; 5864185029Spjd xvattr_t xvap; 5865185029Spjd u_long fflags = 0; 5866185029Spjd int error; 5867168962Spjd 5868185029Spjd xva_init(&xvap); 5869185029Spjd xvap.xva_vattr = *vap; 5870185029Spjd xvap.xva_vattr.va_mask |= AT_XVATTR; 5871185029Spjd 5872185029Spjd /* Convert chflags into ZFS-type flags. */ 5873185029Spjd /* XXX: what about SF_SETTABLE?. */ 5874185029Spjd XVA_SET_REQ(&xvap, XAT_IMMUTABLE); 5875185029Spjd XVA_SET_REQ(&xvap, XAT_APPENDONLY); 5876185029Spjd XVA_SET_REQ(&xvap, XAT_NOUNLINK); 5877185029Spjd XVA_SET_REQ(&xvap, XAT_NODUMP); 5878185029Spjd error = zfs_getattr(ap->a_vp, (vattr_t *)&xvap, 0, ap->a_cred, NULL); 5879185029Spjd if (error != 0) 5880185029Spjd return (error); 5881185029Spjd 5882185029Spjd /* Convert ZFS xattr into chflags. */ 5883185029Spjd#define FLAG_CHECK(fflag, xflag, xfield) do { \ 5884185029Spjd if (XVA_ISSET_RTN(&xvap, (xflag)) && (xfield) != 0) \ 5885185029Spjd fflags |= (fflag); \ 5886185029Spjd} while (0) 5887185029Spjd FLAG_CHECK(SF_IMMUTABLE, XAT_IMMUTABLE, 5888185029Spjd xvap.xva_xoptattrs.xoa_immutable); 5889185029Spjd FLAG_CHECK(SF_APPEND, XAT_APPENDONLY, 5890185029Spjd xvap.xva_xoptattrs.xoa_appendonly); 5891185029Spjd FLAG_CHECK(SF_NOUNLINK, XAT_NOUNLINK, 5892185029Spjd xvap.xva_xoptattrs.xoa_nounlink); 5893185029Spjd FLAG_CHECK(UF_NODUMP, XAT_NODUMP, 5894185029Spjd xvap.xva_xoptattrs.xoa_nodump); 5895185029Spjd#undef FLAG_CHECK 5896185029Spjd *vap = xvap.xva_vattr; 5897185029Spjd vap->va_flags = fflags; 5898185029Spjd return (0); 5899168962Spjd} 5900168962Spjd 5901168962Spjdstatic int 5902168962Spjdzfs_freebsd_setattr(ap) 5903168962Spjd struct vop_setattr_args /* { 5904168962Spjd struct vnode *a_vp; 5905168962Spjd struct vattr *a_vap; 5906168962Spjd struct ucred *a_cred; 5907168962Spjd } */ *ap; 5908168962Spjd{ 5909185172Spjd vnode_t *vp = ap->a_vp; 5910168962Spjd vattr_t *vap = ap->a_vap; 5911185172Spjd cred_t *cred = ap->a_cred; 5912185029Spjd xvattr_t xvap; 5913185029Spjd u_long fflags; 5914185029Spjd uint64_t zflags; 5915168962Spjd 5916168962Spjd vattr_init_mask(vap); 5917170044Spjd vap->va_mask &= ~AT_NOSET; 5918168962Spjd 5919185029Spjd xva_init(&xvap); 5920185029Spjd xvap.xva_vattr = *vap; 5921185029Spjd 5922219089Spjd zflags = VTOZ(vp)->z_pflags; 5923185172Spjd 5924185029Spjd if (vap->va_flags != VNOVAL) { 5925197683Sdelphij zfsvfs_t *zfsvfs = VTOZ(vp)->z_zfsvfs; 5926185172Spjd int error; 5927185172Spjd 5928197683Sdelphij if (zfsvfs->z_use_fuids == B_FALSE) 5929197683Sdelphij return (EOPNOTSUPP); 5930197683Sdelphij 5931185029Spjd fflags = vap->va_flags; 5932185029Spjd if ((fflags & ~(SF_IMMUTABLE|SF_APPEND|SF_NOUNLINK|UF_NODUMP)) != 0) 5933185029Spjd return (EOPNOTSUPP); 5934185172Spjd /* 5935185172Spjd * Unprivileged processes are not permitted to unset system 5936185172Spjd * flags, or modify flags if any system flags are set. 5937185172Spjd * Privileged non-jail processes may not modify system flags 5938185172Spjd * if securelevel > 0 and any existing system flags are set. 5939185172Spjd * Privileged jail processes behave like privileged non-jail 5940185172Spjd * processes if the security.jail.chflags_allowed sysctl is 5941185172Spjd * is non-zero; otherwise, they behave like unprivileged 5942185172Spjd * processes. 5943185172Spjd */ 5944197861Spjd if (secpolicy_fs_owner(vp->v_mount, cred) == 0 || 5945197861Spjd priv_check_cred(cred, PRIV_VFS_SYSFLAGS, 0) == 0) { 5946185172Spjd if (zflags & 5947185172Spjd (ZFS_IMMUTABLE | ZFS_APPENDONLY | ZFS_NOUNLINK)) { 5948185172Spjd error = securelevel_gt(cred, 0); 5949197861Spjd if (error != 0) 5950185172Spjd return (error); 5951185172Spjd } 5952185172Spjd } else { 5953197861Spjd /* 5954197861Spjd * Callers may only modify the file flags on objects they 5955197861Spjd * have VADMIN rights for. 5956197861Spjd */ 5957197861Spjd if ((error = VOP_ACCESS(vp, VADMIN, cred, curthread)) != 0) 5958197861Spjd return (error); 5959185172Spjd if (zflags & 5960185172Spjd (ZFS_IMMUTABLE | ZFS_APPENDONLY | ZFS_NOUNLINK)) { 5961185172Spjd return (EPERM); 5962185172Spjd } 5963185172Spjd if (fflags & 5964185172Spjd (SF_IMMUTABLE | SF_APPEND | SF_NOUNLINK)) { 5965185172Spjd return (EPERM); 5966185172Spjd } 5967185172Spjd } 5968185029Spjd 5969185029Spjd#define FLAG_CHANGE(fflag, zflag, xflag, xfield) do { \ 5970185029Spjd if (((fflags & (fflag)) && !(zflags & (zflag))) || \ 5971185029Spjd ((zflags & (zflag)) && !(fflags & (fflag)))) { \ 5972185029Spjd XVA_SET_REQ(&xvap, (xflag)); \ 5973185029Spjd (xfield) = ((fflags & (fflag)) != 0); \ 5974185029Spjd } \ 5975185029Spjd} while (0) 5976185029Spjd /* Convert chflags into ZFS-type flags. */ 5977185029Spjd /* XXX: what about SF_SETTABLE?. */ 5978185029Spjd FLAG_CHANGE(SF_IMMUTABLE, ZFS_IMMUTABLE, XAT_IMMUTABLE, 5979185029Spjd xvap.xva_xoptattrs.xoa_immutable); 5980185029Spjd FLAG_CHANGE(SF_APPEND, ZFS_APPENDONLY, XAT_APPENDONLY, 5981185029Spjd xvap.xva_xoptattrs.xoa_appendonly); 5982185029Spjd FLAG_CHANGE(SF_NOUNLINK, ZFS_NOUNLINK, XAT_NOUNLINK, 5983185029Spjd xvap.xva_xoptattrs.xoa_nounlink); 5984185029Spjd FLAG_CHANGE(UF_NODUMP, ZFS_NODUMP, XAT_NODUMP, 5985185172Spjd xvap.xva_xoptattrs.xoa_nodump); 5986185029Spjd#undef FLAG_CHANGE 5987185029Spjd } 5988185172Spjd return (zfs_setattr(vp, (vattr_t *)&xvap, 0, cred, NULL)); 5989168962Spjd} 5990168962Spjd 5991168962Spjdstatic int 5992168962Spjdzfs_freebsd_rename(ap) 5993168962Spjd struct vop_rename_args /* { 5994168962Spjd struct vnode *a_fdvp; 5995168962Spjd struct vnode *a_fvp; 5996168962Spjd struct componentname *a_fcnp; 5997168962Spjd struct vnode *a_tdvp; 5998168962Spjd struct vnode *a_tvp; 5999168962Spjd struct componentname *a_tcnp; 6000168962Spjd } */ *ap; 6001168962Spjd{ 6002168962Spjd vnode_t *fdvp = ap->a_fdvp; 6003168962Spjd vnode_t *fvp = ap->a_fvp; 6004168962Spjd vnode_t *tdvp = ap->a_tdvp; 6005168962Spjd vnode_t *tvp = ap->a_tvp; 6006168962Spjd int error; 6007168962Spjd 6008192237Skmacy ASSERT(ap->a_fcnp->cn_flags & (SAVENAME|SAVESTART)); 6009192237Skmacy ASSERT(ap->a_tcnp->cn_flags & (SAVENAME|SAVESTART)); 6010168962Spjd 6011168962Spjd error = zfs_rename(fdvp, ap->a_fcnp->cn_nameptr, tdvp, 6012185029Spjd ap->a_tcnp->cn_nameptr, ap->a_fcnp->cn_cred, NULL, 0); 6013168962Spjd 6014168962Spjd if (tdvp == tvp) 6015168962Spjd VN_RELE(tdvp); 6016168962Spjd else 6017168962Spjd VN_URELE(tdvp); 6018168962Spjd if (tvp) 6019168962Spjd VN_URELE(tvp); 6020168962Spjd VN_RELE(fdvp); 6021168962Spjd VN_RELE(fvp); 6022168962Spjd 6023168962Spjd return (error); 6024168962Spjd} 6025168962Spjd 6026168962Spjdstatic int 6027168962Spjdzfs_freebsd_symlink(ap) 6028168962Spjd struct vop_symlink_args /* { 6029168962Spjd struct vnode *a_dvp; 6030168962Spjd struct vnode **a_vpp; 6031168962Spjd struct componentname *a_cnp; 6032168962Spjd struct vattr *a_vap; 6033168962Spjd char *a_target; 6034168962Spjd } */ *ap; 6035168962Spjd{ 6036168962Spjd struct componentname *cnp = ap->a_cnp; 6037168962Spjd vattr_t *vap = ap->a_vap; 6038168962Spjd 6039168962Spjd ASSERT(cnp->cn_flags & SAVENAME); 6040168962Spjd 6041168962Spjd vap->va_type = VLNK; /* FreeBSD: Syscall only sets va_mode. */ 6042168962Spjd vattr_init_mask(vap); 6043168962Spjd 6044168962Spjd return (zfs_symlink(ap->a_dvp, ap->a_vpp, cnp->cn_nameptr, vap, 6045168962Spjd ap->a_target, cnp->cn_cred, cnp->cn_thread)); 6046168962Spjd} 6047168962Spjd 6048168962Spjdstatic int 6049168962Spjdzfs_freebsd_readlink(ap) 6050168962Spjd struct vop_readlink_args /* { 6051168962Spjd struct vnode *a_vp; 6052168962Spjd struct uio *a_uio; 6053168962Spjd struct ucred *a_cred; 6054168962Spjd } */ *ap; 6055168962Spjd{ 6056168962Spjd 6057185029Spjd return (zfs_readlink(ap->a_vp, ap->a_uio, ap->a_cred, NULL)); 6058168962Spjd} 6059168962Spjd 6060168962Spjdstatic int 6061168962Spjdzfs_freebsd_link(ap) 6062168962Spjd struct vop_link_args /* { 6063168962Spjd struct vnode *a_tdvp; 6064168962Spjd struct vnode *a_vp; 6065168962Spjd struct componentname *a_cnp; 6066168962Spjd } */ *ap; 6067168962Spjd{ 6068168962Spjd struct componentname *cnp = ap->a_cnp; 6069168962Spjd 6070168962Spjd ASSERT(cnp->cn_flags & SAVENAME); 6071168962Spjd 6072185029Spjd return (zfs_link(ap->a_tdvp, ap->a_vp, cnp->cn_nameptr, cnp->cn_cred, NULL, 0)); 6073168962Spjd} 6074168962Spjd 6075168962Spjdstatic int 6076168962Spjdzfs_freebsd_inactive(ap) 6077169170Spjd struct vop_inactive_args /* { 6078169170Spjd struct vnode *a_vp; 6079169170Spjd struct thread *a_td; 6080169170Spjd } */ *ap; 6081168962Spjd{ 6082168962Spjd vnode_t *vp = ap->a_vp; 6083168962Spjd 6084185029Spjd zfs_inactive(vp, ap->a_td->td_ucred, NULL); 6085168962Spjd return (0); 6086168962Spjd} 6087168962Spjd 6088185029Spjdstatic void 6089185029Spjdzfs_reclaim_complete(void *arg, int pending) 6090185029Spjd{ 6091185029Spjd znode_t *zp = arg; 6092185029Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 6093185029Spjd 6094197133Spjd rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER); 6095219089Spjd if (zp->z_sa_hdl != NULL) { 6096197133Spjd ZFS_OBJ_HOLD_ENTER(zfsvfs, zp->z_id); 6097197133Spjd zfs_znode_dmu_fini(zp); 6098197133Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, zp->z_id); 6099197133Spjd } 6100185029Spjd zfs_znode_free(zp); 6101197133Spjd rw_exit(&zfsvfs->z_teardown_inactive_lock); 6102197133Spjd /* 6103197133Spjd * If the file system is being unmounted, there is a process waiting 6104197133Spjd * for us, wake it up. 6105197133Spjd */ 6106197133Spjd if (zfsvfs->z_unmounted) 6107197133Spjd wakeup_one(zfsvfs); 6108185029Spjd} 6109185029Spjd 6110168962Spjdstatic int 6111168962Spjdzfs_freebsd_reclaim(ap) 6112168962Spjd struct vop_reclaim_args /* { 6113168962Spjd struct vnode *a_vp; 6114168962Spjd struct thread *a_td; 6115168962Spjd } */ *ap; 6116168962Spjd{ 6117169170Spjd vnode_t *vp = ap->a_vp; 6118168962Spjd znode_t *zp = VTOZ(vp); 6119197133Spjd zfsvfs_t *zfsvfs = zp->z_zfsvfs; 6120219089Spjd boolean_t rlocked; 6121168962Spjd 6122219089Spjd rlocked = rw_tryenter(&zfsvfs->z_teardown_inactive_lock, RW_READER); 6123197133Spjd 6124169025Spjd ASSERT(zp != NULL); 6125169025Spjd 6126168962Spjd /* 6127168962Spjd * Destroy the vm object and flush associated pages. 6128168962Spjd */ 6129168962Spjd vnode_destroy_vobject(vp); 6130169025Spjd 6131169025Spjd mutex_enter(&zp->z_lock); 6132197153Spjd zp->z_vnode = NULL; 6133196301Spjd mutex_exit(&zp->z_lock); 6134196301Spjd 6135219089Spjd if (zp->z_unlinked) { 6136196301Spjd ; /* Do nothing. */ 6137219089Spjd } else if (!rlocked) { 6138219089Spjd TASK_INIT(&zp->z_task, 0, zfs_reclaim_complete, zp); 6139219089Spjd taskqueue_enqueue(taskqueue_thread, &zp->z_task); 6140219089Spjd } else if (zp->z_sa_hdl == NULL) { 6141196301Spjd zfs_znode_free(zp); 6142219089Spjd } else /* if (!zp->z_unlinked && zp->z_dbuf != NULL) */ { 6143185029Spjd int locked; 6144185029Spjd 6145185029Spjd locked = MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id)) ? 2 : 6146185029Spjd ZFS_OBJ_HOLD_TRYENTER(zfsvfs, zp->z_id); 6147185029Spjd if (locked == 0) { 6148185029Spjd /* 6149185029Spjd * Lock can't be obtained due to deadlock possibility, 6150185029Spjd * so defer znode destruction. 6151185029Spjd */ 6152185029Spjd TASK_INIT(&zp->z_task, 0, zfs_reclaim_complete, zp); 6153185029Spjd taskqueue_enqueue(taskqueue_thread, &zp->z_task); 6154185029Spjd } else { 6155185029Spjd zfs_znode_dmu_fini(zp); 6156185029Spjd if (locked == 1) 6157185029Spjd ZFS_OBJ_HOLD_EXIT(zfsvfs, zp->z_id); 6158185029Spjd zfs_znode_free(zp); 6159185029Spjd } 6160169025Spjd } 6161168962Spjd VI_LOCK(vp); 6162168962Spjd vp->v_data = NULL; 6163171567Spjd ASSERT(vp->v_holdcnt >= 1); 6164171316Sdfr VI_UNLOCK(vp); 6165219089Spjd if (rlocked) 6166219089Spjd rw_exit(&zfsvfs->z_teardown_inactive_lock); 6167168962Spjd return (0); 6168168962Spjd} 6169168962Spjd 6170168962Spjdstatic int 6171168962Spjdzfs_freebsd_fid(ap) 6172168962Spjd struct vop_fid_args /* { 6173168962Spjd struct vnode *a_vp; 6174168962Spjd struct fid *a_fid; 6175168962Spjd } */ *ap; 6176168962Spjd{ 6177168962Spjd 6178185029Spjd return (zfs_fid(ap->a_vp, (void *)ap->a_fid, NULL)); 6179168962Spjd} 6180168962Spjd 6181168962Spjdstatic int 6182168962Spjdzfs_freebsd_pathconf(ap) 6183168962Spjd struct vop_pathconf_args /* { 6184168962Spjd struct vnode *a_vp; 6185168962Spjd int a_name; 6186168962Spjd register_t *a_retval; 6187168962Spjd } */ *ap; 6188168962Spjd{ 6189168962Spjd ulong_t val; 6190168962Spjd int error; 6191168962Spjd 6192185029Spjd error = zfs_pathconf(ap->a_vp, ap->a_name, &val, curthread->td_ucred, NULL); 6193168962Spjd if (error == 0) 6194168962Spjd *ap->a_retval = val; 6195168962Spjd else if (error == EOPNOTSUPP) 6196168962Spjd error = vop_stdpathconf(ap); 6197168962Spjd return (error); 6198168962Spjd} 6199168962Spjd 6200196949Straszstatic int 6201196949Straszzfs_freebsd_fifo_pathconf(ap) 6202196949Strasz struct vop_pathconf_args /* { 6203196949Strasz struct vnode *a_vp; 6204196949Strasz int a_name; 6205196949Strasz register_t *a_retval; 6206196949Strasz } */ *ap; 6207196949Strasz{ 6208196949Strasz 6209196949Strasz switch (ap->a_name) { 6210196949Strasz case _PC_ACL_EXTENDED: 6211196949Strasz case _PC_ACL_NFS4: 6212196949Strasz case _PC_ACL_PATH_MAX: 6213196949Strasz case _PC_MAC_PRESENT: 6214196949Strasz return (zfs_freebsd_pathconf(ap)); 6215196949Strasz default: 6216196949Strasz return (fifo_specops.vop_pathconf(ap)); 6217196949Strasz } 6218196949Strasz} 6219196949Strasz 6220185029Spjd/* 6221185029Spjd * FreeBSD's extended attributes namespace defines file name prefix for ZFS' 6222185029Spjd * extended attribute name: 6223185029Spjd * 6224185029Spjd * NAMESPACE PREFIX 6225185029Spjd * system freebsd:system: 6226185029Spjd * user (none, can be used to access ZFS fsattr(5) attributes 6227185029Spjd * created on Solaris) 6228185029Spjd */ 6229185029Spjdstatic int 6230185029Spjdzfs_create_attrname(int attrnamespace, const char *name, char *attrname, 6231185029Spjd size_t size) 6232185029Spjd{ 6233185029Spjd const char *namespace, *prefix, *suffix; 6234185029Spjd 6235185029Spjd /* We don't allow '/' character in attribute name. */ 6236185029Spjd if (strchr(name, '/') != NULL) 6237185029Spjd return (EINVAL); 6238185029Spjd /* We don't allow attribute names that start with "freebsd:" string. */ 6239185029Spjd if (strncmp(name, "freebsd:", 8) == 0) 6240185029Spjd return (EINVAL); 6241185029Spjd 6242185029Spjd bzero(attrname, size); 6243185029Spjd 6244185029Spjd switch (attrnamespace) { 6245185029Spjd case EXTATTR_NAMESPACE_USER: 6246185029Spjd#if 0 6247185029Spjd prefix = "freebsd:"; 6248185029Spjd namespace = EXTATTR_NAMESPACE_USER_STRING; 6249185029Spjd suffix = ":"; 6250185029Spjd#else 6251185029Spjd /* 6252185029Spjd * This is the default namespace by which we can access all 6253185029Spjd * attributes created on Solaris. 6254185029Spjd */ 6255185029Spjd prefix = namespace = suffix = ""; 6256185029Spjd#endif 6257185029Spjd break; 6258185029Spjd case EXTATTR_NAMESPACE_SYSTEM: 6259185029Spjd prefix = "freebsd:"; 6260185029Spjd namespace = EXTATTR_NAMESPACE_SYSTEM_STRING; 6261185029Spjd suffix = ":"; 6262185029Spjd break; 6263185029Spjd case EXTATTR_NAMESPACE_EMPTY: 6264185029Spjd default: 6265185029Spjd return (EINVAL); 6266185029Spjd } 6267185029Spjd if (snprintf(attrname, size, "%s%s%s%s", prefix, namespace, suffix, 6268185029Spjd name) >= size) { 6269185029Spjd return (ENAMETOOLONG); 6270185029Spjd } 6271185029Spjd return (0); 6272185029Spjd} 6273185029Spjd 6274185029Spjd/* 6275185029Spjd * Vnode operating to retrieve a named extended attribute. 6276185029Spjd */ 6277185029Spjdstatic int 6278185029Spjdzfs_getextattr(struct vop_getextattr_args *ap) 6279185029Spjd/* 6280185029Spjdvop_getextattr { 6281185029Spjd IN struct vnode *a_vp; 6282185029Spjd IN int a_attrnamespace; 6283185029Spjd IN const char *a_name; 6284185029Spjd INOUT struct uio *a_uio; 6285185029Spjd OUT size_t *a_size; 6286185029Spjd IN struct ucred *a_cred; 6287185029Spjd IN struct thread *a_td; 6288185029Spjd}; 6289185029Spjd*/ 6290185029Spjd{ 6291185029Spjd zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs; 6292185029Spjd struct thread *td = ap->a_td; 6293185029Spjd struct nameidata nd; 6294185029Spjd char attrname[255]; 6295185029Spjd struct vattr va; 6296185029Spjd vnode_t *xvp = NULL, *vp; 6297185029Spjd int error, flags; 6298185029Spjd 6299195785Strasz error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace, 6300195785Strasz ap->a_cred, ap->a_td, VREAD); 6301195785Strasz if (error != 0) 6302195785Strasz return (error); 6303195785Strasz 6304185029Spjd error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname, 6305185029Spjd sizeof(attrname)); 6306185029Spjd if (error != 0) 6307185029Spjd return (error); 6308185029Spjd 6309185029Spjd ZFS_ENTER(zfsvfs); 6310185029Spjd 6311185029Spjd error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td, 6312185029Spjd LOOKUP_XATTR); 6313185029Spjd if (error != 0) { 6314185029Spjd ZFS_EXIT(zfsvfs); 6315185029Spjd return (error); 6316185029Spjd } 6317185029Spjd 6318185029Spjd flags = FREAD; 6319185029Spjd NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW | MPSAFE, UIO_SYSSPACE, attrname, 6320185029Spjd xvp, td); 6321194586Skib error = vn_open_cred(&nd, &flags, 0, 0, ap->a_cred, NULL); 6322185029Spjd vp = nd.ni_vp; 6323185029Spjd NDFREE(&nd, NDF_ONLY_PNBUF); 6324185029Spjd if (error != 0) { 6325196303Spjd ZFS_EXIT(zfsvfs); 6326195785Strasz if (error == ENOENT) 6327195785Strasz error = ENOATTR; 6328185029Spjd return (error); 6329185029Spjd } 6330185029Spjd 6331185029Spjd if (ap->a_size != NULL) { 6332185029Spjd error = VOP_GETATTR(vp, &va, ap->a_cred); 6333185029Spjd if (error == 0) 6334185029Spjd *ap->a_size = (size_t)va.va_size; 6335185029Spjd } else if (ap->a_uio != NULL) 6336185029Spjd error = VOP_READ(vp, ap->a_uio, IO_UNIT | IO_SYNC, ap->a_cred); 6337185029Spjd 6338185029Spjd VOP_UNLOCK(vp, 0); 6339185029Spjd vn_close(vp, flags, ap->a_cred, td); 6340185029Spjd ZFS_EXIT(zfsvfs); 6341185029Spjd 6342185029Spjd return (error); 6343185029Spjd} 6344185029Spjd 6345185029Spjd/* 6346185029Spjd * Vnode operation to remove a named attribute. 6347185029Spjd */ 6348185029Spjdint 6349185029Spjdzfs_deleteextattr(struct vop_deleteextattr_args *ap) 6350185029Spjd/* 6351185029Spjdvop_deleteextattr { 6352185029Spjd IN struct vnode *a_vp; 6353185029Spjd IN int a_attrnamespace; 6354185029Spjd IN const char *a_name; 6355185029Spjd IN struct ucred *a_cred; 6356185029Spjd IN struct thread *a_td; 6357185029Spjd}; 6358185029Spjd*/ 6359185029Spjd{ 6360185029Spjd zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs; 6361185029Spjd struct thread *td = ap->a_td; 6362185029Spjd struct nameidata nd; 6363185029Spjd char attrname[255]; 6364185029Spjd struct vattr va; 6365185029Spjd vnode_t *xvp = NULL, *vp; 6366185029Spjd int error, flags; 6367185029Spjd 6368195785Strasz error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace, 6369195785Strasz ap->a_cred, ap->a_td, VWRITE); 6370195785Strasz if (error != 0) 6371195785Strasz return (error); 6372195785Strasz 6373185029Spjd error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname, 6374185029Spjd sizeof(attrname)); 6375185029Spjd if (error != 0) 6376185029Spjd return (error); 6377185029Spjd 6378185029Spjd ZFS_ENTER(zfsvfs); 6379185029Spjd 6380185029Spjd error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td, 6381185029Spjd LOOKUP_XATTR); 6382185029Spjd if (error != 0) { 6383185029Spjd ZFS_EXIT(zfsvfs); 6384185029Spjd return (error); 6385185029Spjd } 6386185029Spjd 6387185029Spjd NDINIT_ATVP(&nd, DELETE, NOFOLLOW | LOCKPARENT | LOCKLEAF | MPSAFE, 6388185029Spjd UIO_SYSSPACE, attrname, xvp, td); 6389185029Spjd error = namei(&nd); 6390185029Spjd vp = nd.ni_vp; 6391185029Spjd NDFREE(&nd, NDF_ONLY_PNBUF); 6392185029Spjd if (error != 0) { 6393196303Spjd ZFS_EXIT(zfsvfs); 6394195785Strasz if (error == ENOENT) 6395195785Strasz error = ENOATTR; 6396185029Spjd return (error); 6397185029Spjd } 6398185029Spjd error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd); 6399185029Spjd 6400185029Spjd vput(nd.ni_dvp); 6401185029Spjd if (vp == nd.ni_dvp) 6402185029Spjd vrele(vp); 6403185029Spjd else 6404185029Spjd vput(vp); 6405185029Spjd ZFS_EXIT(zfsvfs); 6406185029Spjd 6407185029Spjd return (error); 6408185029Spjd} 6409185029Spjd 6410185029Spjd/* 6411185029Spjd * Vnode operation to set a named attribute. 6412185029Spjd */ 6413185029Spjdstatic int 6414185029Spjdzfs_setextattr(struct vop_setextattr_args *ap) 6415185029Spjd/* 6416185029Spjdvop_setextattr { 6417185029Spjd IN struct vnode *a_vp; 6418185029Spjd IN int a_attrnamespace; 6419185029Spjd IN const char *a_name; 6420185029Spjd INOUT struct uio *a_uio; 6421185029Spjd IN struct ucred *a_cred; 6422185029Spjd IN struct thread *a_td; 6423185029Spjd}; 6424185029Spjd*/ 6425185029Spjd{ 6426185029Spjd zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs; 6427185029Spjd struct thread *td = ap->a_td; 6428185029Spjd struct nameidata nd; 6429185029Spjd char attrname[255]; 6430185029Spjd struct vattr va; 6431185029Spjd vnode_t *xvp = NULL, *vp; 6432185029Spjd int error, flags; 6433185029Spjd 6434195785Strasz error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace, 6435195785Strasz ap->a_cred, ap->a_td, VWRITE); 6436195785Strasz if (error != 0) 6437195785Strasz return (error); 6438195785Strasz 6439185029Spjd error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname, 6440185029Spjd sizeof(attrname)); 6441185029Spjd if (error != 0) 6442185029Spjd return (error); 6443185029Spjd 6444185029Spjd ZFS_ENTER(zfsvfs); 6445185029Spjd 6446185029Spjd error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td, 6447195785Strasz LOOKUP_XATTR | CREATE_XATTR_DIR); 6448185029Spjd if (error != 0) { 6449185029Spjd ZFS_EXIT(zfsvfs); 6450185029Spjd return (error); 6451185029Spjd } 6452185029Spjd 6453185029Spjd flags = FFLAGS(O_WRONLY | O_CREAT); 6454185029Spjd NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW | MPSAFE, UIO_SYSSPACE, attrname, 6455185029Spjd xvp, td); 6456194586Skib error = vn_open_cred(&nd, &flags, 0600, 0, ap->a_cred, NULL); 6457185029Spjd vp = nd.ni_vp; 6458185029Spjd NDFREE(&nd, NDF_ONLY_PNBUF); 6459185029Spjd if (error != 0) { 6460185029Spjd ZFS_EXIT(zfsvfs); 6461185029Spjd return (error); 6462185029Spjd } 6463185029Spjd 6464185029Spjd VATTR_NULL(&va); 6465185029Spjd va.va_size = 0; 6466185029Spjd error = VOP_SETATTR(vp, &va, ap->a_cred); 6467185029Spjd if (error == 0) 6468185029Spjd VOP_WRITE(vp, ap->a_uio, IO_UNIT | IO_SYNC, ap->a_cred); 6469185029Spjd 6470185029Spjd VOP_UNLOCK(vp, 0); 6471185029Spjd vn_close(vp, flags, ap->a_cred, td); 6472185029Spjd ZFS_EXIT(zfsvfs); 6473185029Spjd 6474185029Spjd return (error); 6475185029Spjd} 6476185029Spjd 6477185029Spjd/* 6478185029Spjd * Vnode operation to retrieve extended attributes on a vnode. 6479185029Spjd */ 6480185029Spjdstatic int 6481185029Spjdzfs_listextattr(struct vop_listextattr_args *ap) 6482185029Spjd/* 6483185029Spjdvop_listextattr { 6484185029Spjd IN struct vnode *a_vp; 6485185029Spjd IN int a_attrnamespace; 6486185029Spjd INOUT struct uio *a_uio; 6487185029Spjd OUT size_t *a_size; 6488185029Spjd IN struct ucred *a_cred; 6489185029Spjd IN struct thread *a_td; 6490185029Spjd}; 6491185029Spjd*/ 6492185029Spjd{ 6493185029Spjd zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs; 6494185029Spjd struct thread *td = ap->a_td; 6495185029Spjd struct nameidata nd; 6496185029Spjd char attrprefix[16]; 6497185029Spjd u_char dirbuf[sizeof(struct dirent)]; 6498185029Spjd struct dirent *dp; 6499185029Spjd struct iovec aiov; 6500185029Spjd struct uio auio, *uio = ap->a_uio; 6501185029Spjd size_t *sizep = ap->a_size; 6502185029Spjd size_t plen; 6503185029Spjd vnode_t *xvp = NULL, *vp; 6504185029Spjd int done, error, eof, pos; 6505185029Spjd 6506195785Strasz error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace, 6507195785Strasz ap->a_cred, ap->a_td, VREAD); 6508196303Spjd if (error != 0) 6509195785Strasz return (error); 6510195785Strasz 6511185029Spjd error = zfs_create_attrname(ap->a_attrnamespace, "", attrprefix, 6512185029Spjd sizeof(attrprefix)); 6513185029Spjd if (error != 0) 6514185029Spjd return (error); 6515185029Spjd plen = strlen(attrprefix); 6516185029Spjd 6517185029Spjd ZFS_ENTER(zfsvfs); 6518185029Spjd 6519195822Strasz if (sizep != NULL) 6520195822Strasz *sizep = 0; 6521195822Strasz 6522185029Spjd error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td, 6523185029Spjd LOOKUP_XATTR); 6524185029Spjd if (error != 0) { 6525196303Spjd ZFS_EXIT(zfsvfs); 6526195785Strasz /* 6527195785Strasz * ENOATTR means that the EA directory does not yet exist, 6528195785Strasz * i.e. there are no extended attributes there. 6529195785Strasz */ 6530195785Strasz if (error == ENOATTR) 6531195785Strasz error = 0; 6532185029Spjd return (error); 6533185029Spjd } 6534185029Spjd 6535188588Sjhb NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | LOCKSHARED | MPSAFE, 6536188588Sjhb UIO_SYSSPACE, ".", xvp, td); 6537185029Spjd error = namei(&nd); 6538185029Spjd vp = nd.ni_vp; 6539185029Spjd NDFREE(&nd, NDF_ONLY_PNBUF); 6540185029Spjd if (error != 0) { 6541185029Spjd ZFS_EXIT(zfsvfs); 6542185029Spjd return (error); 6543185029Spjd } 6544185029Spjd 6545185029Spjd auio.uio_iov = &aiov; 6546185029Spjd auio.uio_iovcnt = 1; 6547185029Spjd auio.uio_segflg = UIO_SYSSPACE; 6548185029Spjd auio.uio_td = td; 6549185029Spjd auio.uio_rw = UIO_READ; 6550185029Spjd auio.uio_offset = 0; 6551185029Spjd 6552185029Spjd do { 6553185029Spjd u_char nlen; 6554185029Spjd 6555185029Spjd aiov.iov_base = (void *)dirbuf; 6556185029Spjd aiov.iov_len = sizeof(dirbuf); 6557185029Spjd auio.uio_resid = sizeof(dirbuf); 6558185029Spjd error = VOP_READDIR(vp, &auio, ap->a_cred, &eof, NULL, NULL); 6559185029Spjd done = sizeof(dirbuf) - auio.uio_resid; 6560185029Spjd if (error != 0) 6561185029Spjd break; 6562185029Spjd for (pos = 0; pos < done;) { 6563185029Spjd dp = (struct dirent *)(dirbuf + pos); 6564185029Spjd pos += dp->d_reclen; 6565185029Spjd /* 6566185029Spjd * XXX: Temporarily we also accept DT_UNKNOWN, as this 6567185029Spjd * is what we get when attribute was created on Solaris. 6568185029Spjd */ 6569185029Spjd if (dp->d_type != DT_REG && dp->d_type != DT_UNKNOWN) 6570185029Spjd continue; 6571185029Spjd if (plen == 0 && strncmp(dp->d_name, "freebsd:", 8) == 0) 6572185029Spjd continue; 6573185029Spjd else if (strncmp(dp->d_name, attrprefix, plen) != 0) 6574185029Spjd continue; 6575185029Spjd nlen = dp->d_namlen - plen; 6576185029Spjd if (sizep != NULL) 6577185029Spjd *sizep += 1 + nlen; 6578185029Spjd else if (uio != NULL) { 6579185029Spjd /* 6580185029Spjd * Format of extattr name entry is one byte for 6581185029Spjd * length and the rest for name. 6582185029Spjd */ 6583185029Spjd error = uiomove(&nlen, 1, uio->uio_rw, uio); 6584185029Spjd if (error == 0) { 6585185029Spjd error = uiomove(dp->d_name + plen, nlen, 6586185029Spjd uio->uio_rw, uio); 6587185029Spjd } 6588185029Spjd if (error != 0) 6589185029Spjd break; 6590185029Spjd } 6591185029Spjd } 6592185029Spjd } while (!eof && error == 0); 6593185029Spjd 6594185029Spjd vput(vp); 6595185029Spjd ZFS_EXIT(zfsvfs); 6596185029Spjd 6597185029Spjd return (error); 6598185029Spjd} 6599185029Spjd 6600192800Straszint 6601192800Straszzfs_freebsd_getacl(ap) 6602192800Strasz struct vop_getacl_args /* { 6603192800Strasz struct vnode *vp; 6604192800Strasz acl_type_t type; 6605192800Strasz struct acl *aclp; 6606192800Strasz struct ucred *cred; 6607192800Strasz struct thread *td; 6608192800Strasz } */ *ap; 6609192800Strasz{ 6610192800Strasz int error; 6611192800Strasz vsecattr_t vsecattr; 6612192800Strasz 6613192800Strasz if (ap->a_type != ACL_TYPE_NFS4) 6614197435Strasz return (EINVAL); 6615192800Strasz 6616192800Strasz vsecattr.vsa_mask = VSA_ACE | VSA_ACECNT; 6617192800Strasz if (error = zfs_getsecattr(ap->a_vp, &vsecattr, 0, ap->a_cred, NULL)) 6618192800Strasz return (error); 6619192800Strasz 6620192800Strasz error = acl_from_aces(ap->a_aclp, vsecattr.vsa_aclentp, vsecattr.vsa_aclcnt); 6621196303Spjd if (vsecattr.vsa_aclentp != NULL) 6622196303Spjd kmem_free(vsecattr.vsa_aclentp, vsecattr.vsa_aclentsz); 6623192800Strasz 6624196303Spjd return (error); 6625192800Strasz} 6626192800Strasz 6627192800Straszint 6628192800Straszzfs_freebsd_setacl(ap) 6629192800Strasz struct vop_setacl_args /* { 6630192800Strasz struct vnode *vp; 6631192800Strasz acl_type_t type; 6632192800Strasz struct acl *aclp; 6633192800Strasz struct ucred *cred; 6634192800Strasz struct thread *td; 6635192800Strasz } */ *ap; 6636192800Strasz{ 6637192800Strasz int error; 6638192800Strasz vsecattr_t vsecattr; 6639192800Strasz int aclbsize; /* size of acl list in bytes */ 6640192800Strasz aclent_t *aaclp; 6641192800Strasz 6642192800Strasz if (ap->a_type != ACL_TYPE_NFS4) 6643197435Strasz return (EINVAL); 6644192800Strasz 6645192800Strasz if (ap->a_aclp->acl_cnt < 1 || ap->a_aclp->acl_cnt > MAX_ACL_ENTRIES) 6646192800Strasz return (EINVAL); 6647192800Strasz 6648192800Strasz /* 6649196949Strasz * With NFSv4 ACLs, chmod(2) may need to add additional entries, 6650192800Strasz * splitting every entry into two and appending "canonical six" 6651192800Strasz * entries at the end. Don't allow for setting an ACL that would 6652192800Strasz * cause chmod(2) to run out of ACL entries. 6653192800Strasz */ 6654192800Strasz if (ap->a_aclp->acl_cnt * 2 + 6 > ACL_MAX_ENTRIES) 6655192800Strasz return (ENOSPC); 6656192800Strasz 6657208030Strasz error = acl_nfs4_check(ap->a_aclp, ap->a_vp->v_type == VDIR); 6658208030Strasz if (error != 0) 6659208030Strasz return (error); 6660208030Strasz 6661192800Strasz vsecattr.vsa_mask = VSA_ACE; 6662192800Strasz aclbsize = ap->a_aclp->acl_cnt * sizeof(ace_t); 6663192800Strasz vsecattr.vsa_aclentp = kmem_alloc(aclbsize, KM_SLEEP); 6664192800Strasz aaclp = vsecattr.vsa_aclentp; 6665192800Strasz vsecattr.vsa_aclentsz = aclbsize; 6666192800Strasz 6667192800Strasz aces_from_acl(vsecattr.vsa_aclentp, &vsecattr.vsa_aclcnt, ap->a_aclp); 6668192800Strasz error = zfs_setsecattr(ap->a_vp, &vsecattr, 0, ap->a_cred, NULL); 6669192800Strasz kmem_free(aaclp, aclbsize); 6670192800Strasz 6671192800Strasz return (error); 6672192800Strasz} 6673192800Strasz 6674192800Straszint 6675192800Straszzfs_freebsd_aclcheck(ap) 6676192800Strasz struct vop_aclcheck_args /* { 6677192800Strasz struct vnode *vp; 6678192800Strasz acl_type_t type; 6679192800Strasz struct acl *aclp; 6680192800Strasz struct ucred *cred; 6681192800Strasz struct thread *td; 6682192800Strasz } */ *ap; 6683192800Strasz{ 6684192800Strasz 6685192800Strasz return (EOPNOTSUPP); 6686192800Strasz} 6687192800Strasz 6688168404Spjdstruct vop_vector zfs_vnodeops; 6689168404Spjdstruct vop_vector zfs_fifoops; 6690209962Smmstruct vop_vector zfs_shareops; 6691168404Spjd 6692168404Spjdstruct vop_vector zfs_vnodeops = { 6693185029Spjd .vop_default = &default_vnodeops, 6694185029Spjd .vop_inactive = zfs_freebsd_inactive, 6695185029Spjd .vop_reclaim = zfs_freebsd_reclaim, 6696185029Spjd .vop_access = zfs_freebsd_access, 6697168404Spjd#ifdef FREEBSD_NAMECACHE 6698185029Spjd .vop_lookup = vfs_cache_lookup, 6699185029Spjd .vop_cachedlookup = zfs_freebsd_lookup, 6700168404Spjd#else 6701185029Spjd .vop_lookup = zfs_freebsd_lookup, 6702168404Spjd#endif 6703185029Spjd .vop_getattr = zfs_freebsd_getattr, 6704185029Spjd .vop_setattr = zfs_freebsd_setattr, 6705185029Spjd .vop_create = zfs_freebsd_create, 6706185029Spjd .vop_mknod = zfs_freebsd_create, 6707185029Spjd .vop_mkdir = zfs_freebsd_mkdir, 6708185029Spjd .vop_readdir = zfs_freebsd_readdir, 6709185029Spjd .vop_fsync = zfs_freebsd_fsync, 6710185029Spjd .vop_open = zfs_freebsd_open, 6711185029Spjd .vop_close = zfs_freebsd_close, 6712185029Spjd .vop_rmdir = zfs_freebsd_rmdir, 6713185029Spjd .vop_ioctl = zfs_freebsd_ioctl, 6714185029Spjd .vop_link = zfs_freebsd_link, 6715185029Spjd .vop_symlink = zfs_freebsd_symlink, 6716185029Spjd .vop_readlink = zfs_freebsd_readlink, 6717185029Spjd .vop_read = zfs_freebsd_read, 6718185029Spjd .vop_write = zfs_freebsd_write, 6719185029Spjd .vop_remove = zfs_freebsd_remove, 6720185029Spjd .vop_rename = zfs_freebsd_rename, 6721185029Spjd .vop_pathconf = zfs_freebsd_pathconf, 6722185029Spjd .vop_bmap = VOP_EOPNOTSUPP, 6723185029Spjd .vop_fid = zfs_freebsd_fid, 6724185029Spjd .vop_getextattr = zfs_getextattr, 6725185029Spjd .vop_deleteextattr = zfs_deleteextattr, 6726185029Spjd .vop_setextattr = zfs_setextattr, 6727185029Spjd .vop_listextattr = zfs_listextattr, 6728192800Strasz .vop_getacl = zfs_freebsd_getacl, 6729192800Strasz .vop_setacl = zfs_freebsd_setacl, 6730192800Strasz .vop_aclcheck = zfs_freebsd_aclcheck, 6731213937Savg .vop_getpages = zfs_freebsd_getpages, 6732168404Spjd}; 6733168404Spjd 6734169170Spjdstruct vop_vector zfs_fifoops = { 6735185029Spjd .vop_default = &fifo_specops, 6736200162Skib .vop_fsync = zfs_freebsd_fsync, 6737185029Spjd .vop_access = zfs_freebsd_access, 6738185029Spjd .vop_getattr = zfs_freebsd_getattr, 6739185029Spjd .vop_inactive = zfs_freebsd_inactive, 6740185029Spjd .vop_read = VOP_PANIC, 6741185029Spjd .vop_reclaim = zfs_freebsd_reclaim, 6742185029Spjd .vop_setattr = zfs_freebsd_setattr, 6743185029Spjd .vop_write = VOP_PANIC, 6744196949Strasz .vop_pathconf = zfs_freebsd_fifo_pathconf, 6745185029Spjd .vop_fid = zfs_freebsd_fid, 6746192800Strasz .vop_getacl = zfs_freebsd_getacl, 6747192800Strasz .vop_setacl = zfs_freebsd_setacl, 6748192800Strasz .vop_aclcheck = zfs_freebsd_aclcheck, 6749168404Spjd}; 6750209962Smm 6751209962Smm/* 6752209962Smm * special share hidden files vnode operations template 6753209962Smm */ 6754209962Smmstruct vop_vector zfs_shareops = { 6755209962Smm .vop_default = &default_vnodeops, 6756209962Smm .vop_access = zfs_freebsd_access, 6757209962Smm .vop_inactive = zfs_freebsd_inactive, 6758209962Smm .vop_reclaim = zfs_freebsd_reclaim, 6759209962Smm .vop_fid = zfs_freebsd_fid, 6760209962Smm .vop_pathconf = zfs_freebsd_pathconf, 6761209962Smm}; 6762