zfs_vnops.c revision 210470
1/* 2 * CDDL HEADER START 3 * 4 * The contents of this file are subject to the terms of the 5 * Common Development and Distribution License (the "License"). 6 * You may not use this file except in compliance with the License. 7 * 8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9 * or http://www.opensolaris.org/os/licensing. 10 * See the License for the specific language governing permissions 11 * and limitations under the License. 12 * 13 * When distributing Covered Code, include this CDDL HEADER in each 14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15 * If applicable, add the following below this CDDL HEADER, with the 16 * fields enclosed by brackets "[]" replaced with your own identifying 17 * information: Portions Copyright [yyyy] [name of copyright owner] 18 * 19 * CDDL HEADER END 20 */ 21/* 22 * Copyright 2009 Sun Microsystems, Inc. All rights reserved. 23 * Use is subject to license terms. 24 */ 25 26/* Portions Copyright 2007 Jeremy Teo */ 27 28#include <sys/types.h> 29#include <sys/param.h> 30#include <sys/time.h> 31#include <sys/systm.h> 32#include <sys/sysmacros.h> 33#include <sys/resource.h> 34#include <sys/vfs.h> 35#include <sys/vnode.h> 36#include <sys/file.h> 37#include <sys/stat.h> 38#include <sys/kmem.h> 39#include <sys/taskq.h> 40#include <sys/uio.h> 41#include <sys/atomic.h> 42#include <sys/namei.h> 43#include <sys/mman.h> 44#include <sys/cmn_err.h> 45#include <sys/errno.h> 46#include <sys/unistd.h> 47#include <sys/zfs_dir.h> 48#include <sys/zfs_ioctl.h> 49#include <sys/fs/zfs.h> 50#include <sys/dmu.h> 51#include <sys/spa.h> 52#include <sys/txg.h> 53#include <sys/dbuf.h> 54#include <sys/zap.h> 55#include <sys/dirent.h> 56#include <sys/policy.h> 57#include <sys/sunddi.h> 58#include <sys/filio.h> 59#include <sys/sid.h> 60#include <sys/zfs_ctldir.h> 61#include <sys/zfs_fuid.h> 62#include <sys/dnlc.h> 63#include <sys/zfs_rlock.h> 64#include <sys/extdirent.h> 65#include <sys/kidmap.h> 66#include <sys/bio.h> 67#include <sys/buf.h> 68#include <sys/sf_buf.h> 69#include <sys/sched.h> 70#include <sys/acl.h> 71 72/* 73 * Programming rules. 74 * 75 * Each vnode op performs some logical unit of work. To do this, the ZPL must 76 * properly lock its in-core state, create a DMU transaction, do the work, 77 * record this work in the intent log (ZIL), commit the DMU transaction, 78 * and wait for the intent log to commit if it is a synchronous operation. 79 * Moreover, the vnode ops must work in both normal and log replay context. 80 * The ordering of events is important to avoid deadlocks and references 81 * to freed memory. The example below illustrates the following Big Rules: 82 * 83 * (1) A check must be made in each zfs thread for a mounted file system. 84 * This is done avoiding races using ZFS_ENTER(zfsvfs). 85 * A ZFS_EXIT(zfsvfs) is needed before all returns. Any znodes 86 * must be checked with ZFS_VERIFY_ZP(zp). Both of these macros 87 * can return EIO from the calling function. 88 * 89 * (2) VN_RELE() should always be the last thing except for zil_commit() 90 * (if necessary) and ZFS_EXIT(). This is for 3 reasons: 91 * First, if it's the last reference, the vnode/znode 92 * can be freed, so the zp may point to freed memory. Second, the last 93 * reference will call zfs_zinactive(), which may induce a lot of work -- 94 * pushing cached pages (which acquires range locks) and syncing out 95 * cached atime changes. Third, zfs_zinactive() may require a new tx, 96 * which could deadlock the system if you were already holding one. 97 * If you must call VN_RELE() within a tx then use VN_RELE_ASYNC(). 98 * 99 * (3) All range locks must be grabbed before calling dmu_tx_assign(), 100 * as they can span dmu_tx_assign() calls. 101 * 102 * (4) Always pass TXG_NOWAIT as the second argument to dmu_tx_assign(). 103 * This is critical because we don't want to block while holding locks. 104 * Note, in particular, that if a lock is sometimes acquired before 105 * the tx assigns, and sometimes after (e.g. z_lock), then failing to 106 * use a non-blocking assign can deadlock the system. The scenario: 107 * 108 * Thread A has grabbed a lock before calling dmu_tx_assign(). 109 * Thread B is in an already-assigned tx, and blocks for this lock. 110 * Thread A calls dmu_tx_assign(TXG_WAIT) and blocks in txg_wait_open() 111 * forever, because the previous txg can't quiesce until B's tx commits. 112 * 113 * If dmu_tx_assign() returns ERESTART and zfsvfs->z_assign is TXG_NOWAIT, 114 * then drop all locks, call dmu_tx_wait(), and try again. 115 * 116 * (5) If the operation succeeded, generate the intent log entry for it 117 * before dropping locks. This ensures that the ordering of events 118 * in the intent log matches the order in which they actually occurred. 119 * During ZIL replay the zfs_log_* functions will update the sequence 120 * number to indicate the zil transaction has replayed. 121 * 122 * (6) At the end of each vnode op, the DMU tx must always commit, 123 * regardless of whether there were any errors. 124 * 125 * (7) After dropping all locks, invoke zil_commit(zilog, seq, foid) 126 * to ensure that synchronous semantics are provided when necessary. 127 * 128 * In general, this is how things should be ordered in each vnode op: 129 * 130 * ZFS_ENTER(zfsvfs); // exit if unmounted 131 * top: 132 * zfs_dirent_lock(&dl, ...) // lock directory entry (may VN_HOLD()) 133 * rw_enter(...); // grab any other locks you need 134 * tx = dmu_tx_create(...); // get DMU tx 135 * dmu_tx_hold_*(); // hold each object you might modify 136 * error = dmu_tx_assign(tx, TXG_NOWAIT); // try to assign 137 * if (error) { 138 * rw_exit(...); // drop locks 139 * zfs_dirent_unlock(dl); // unlock directory entry 140 * VN_RELE(...); // release held vnodes 141 * if (error == ERESTART) { 142 * dmu_tx_wait(tx); 143 * dmu_tx_abort(tx); 144 * goto top; 145 * } 146 * dmu_tx_abort(tx); // abort DMU tx 147 * ZFS_EXIT(zfsvfs); // finished in zfs 148 * return (error); // really out of space 149 * } 150 * error = do_real_work(); // do whatever this VOP does 151 * if (error == 0) 152 * zfs_log_*(...); // on success, make ZIL entry 153 * dmu_tx_commit(tx); // commit DMU tx -- error or not 154 * rw_exit(...); // drop locks 155 * zfs_dirent_unlock(dl); // unlock directory entry 156 * VN_RELE(...); // release held vnodes 157 * zil_commit(zilog, seq, foid); // synchronous when necessary 158 * ZFS_EXIT(zfsvfs); // finished in zfs 159 * return (error); // done, report error 160 */ 161 162/* ARGSUSED */ 163static int 164zfs_open(vnode_t **vpp, int flag, cred_t *cr, caller_context_t *ct) 165{ 166 znode_t *zp = VTOZ(*vpp); 167 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 168 169 ZFS_ENTER(zfsvfs); 170 ZFS_VERIFY_ZP(zp); 171 172 if ((flag & FWRITE) && (zp->z_phys->zp_flags & ZFS_APPENDONLY) && 173 ((flag & FAPPEND) == 0)) { 174 ZFS_EXIT(zfsvfs); 175 return (EPERM); 176 } 177 178 if (!zfs_has_ctldir(zp) && zp->z_zfsvfs->z_vscan && 179 ZTOV(zp)->v_type == VREG && 180 !(zp->z_phys->zp_flags & ZFS_AV_QUARANTINED) && 181 zp->z_phys->zp_size > 0) { 182 if (fs_vscan(*vpp, cr, 0) != 0) { 183 ZFS_EXIT(zfsvfs); 184 return (EACCES); 185 } 186 } 187 188 /* Keep a count of the synchronous opens in the znode */ 189 if (flag & (FSYNC | FDSYNC)) 190 atomic_inc_32(&zp->z_sync_cnt); 191 192 ZFS_EXIT(zfsvfs); 193 return (0); 194} 195 196/* ARGSUSED */ 197static int 198zfs_close(vnode_t *vp, int flag, int count, offset_t offset, cred_t *cr, 199 caller_context_t *ct) 200{ 201 znode_t *zp = VTOZ(vp); 202 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 203 204 /* 205 * Clean up any locks held by this process on the vp. 206 */ 207 cleanlocks(vp, ddi_get_pid(), 0); 208 cleanshares(vp, ddi_get_pid()); 209 210 ZFS_ENTER(zfsvfs); 211 ZFS_VERIFY_ZP(zp); 212 213 /* Decrement the synchronous opens in the znode */ 214 if ((flag & (FSYNC | FDSYNC)) && (count == 1)) 215 atomic_dec_32(&zp->z_sync_cnt); 216 217 if (!zfs_has_ctldir(zp) && zp->z_zfsvfs->z_vscan && 218 ZTOV(zp)->v_type == VREG && 219 !(zp->z_phys->zp_flags & ZFS_AV_QUARANTINED) && 220 zp->z_phys->zp_size > 0) 221 VERIFY(fs_vscan(vp, cr, 1) == 0); 222 223 ZFS_EXIT(zfsvfs); 224 return (0); 225} 226 227/* 228 * Lseek support for finding holes (cmd == _FIO_SEEK_HOLE) and 229 * data (cmd == _FIO_SEEK_DATA). "off" is an in/out parameter. 230 */ 231static int 232zfs_holey(vnode_t *vp, u_long cmd, offset_t *off) 233{ 234 znode_t *zp = VTOZ(vp); 235 uint64_t noff = (uint64_t)*off; /* new offset */ 236 uint64_t file_sz; 237 int error; 238 boolean_t hole; 239 240 file_sz = zp->z_phys->zp_size; 241 if (noff >= file_sz) { 242 return (ENXIO); 243 } 244 245 if (cmd == _FIO_SEEK_HOLE) 246 hole = B_TRUE; 247 else 248 hole = B_FALSE; 249 250 error = dmu_offset_next(zp->z_zfsvfs->z_os, zp->z_id, hole, &noff); 251 252 /* end of file? */ 253 if ((error == ESRCH) || (noff > file_sz)) { 254 /* 255 * Handle the virtual hole at the end of file. 256 */ 257 if (hole) { 258 *off = file_sz; 259 return (0); 260 } 261 return (ENXIO); 262 } 263 264 if (noff < *off) 265 return (error); 266 *off = noff; 267 return (error); 268} 269 270/* ARGSUSED */ 271static int 272zfs_ioctl(vnode_t *vp, u_long com, intptr_t data, int flag, cred_t *cred, 273 int *rvalp, caller_context_t *ct) 274{ 275 offset_t off; 276 int error; 277 zfsvfs_t *zfsvfs; 278 znode_t *zp; 279 280 switch (com) { 281 case _FIOFFS: 282 return (0); 283 284 /* 285 * The following two ioctls are used by bfu. Faking out, 286 * necessary to avoid bfu errors. 287 */ 288 case _FIOGDIO: 289 case _FIOSDIO: 290 return (0); 291 292 case _FIO_SEEK_DATA: 293 case _FIO_SEEK_HOLE: 294 if (ddi_copyin((void *)data, &off, sizeof (off), flag)) 295 return (EFAULT); 296 297 zp = VTOZ(vp); 298 zfsvfs = zp->z_zfsvfs; 299 ZFS_ENTER(zfsvfs); 300 ZFS_VERIFY_ZP(zp); 301 302 /* offset parameter is in/out */ 303 error = zfs_holey(vp, com, &off); 304 ZFS_EXIT(zfsvfs); 305 if (error) 306 return (error); 307 if (ddi_copyout(&off, (void *)data, sizeof (off), flag)) 308 return (EFAULT); 309 return (0); 310 } 311 return (ENOTTY); 312} 313 314static vm_page_t 315page_lookup(vnode_t *vp, int64_t start, int64_t off, int64_t nbytes) 316{ 317 vm_object_t obj; 318 vm_page_t pp; 319 320 obj = vp->v_object; 321 VM_OBJECT_LOCK_ASSERT(obj, MA_OWNED); 322 323 for (;;) { 324 if ((pp = vm_page_lookup(obj, OFF_TO_IDX(start))) != NULL && 325 vm_page_is_valid(pp, (vm_offset_t)off, nbytes)) { 326 if (vm_page_sleep_if_busy(pp, FALSE, "zfsmwb")) 327 continue; 328 vm_page_busy(pp); 329 vm_page_lock_queues(); 330 vm_page_undirty(pp); 331 vm_page_unlock_queues(); 332 } else { 333 if (__predict_false(obj->cache != NULL)) { 334 vm_page_cache_free(obj, OFF_TO_IDX(start), 335 OFF_TO_IDX(start) + 1); 336 } 337 pp = NULL; 338 } 339 break; 340 } 341 return (pp); 342} 343 344static void 345page_unlock(vm_page_t pp) 346{ 347 348 vm_page_wakeup(pp); 349} 350 351static caddr_t 352zfs_map_page(vm_page_t pp, struct sf_buf **sfp) 353{ 354 355 sched_pin(); 356 *sfp = sf_buf_alloc(pp, SFB_CPUPRIVATE); 357 return ((caddr_t)sf_buf_kva(*sfp)); 358} 359 360static void 361zfs_unmap_page(struct sf_buf *sf) 362{ 363 364 sf_buf_free(sf); 365 sched_unpin(); 366} 367 368 369/* 370 * When a file is memory mapped, we must keep the IO data synchronized 371 * between the DMU cache and the memory mapped pages. What this means: 372 * 373 * On Write: If we find a memory mapped page, we write to *both* 374 * the page and the dmu buffer. 375 */ 376 377static void 378update_pages(vnode_t *vp, int64_t start, int len, objset_t *os, uint64_t oid, 379 int segflg, dmu_tx_t *tx) 380{ 381 vm_object_t obj; 382 struct sf_buf *sf; 383 int64_t off; 384 385 ASSERT(vp->v_mount != NULL); 386 obj = vp->v_object; 387 ASSERT(obj != NULL); 388 389 off = start & PAGEOFFSET; 390 VM_OBJECT_LOCK(obj); 391 for (start &= PAGEMASK; len > 0; start += PAGESIZE) { 392 vm_page_t pp; 393 uint64_t nbytes = MIN(PAGESIZE - off, len); 394 395 if ((pp = page_lookup(vp, start, off, nbytes)) != NULL) { 396 caddr_t va; 397 398 VM_OBJECT_UNLOCK(obj); 399 va = zfs_map_page(pp, &sf); 400 if (segflg == UIO_NOCOPY) { 401 (void) dmu_write(os, oid, start+off, nbytes, 402 va+off, tx); 403 } else { 404 (void) dmu_read(os, oid, start+off, nbytes, 405 va+off, DMU_READ_PREFETCH);; 406 } 407 zfs_unmap_page(sf); 408 VM_OBJECT_LOCK(obj); 409 page_unlock(pp); 410 411 } 412 len -= nbytes; 413 off = 0; 414 } 415 VM_OBJECT_UNLOCK(obj); 416} 417 418/* 419 * When a file is memory mapped, we must keep the IO data synchronized 420 * between the DMU cache and the memory mapped pages. What this means: 421 * 422 * On Read: We "read" preferentially from memory mapped pages, 423 * else we default from the dmu buffer. 424 * 425 * NOTE: We will always "break up" the IO into PAGESIZE uiomoves when 426 * the file is memory mapped. 427 */ 428static int 429mappedread(vnode_t *vp, int nbytes, uio_t *uio) 430{ 431 znode_t *zp = VTOZ(vp); 432 objset_t *os = zp->z_zfsvfs->z_os; 433 vm_object_t obj; 434 vm_page_t m; 435 struct sf_buf *sf; 436 int64_t start, off; 437 caddr_t va; 438 int len = nbytes; 439 int error = 0; 440 uint64_t dirbytes; 441 442 ASSERT(vp->v_mount != NULL); 443 obj = vp->v_object; 444 ASSERT(obj != NULL); 445 446 start = uio->uio_loffset; 447 off = start & PAGEOFFSET; 448 dirbytes = 0; 449 VM_OBJECT_LOCK(obj); 450 for (start &= PAGEMASK; len > 0; start += PAGESIZE) { 451 uint64_t bytes = MIN(PAGESIZE - off, len); 452 453again: 454 if ((m = vm_page_lookup(obj, OFF_TO_IDX(start))) != NULL && 455 vm_page_is_valid(m, (vm_offset_t)off, bytes)) { 456 if (vm_page_sleep_if_busy(m, FALSE, "zfsmrb")) 457 goto again; 458 vm_page_busy(m); 459 VM_OBJECT_UNLOCK(obj); 460 if (dirbytes > 0) { 461 error = dmu_read_uio(os, zp->z_id, uio, 462 dirbytes); 463 dirbytes = 0; 464 } 465 if (error == 0) { 466 sched_pin(); 467 sf = sf_buf_alloc(m, SFB_CPUPRIVATE); 468 va = (caddr_t)sf_buf_kva(sf); 469 error = uiomove(va + off, bytes, UIO_READ, uio); 470 sf_buf_free(sf); 471 sched_unpin(); 472 } 473 VM_OBJECT_LOCK(obj); 474 vm_page_wakeup(m); 475 } else if (m != NULL && uio->uio_segflg == UIO_NOCOPY) { 476 /* 477 * The code below is here to make sendfile(2) work 478 * correctly with ZFS. As pointed out by ups@ 479 * sendfile(2) should be changed to use VOP_GETPAGES(), 480 * but it pessimize performance of sendfile/UFS, that's 481 * why I handle this special case in ZFS code. 482 */ 483 if (vm_page_sleep_if_busy(m, FALSE, "zfsmrb")) 484 goto again; 485 vm_page_busy(m); 486 VM_OBJECT_UNLOCK(obj); 487 if (dirbytes > 0) { 488 error = dmu_read_uio(os, zp->z_id, uio, 489 dirbytes); 490 dirbytes = 0; 491 } 492 if (error == 0) { 493 sched_pin(); 494 sf = sf_buf_alloc(m, SFB_CPUPRIVATE); 495 va = (caddr_t)sf_buf_kva(sf); 496 error = dmu_read(os, zp->z_id, start + off, 497 bytes, (void *)(va + off), 498 DMU_READ_PREFETCH); 499 sf_buf_free(sf); 500 sched_unpin(); 501 } 502 VM_OBJECT_LOCK(obj); 503 vm_page_wakeup(m); 504 if (error == 0) 505 uio->uio_resid -= bytes; 506 } else { 507 dirbytes += bytes; 508 } 509 len -= bytes; 510 off = 0; 511 if (error) 512 break; 513 } 514 VM_OBJECT_UNLOCK(obj); 515 if (error == 0 && dirbytes > 0) 516 error = dmu_read_uio(os, zp->z_id, uio, dirbytes); 517 return (error); 518} 519 520offset_t zfs_read_chunk_size = 1024 * 1024; /* Tunable */ 521 522/* 523 * Read bytes from specified file into supplied buffer. 524 * 525 * IN: vp - vnode of file to be read from. 526 * uio - structure supplying read location, range info, 527 * and return buffer. 528 * ioflag - SYNC flags; used to provide FRSYNC semantics. 529 * cr - credentials of caller. 530 * ct - caller context 531 * 532 * OUT: uio - updated offset and range, buffer filled. 533 * 534 * RETURN: 0 if success 535 * error code if failure 536 * 537 * Side Effects: 538 * vp - atime updated if byte count > 0 539 */ 540/* ARGSUSED */ 541static int 542zfs_read(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct) 543{ 544 znode_t *zp = VTOZ(vp); 545 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 546 objset_t *os; 547 ssize_t n, nbytes; 548 int error; 549 rl_t *rl; 550 551 ZFS_ENTER(zfsvfs); 552 ZFS_VERIFY_ZP(zp); 553 os = zfsvfs->z_os; 554 555 if (zp->z_phys->zp_flags & ZFS_AV_QUARANTINED) { 556 ZFS_EXIT(zfsvfs); 557 return (EACCES); 558 } 559 560 /* 561 * Validate file offset 562 */ 563 if (uio->uio_loffset < (offset_t)0) { 564 ZFS_EXIT(zfsvfs); 565 return (EINVAL); 566 } 567 568 /* 569 * Fasttrack empty reads 570 */ 571 if (uio->uio_resid == 0) { 572 ZFS_EXIT(zfsvfs); 573 return (0); 574 } 575 576 /* 577 * Check for mandatory locks 578 */ 579 if (MANDMODE((mode_t)zp->z_phys->zp_mode)) { 580 if (error = chklock(vp, FREAD, 581 uio->uio_loffset, uio->uio_resid, uio->uio_fmode, ct)) { 582 ZFS_EXIT(zfsvfs); 583 return (error); 584 } 585 } 586 587 /* 588 * If we're in FRSYNC mode, sync out this znode before reading it. 589 */ 590 if (ioflag & FRSYNC) 591 zil_commit(zfsvfs->z_log, zp->z_last_itx, zp->z_id); 592 593 /* 594 * Lock the range against changes. 595 */ 596 rl = zfs_range_lock(zp, uio->uio_loffset, uio->uio_resid, RL_READER); 597 598 /* 599 * If we are reading past end-of-file we can skip 600 * to the end; but we might still need to set atime. 601 */ 602 if (uio->uio_loffset >= zp->z_phys->zp_size) { 603 error = 0; 604 goto out; 605 } 606 607 ASSERT(uio->uio_loffset < zp->z_phys->zp_size); 608 n = MIN(uio->uio_resid, zp->z_phys->zp_size - uio->uio_loffset); 609 610 while (n > 0) { 611 nbytes = MIN(n, zfs_read_chunk_size - 612 P2PHASE(uio->uio_loffset, zfs_read_chunk_size)); 613 614 if (vn_has_cached_data(vp)) 615 error = mappedread(vp, nbytes, uio); 616 else 617 error = dmu_read_uio(os, zp->z_id, uio, nbytes); 618 if (error) { 619 /* convert checksum errors into IO errors */ 620 if (error == ECKSUM) 621 error = EIO; 622 break; 623 } 624 625 n -= nbytes; 626 } 627 628out: 629 zfs_range_unlock(rl); 630 631 ZFS_ACCESSTIME_STAMP(zfsvfs, zp); 632 ZFS_EXIT(zfsvfs); 633 return (error); 634} 635 636/* 637 * Fault in the pages of the first n bytes specified by the uio structure. 638 * 1 byte in each page is touched and the uio struct is unmodified. 639 * Any error will exit this routine as this is only a best 640 * attempt to get the pages resident. This is a copy of ufs_trans_touch(). 641 */ 642static void 643zfs_prefault_write(ssize_t n, struct uio *uio) 644{ 645 struct iovec *iov; 646 ulong_t cnt, incr; 647 caddr_t p; 648 649 if (uio->uio_segflg != UIO_USERSPACE) 650 return; 651 652 iov = uio->uio_iov; 653 654 while (n) { 655 cnt = MIN(iov->iov_len, n); 656 if (cnt == 0) { 657 /* empty iov entry */ 658 iov++; 659 continue; 660 } 661 n -= cnt; 662 /* 663 * touch each page in this segment. 664 */ 665 p = iov->iov_base; 666 while (cnt) { 667 if (fubyte(p) == -1) 668 return; 669 incr = MIN(cnt, PAGESIZE); 670 p += incr; 671 cnt -= incr; 672 } 673 /* 674 * touch the last byte in case it straddles a page. 675 */ 676 p--; 677 if (fubyte(p) == -1) 678 return; 679 iov++; 680 } 681} 682 683/* 684 * Write the bytes to a file. 685 * 686 * IN: vp - vnode of file to be written to. 687 * uio - structure supplying write location, range info, 688 * and data buffer. 689 * ioflag - IO_APPEND flag set if in append mode. 690 * cr - credentials of caller. 691 * ct - caller context (NFS/CIFS fem monitor only) 692 * 693 * OUT: uio - updated offset and range. 694 * 695 * RETURN: 0 if success 696 * error code if failure 697 * 698 * Timestamps: 699 * vp - ctime|mtime updated if byte count > 0 700 */ 701/* ARGSUSED */ 702static int 703zfs_write(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct) 704{ 705 znode_t *zp = VTOZ(vp); 706 rlim64_t limit = MAXOFFSET_T; 707 ssize_t start_resid = uio->uio_resid; 708 ssize_t tx_bytes; 709 uint64_t end_size; 710 dmu_tx_t *tx; 711 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 712 zilog_t *zilog; 713 offset_t woff; 714 ssize_t n, nbytes; 715 rl_t *rl; 716 int max_blksz = zfsvfs->z_max_blksz; 717 uint64_t pflags; 718 int error; 719 arc_buf_t *abuf; 720 721 /* 722 * Fasttrack empty write 723 */ 724 n = start_resid; 725 if (n == 0) 726 return (0); 727 728 if (limit == RLIM64_INFINITY || limit > MAXOFFSET_T) 729 limit = MAXOFFSET_T; 730 731 ZFS_ENTER(zfsvfs); 732 ZFS_VERIFY_ZP(zp); 733 734 /* 735 * If immutable or not appending then return EPERM 736 */ 737 pflags = zp->z_phys->zp_flags; 738 if ((pflags & (ZFS_IMMUTABLE | ZFS_READONLY)) || 739 ((pflags & ZFS_APPENDONLY) && !(ioflag & FAPPEND) && 740 (uio->uio_loffset < zp->z_phys->zp_size))) { 741 ZFS_EXIT(zfsvfs); 742 return (EPERM); 743 } 744 745 zilog = zfsvfs->z_log; 746 747 /* 748 * Pre-fault the pages to ensure slow (eg NFS) pages 749 * don't hold up txg. 750 */ 751 zfs_prefault_write(n, uio); 752 753 /* 754 * If in append mode, set the io offset pointer to eof. 755 */ 756 if (ioflag & IO_APPEND) { 757 /* 758 * Range lock for a file append: 759 * The value for the start of range will be determined by 760 * zfs_range_lock() (to guarantee append semantics). 761 * If this write will cause the block size to increase, 762 * zfs_range_lock() will lock the entire file, so we must 763 * later reduce the range after we grow the block size. 764 */ 765 rl = zfs_range_lock(zp, 0, n, RL_APPEND); 766 if (rl->r_len == UINT64_MAX) { 767 /* overlocked, zp_size can't change */ 768 woff = uio->uio_loffset = zp->z_phys->zp_size; 769 } else { 770 woff = uio->uio_loffset = rl->r_off; 771 } 772 } else { 773 woff = uio->uio_loffset; 774 /* 775 * Validate file offset 776 */ 777 if (woff < 0) { 778 ZFS_EXIT(zfsvfs); 779 return (EINVAL); 780 } 781 782 /* 783 * If we need to grow the block size then zfs_range_lock() 784 * will lock a wider range than we request here. 785 * Later after growing the block size we reduce the range. 786 */ 787 rl = zfs_range_lock(zp, woff, n, RL_WRITER); 788 } 789 790 if (woff >= limit) { 791 zfs_range_unlock(rl); 792 ZFS_EXIT(zfsvfs); 793 return (EFBIG); 794 } 795 796 if ((woff + n) > limit || woff > (limit - n)) 797 n = limit - woff; 798 799 /* 800 * Check for mandatory locks 801 */ 802 if (MANDMODE((mode_t)zp->z_phys->zp_mode) && 803 (error = chklock(vp, FWRITE, woff, n, uio->uio_fmode, ct)) != 0) { 804 zfs_range_unlock(rl); 805 ZFS_EXIT(zfsvfs); 806 return (error); 807 } 808 end_size = MAX(zp->z_phys->zp_size, woff + n); 809 810 /* 811 * Write the file in reasonable size chunks. Each chunk is written 812 * in a separate transaction; this keeps the intent log records small 813 * and allows us to do more fine-grained space accounting. 814 */ 815 while (n > 0) { 816 abuf = NULL; 817 woff = uio->uio_loffset; 818 819again: 820 if (zfs_usergroup_overquota(zfsvfs, 821 B_FALSE, zp->z_phys->zp_uid) || 822 zfs_usergroup_overquota(zfsvfs, 823 B_TRUE, zp->z_phys->zp_gid)) { 824 if (abuf != NULL) 825 dmu_return_arcbuf(abuf); 826 error = EDQUOT; 827 break; 828 } 829 830 /* 831 * If dmu_assign_arcbuf() is expected to execute with minimum 832 * overhead loan an arc buffer and copy user data to it before 833 * we enter a txg. This avoids holding a txg forever while we 834 * pagefault on a hanging NFS server mapping. 835 */ 836 if (abuf == NULL && n >= max_blksz && 837 woff >= zp->z_phys->zp_size && 838 P2PHASE(woff, max_blksz) == 0 && 839 zp->z_blksz == max_blksz) { 840 size_t cbytes; 841 842 abuf = dmu_request_arcbuf(zp->z_dbuf, max_blksz); 843 ASSERT(abuf != NULL); 844 ASSERT(arc_buf_size(abuf) == max_blksz); 845 if (error = uiocopy(abuf->b_data, max_blksz, 846 UIO_WRITE, uio, &cbytes)) { 847 dmu_return_arcbuf(abuf); 848 break; 849 } 850 ASSERT(cbytes == max_blksz); 851 } 852 853 /* 854 * Start a transaction. 855 */ 856 tx = dmu_tx_create(zfsvfs->z_os); 857 dmu_tx_hold_bonus(tx, zp->z_id); 858 dmu_tx_hold_write(tx, zp->z_id, woff, MIN(n, max_blksz)); 859 error = dmu_tx_assign(tx, TXG_NOWAIT); 860 if (error) { 861 if (error == ERESTART) { 862 dmu_tx_wait(tx); 863 dmu_tx_abort(tx); 864 goto again; 865 } 866 dmu_tx_abort(tx); 867 if (abuf != NULL) 868 dmu_return_arcbuf(abuf); 869 break; 870 } 871 872 /* 873 * If zfs_range_lock() over-locked we grow the blocksize 874 * and then reduce the lock range. This will only happen 875 * on the first iteration since zfs_range_reduce() will 876 * shrink down r_len to the appropriate size. 877 */ 878 if (rl->r_len == UINT64_MAX) { 879 uint64_t new_blksz; 880 881 if (zp->z_blksz > max_blksz) { 882 ASSERT(!ISP2(zp->z_blksz)); 883 new_blksz = MIN(end_size, SPA_MAXBLOCKSIZE); 884 } else { 885 new_blksz = MIN(end_size, max_blksz); 886 } 887 zfs_grow_blocksize(zp, new_blksz, tx); 888 zfs_range_reduce(rl, woff, n); 889 } 890 891 /* 892 * XXX - should we really limit each write to z_max_blksz? 893 * Perhaps we should use SPA_MAXBLOCKSIZE chunks? 894 */ 895 nbytes = MIN(n, max_blksz - P2PHASE(woff, max_blksz)); 896 897 if (woff + nbytes > zp->z_phys->zp_size) 898 vnode_pager_setsize(vp, woff + nbytes); 899 900 if (abuf == NULL) { 901 tx_bytes = uio->uio_resid; 902 error = dmu_write_uio(zfsvfs->z_os, zp->z_id, uio, 903 nbytes, tx); 904 tx_bytes -= uio->uio_resid; 905 } else { 906 tx_bytes = nbytes; 907 ASSERT(tx_bytes == max_blksz); 908 dmu_assign_arcbuf(zp->z_dbuf, woff, abuf, tx); 909 ASSERT(tx_bytes <= uio->uio_resid); 910 uioskip(uio, tx_bytes); 911 } 912 913 /* 914 * XXXPJD: There are some cases (triggered by fsx) where 915 * vn_has_cached_data(vp) returns false when it should 916 * return true. This should be investigated. 917 */ 918#if 0 919 if (tx_bytes && vn_has_cached_data(vp)) 920#else 921 if (tx_bytes && vp->v_object != NULL) 922#endif 923 { 924 update_pages(vp, woff, tx_bytes, zfsvfs->z_os, 925 zp->z_id, uio->uio_segflg, tx); 926 } 927 928 /* 929 * If we made no progress, we're done. If we made even 930 * partial progress, update the znode and ZIL accordingly. 931 */ 932 if (tx_bytes == 0) { 933 dmu_tx_commit(tx); 934 ASSERT(error != 0); 935 break; 936 } 937 938 /* 939 * Clear Set-UID/Set-GID bits on successful write if not 940 * privileged and at least one of the excute bits is set. 941 * 942 * It would be nice to to this after all writes have 943 * been done, but that would still expose the ISUID/ISGID 944 * to another app after the partial write is committed. 945 * 946 * Note: we don't call zfs_fuid_map_id() here because 947 * user 0 is not an ephemeral uid. 948 */ 949 mutex_enter(&zp->z_acl_lock); 950 if ((zp->z_phys->zp_mode & (S_IXUSR | (S_IXUSR >> 3) | 951 (S_IXUSR >> 6))) != 0 && 952 (zp->z_phys->zp_mode & (S_ISUID | S_ISGID)) != 0 && 953 secpolicy_vnode_setid_retain(vp, cr, 954 (zp->z_phys->zp_mode & S_ISUID) != 0 && 955 zp->z_phys->zp_uid == 0) != 0) { 956 zp->z_phys->zp_mode &= ~(S_ISUID | S_ISGID); 957 } 958 mutex_exit(&zp->z_acl_lock); 959 960 /* 961 * Update time stamp. NOTE: This marks the bonus buffer as 962 * dirty, so we don't have to do it again for zp_size. 963 */ 964 zfs_time_stamper(zp, CONTENT_MODIFIED, tx); 965 966 /* 967 * Update the file size (zp_size) if it has changed; 968 * account for possible concurrent updates. 969 */ 970 while ((end_size = zp->z_phys->zp_size) < uio->uio_loffset) 971 (void) atomic_cas_64(&zp->z_phys->zp_size, end_size, 972 uio->uio_loffset); 973 zfs_log_write(zilog, tx, TX_WRITE, zp, woff, tx_bytes, ioflag); 974 dmu_tx_commit(tx); 975 976 if (error != 0) 977 break; 978 ASSERT(tx_bytes == nbytes); 979 n -= nbytes; 980 } 981 982 zfs_range_unlock(rl); 983 984 /* 985 * If we're in replay mode, or we made no progress, return error. 986 * Otherwise, it's at least a partial write, so it's successful. 987 */ 988 if (zfsvfs->z_replay || uio->uio_resid == start_resid) { 989 ZFS_EXIT(zfsvfs); 990 return (error); 991 } 992 993 if (ioflag & (FSYNC | FDSYNC)) 994 zil_commit(zilog, zp->z_last_itx, zp->z_id); 995 996 ZFS_EXIT(zfsvfs); 997 return (0); 998} 999 1000void 1001zfs_get_done(dmu_buf_t *db, void *vzgd) 1002{ 1003 zgd_t *zgd = (zgd_t *)vzgd; 1004 rl_t *rl = zgd->zgd_rl; 1005 vnode_t *vp = ZTOV(rl->r_zp); 1006 objset_t *os = rl->r_zp->z_zfsvfs->z_os; 1007 int vfslocked; 1008 1009 vfslocked = VFS_LOCK_GIANT(vp->v_vfsp); 1010 dmu_buf_rele(db, vzgd); 1011 zfs_range_unlock(rl); 1012 /* 1013 * Release the vnode asynchronously as we currently have the 1014 * txg stopped from syncing. 1015 */ 1016 VN_RELE_ASYNC(vp, dsl_pool_vnrele_taskq(dmu_objset_pool(os))); 1017 zil_add_block(zgd->zgd_zilog, zgd->zgd_bp); 1018 kmem_free(zgd, sizeof (zgd_t)); 1019 VFS_UNLOCK_GIANT(vfslocked); 1020} 1021 1022/* 1023 * Get data to generate a TX_WRITE intent log record. 1024 */ 1025int 1026zfs_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio) 1027{ 1028 zfsvfs_t *zfsvfs = arg; 1029 objset_t *os = zfsvfs->z_os; 1030 znode_t *zp; 1031 uint64_t off = lr->lr_offset; 1032 dmu_buf_t *db; 1033 rl_t *rl; 1034 zgd_t *zgd; 1035 int dlen = lr->lr_length; /* length of user data */ 1036 int error = 0; 1037 1038 ASSERT(zio); 1039 ASSERT(dlen != 0); 1040 1041 /* 1042 * Nothing to do if the file has been removed 1043 */ 1044 if (zfs_zget(zfsvfs, lr->lr_foid, &zp) != 0) 1045 return (ENOENT); 1046 if (zp->z_unlinked) { 1047 /* 1048 * Release the vnode asynchronously as we currently have the 1049 * txg stopped from syncing. 1050 */ 1051 VN_RELE_ASYNC(ZTOV(zp), 1052 dsl_pool_vnrele_taskq(dmu_objset_pool(os))); 1053 return (ENOENT); 1054 } 1055 1056 /* 1057 * Write records come in two flavors: immediate and indirect. 1058 * For small writes it's cheaper to store the data with the 1059 * log record (immediate); for large writes it's cheaper to 1060 * sync the data and get a pointer to it (indirect) so that 1061 * we don't have to write the data twice. 1062 */ 1063 if (buf != NULL) { /* immediate write */ 1064 rl = zfs_range_lock(zp, off, dlen, RL_READER); 1065 /* test for truncation needs to be done while range locked */ 1066 if (off >= zp->z_phys->zp_size) { 1067 error = ENOENT; 1068 goto out; 1069 } 1070 VERIFY(0 == dmu_read(os, lr->lr_foid, off, dlen, buf, 1071 DMU_READ_NO_PREFETCH)); 1072 } else { /* indirect write */ 1073 uint64_t boff; /* block starting offset */ 1074 1075 /* 1076 * Have to lock the whole block to ensure when it's 1077 * written out and it's checksum is being calculated 1078 * that no one can change the data. We need to re-check 1079 * blocksize after we get the lock in case it's changed! 1080 */ 1081 for (;;) { 1082 if (ISP2(zp->z_blksz)) { 1083 boff = P2ALIGN_TYPED(off, zp->z_blksz, 1084 uint64_t); 1085 } else { 1086 boff = 0; 1087 } 1088 dlen = zp->z_blksz; 1089 rl = zfs_range_lock(zp, boff, dlen, RL_READER); 1090 if (zp->z_blksz == dlen) 1091 break; 1092 zfs_range_unlock(rl); 1093 } 1094 /* test for truncation needs to be done while range locked */ 1095 if (off >= zp->z_phys->zp_size) { 1096 error = ENOENT; 1097 goto out; 1098 } 1099 zgd = (zgd_t *)kmem_alloc(sizeof (zgd_t), KM_SLEEP); 1100 zgd->zgd_rl = rl; 1101 zgd->zgd_zilog = zfsvfs->z_log; 1102 zgd->zgd_bp = &lr->lr_blkptr; 1103 VERIFY(0 == dmu_buf_hold(os, lr->lr_foid, boff, zgd, &db)); 1104 ASSERT(boff == db->db_offset); 1105 lr->lr_blkoff = off - boff; 1106 error = dmu_sync(zio, db, &lr->lr_blkptr, 1107 lr->lr_common.lrc_txg, zfs_get_done, zgd); 1108 ASSERT((error && error != EINPROGRESS) || 1109 lr->lr_length <= zp->z_blksz); 1110 if (error == 0) { 1111 /* 1112 * dmu_sync() can compress a block of zeros to a null 1113 * blkptr but the block size still needs to be passed 1114 * through to replay. 1115 */ 1116 BP_SET_LSIZE(&lr->lr_blkptr, db->db_size); 1117 zil_add_block(zfsvfs->z_log, &lr->lr_blkptr); 1118 } 1119 1120 /* 1121 * If we get EINPROGRESS, then we need to wait for a 1122 * write IO initiated by dmu_sync() to complete before 1123 * we can release this dbuf. We will finish everything 1124 * up in the zfs_get_done() callback. 1125 */ 1126 if (error == EINPROGRESS) { 1127 return (0); 1128 } else if (error == EALREADY) { 1129 lr->lr_common.lrc_txtype = TX_WRITE2; 1130 error = 0; 1131 } 1132 dmu_buf_rele(db, zgd); 1133 kmem_free(zgd, sizeof (zgd_t)); 1134 } 1135out: 1136 zfs_range_unlock(rl); 1137 /* 1138 * Release the vnode asynchronously as we currently have the 1139 * txg stopped from syncing. 1140 */ 1141 VN_RELE_ASYNC(ZTOV(zp), dsl_pool_vnrele_taskq(dmu_objset_pool(os))); 1142 return (error); 1143} 1144 1145/*ARGSUSED*/ 1146static int 1147zfs_access(vnode_t *vp, int mode, int flag, cred_t *cr, 1148 caller_context_t *ct) 1149{ 1150 znode_t *zp = VTOZ(vp); 1151 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1152 int error; 1153 1154 ZFS_ENTER(zfsvfs); 1155 ZFS_VERIFY_ZP(zp); 1156 1157 if (flag & V_ACE_MASK) 1158 error = zfs_zaccess(zp, mode, flag, B_FALSE, cr); 1159 else 1160 error = zfs_zaccess_rwx(zp, mode, flag, cr); 1161 1162 ZFS_EXIT(zfsvfs); 1163 return (error); 1164} 1165 1166/* 1167 * Lookup an entry in a directory, or an extended attribute directory. 1168 * If it exists, return a held vnode reference for it. 1169 * 1170 * IN: dvp - vnode of directory to search. 1171 * nm - name of entry to lookup. 1172 * pnp - full pathname to lookup [UNUSED]. 1173 * flags - LOOKUP_XATTR set if looking for an attribute. 1174 * rdir - root directory vnode [UNUSED]. 1175 * cr - credentials of caller. 1176 * ct - caller context 1177 * direntflags - directory lookup flags 1178 * realpnp - returned pathname. 1179 * 1180 * OUT: vpp - vnode of located entry, NULL if not found. 1181 * 1182 * RETURN: 0 if success 1183 * error code if failure 1184 * 1185 * Timestamps: 1186 * NA 1187 */ 1188/* ARGSUSED */ 1189static int 1190zfs_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, struct componentname *cnp, 1191 int nameiop, cred_t *cr, kthread_t *td, int flags) 1192{ 1193 znode_t *zdp = VTOZ(dvp); 1194 zfsvfs_t *zfsvfs = zdp->z_zfsvfs; 1195 int error; 1196 int *direntflags = NULL; 1197 void *realpnp = NULL; 1198 1199 ZFS_ENTER(zfsvfs); 1200 ZFS_VERIFY_ZP(zdp); 1201 1202 *vpp = NULL; 1203 1204 if (flags & LOOKUP_XATTR) { 1205#ifdef TODO 1206 /* 1207 * If the xattr property is off, refuse the lookup request. 1208 */ 1209 if (!(zfsvfs->z_vfs->vfs_flag & VFS_XATTR)) { 1210 ZFS_EXIT(zfsvfs); 1211 return (EINVAL); 1212 } 1213#endif 1214 1215 /* 1216 * We don't allow recursive attributes.. 1217 * Maybe someday we will. 1218 */ 1219 if (zdp->z_phys->zp_flags & ZFS_XATTR) { 1220 ZFS_EXIT(zfsvfs); 1221 return (EINVAL); 1222 } 1223 1224 if (error = zfs_get_xattrdir(VTOZ(dvp), vpp, cr, flags)) { 1225 ZFS_EXIT(zfsvfs); 1226 return (error); 1227 } 1228 1229 /* 1230 * Do we have permission to get into attribute directory? 1231 */ 1232 1233 if (error = zfs_zaccess(VTOZ(*vpp), ACE_EXECUTE, 0, 1234 B_FALSE, cr)) { 1235 VN_RELE(*vpp); 1236 *vpp = NULL; 1237 } 1238 1239 ZFS_EXIT(zfsvfs); 1240 return (error); 1241 } 1242 1243 if (dvp->v_type != VDIR) { 1244 ZFS_EXIT(zfsvfs); 1245 return (ENOTDIR); 1246 } 1247 1248 /* 1249 * Check accessibility of directory. 1250 */ 1251 1252 if (error = zfs_zaccess(zdp, ACE_EXECUTE, 0, B_FALSE, cr)) { 1253 ZFS_EXIT(zfsvfs); 1254 return (error); 1255 } 1256 1257 if (zfsvfs->z_utf8 && u8_validate(nm, strlen(nm), 1258 NULL, U8_VALIDATE_ENTIRE, &error) < 0) { 1259 ZFS_EXIT(zfsvfs); 1260 return (EILSEQ); 1261 } 1262 1263 error = zfs_dirlook(zdp, nm, vpp, flags, direntflags, realpnp); 1264 if (error == 0) { 1265 /* 1266 * Convert device special files 1267 */ 1268 if (IS_DEVVP(*vpp)) { 1269 vnode_t *svp; 1270 1271 svp = specvp(*vpp, (*vpp)->v_rdev, (*vpp)->v_type, cr); 1272 VN_RELE(*vpp); 1273 if (svp == NULL) 1274 error = ENOSYS; 1275 else 1276 *vpp = svp; 1277 } 1278 } 1279 1280 /* Translate errors and add SAVENAME when needed. */ 1281 if (cnp->cn_flags & ISLASTCN) { 1282 switch (nameiop) { 1283 case CREATE: 1284 case RENAME: 1285 if (error == ENOENT) { 1286 error = EJUSTRETURN; 1287 cnp->cn_flags |= SAVENAME; 1288 break; 1289 } 1290 /* FALLTHROUGH */ 1291 case DELETE: 1292 if (error == 0) 1293 cnp->cn_flags |= SAVENAME; 1294 break; 1295 } 1296 } 1297 if (error == 0 && (nm[0] != '.' || nm[1] != '\0')) { 1298 int ltype = 0; 1299 1300 if (cnp->cn_flags & ISDOTDOT) { 1301 ltype = VOP_ISLOCKED(dvp); 1302 VOP_UNLOCK(dvp, 0); 1303 } 1304 ZFS_EXIT(zfsvfs); 1305 error = vn_lock(*vpp, cnp->cn_lkflags); 1306 if (cnp->cn_flags & ISDOTDOT) 1307 vn_lock(dvp, ltype | LK_RETRY); 1308 if (error != 0) { 1309 VN_RELE(*vpp); 1310 *vpp = NULL; 1311 return (error); 1312 } 1313 } else { 1314 ZFS_EXIT(zfsvfs); 1315 } 1316 1317#ifdef FREEBSD_NAMECACHE 1318 /* 1319 * Insert name into cache (as non-existent) if appropriate. 1320 */ 1321 if (error == ENOENT && (cnp->cn_flags & MAKEENTRY) && nameiop != CREATE) 1322 cache_enter(dvp, *vpp, cnp); 1323 /* 1324 * Insert name into cache if appropriate. 1325 */ 1326 if (error == 0 && (cnp->cn_flags & MAKEENTRY)) { 1327 if (!(cnp->cn_flags & ISLASTCN) || 1328 (nameiop != DELETE && nameiop != RENAME)) { 1329 cache_enter(dvp, *vpp, cnp); 1330 } 1331 } 1332#endif 1333 1334 return (error); 1335} 1336 1337/* 1338 * Attempt to create a new entry in a directory. If the entry 1339 * already exists, truncate the file if permissible, else return 1340 * an error. Return the vp of the created or trunc'd file. 1341 * 1342 * IN: dvp - vnode of directory to put new file entry in. 1343 * name - name of new file entry. 1344 * vap - attributes of new file. 1345 * excl - flag indicating exclusive or non-exclusive mode. 1346 * mode - mode to open file with. 1347 * cr - credentials of caller. 1348 * flag - large file flag [UNUSED]. 1349 * ct - caller context 1350 * vsecp - ACL to be set 1351 * 1352 * OUT: vpp - vnode of created or trunc'd entry. 1353 * 1354 * RETURN: 0 if success 1355 * error code if failure 1356 * 1357 * Timestamps: 1358 * dvp - ctime|mtime updated if new entry created 1359 * vp - ctime|mtime always, atime if new 1360 */ 1361 1362/* ARGSUSED */ 1363static int 1364zfs_create(vnode_t *dvp, char *name, vattr_t *vap, int excl, int mode, 1365 vnode_t **vpp, cred_t *cr, kthread_t *td) 1366{ 1367 znode_t *zp, *dzp = VTOZ(dvp); 1368 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 1369 zilog_t *zilog; 1370 objset_t *os; 1371 zfs_dirlock_t *dl; 1372 dmu_tx_t *tx; 1373 int error; 1374 ksid_t *ksid; 1375 uid_t uid; 1376 gid_t gid = crgetgid(cr); 1377 zfs_acl_ids_t acl_ids; 1378 boolean_t fuid_dirtied; 1379 void *vsecp = NULL; 1380 int flag = 0; 1381 1382 /* 1383 * If we have an ephemeral id, ACL, or XVATTR then 1384 * make sure file system is at proper version 1385 */ 1386 1387 ksid = crgetsid(cr, KSID_OWNER); 1388 if (ksid) 1389 uid = ksid_getid(ksid); 1390 else 1391 uid = crgetuid(cr); 1392 if (zfsvfs->z_use_fuids == B_FALSE && 1393 (vsecp || (vap->va_mask & AT_XVATTR) || 1394 IS_EPHEMERAL(crgetuid(cr)) || IS_EPHEMERAL(crgetgid(cr)))) 1395 return (EINVAL); 1396 1397 ZFS_ENTER(zfsvfs); 1398 ZFS_VERIFY_ZP(dzp); 1399 os = zfsvfs->z_os; 1400 zilog = zfsvfs->z_log; 1401 1402 if (zfsvfs->z_utf8 && u8_validate(name, strlen(name), 1403 NULL, U8_VALIDATE_ENTIRE, &error) < 0) { 1404 ZFS_EXIT(zfsvfs); 1405 return (EILSEQ); 1406 } 1407 1408 if (vap->va_mask & AT_XVATTR) { 1409 if ((error = secpolicy_xvattr(dvp, (xvattr_t *)vap, 1410 crgetuid(cr), cr, vap->va_type)) != 0) { 1411 ZFS_EXIT(zfsvfs); 1412 return (error); 1413 } 1414 } 1415top: 1416 *vpp = NULL; 1417 1418 if ((vap->va_mode & S_ISVTX) && secpolicy_vnode_stky_modify(cr)) 1419 vap->va_mode &= ~S_ISVTX; 1420 1421 if (*name == '\0') { 1422 /* 1423 * Null component name refers to the directory itself. 1424 */ 1425 VN_HOLD(dvp); 1426 zp = dzp; 1427 dl = NULL; 1428 error = 0; 1429 } else { 1430 /* possible VN_HOLD(zp) */ 1431 int zflg = 0; 1432 1433 if (flag & FIGNORECASE) 1434 zflg |= ZCILOOK; 1435 1436 error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, 1437 NULL, NULL); 1438 if (error) { 1439 if (strcmp(name, "..") == 0) 1440 error = EISDIR; 1441 ZFS_EXIT(zfsvfs); 1442 return (error); 1443 } 1444 } 1445 if (zp == NULL) { 1446 uint64_t txtype; 1447 1448 /* 1449 * Create a new file object and update the directory 1450 * to reference it. 1451 */ 1452 if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) { 1453 goto out; 1454 } 1455 1456 /* 1457 * We only support the creation of regular files in 1458 * extended attribute directories. 1459 */ 1460 if ((dzp->z_phys->zp_flags & ZFS_XATTR) && 1461 (vap->va_type != VREG)) { 1462 error = EINVAL; 1463 goto out; 1464 } 1465 1466 1467 if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, vsecp, 1468 &acl_ids)) != 0) 1469 goto out; 1470 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) { 1471 error = EDQUOT; 1472 goto out; 1473 } 1474 1475 tx = dmu_tx_create(os); 1476 dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT); 1477 fuid_dirtied = zfsvfs->z_fuid_dirty; 1478 if (fuid_dirtied) 1479 zfs_fuid_txhold(zfsvfs, tx); 1480 dmu_tx_hold_bonus(tx, dzp->z_id); 1481 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); 1482 if (acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) { 1483 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 1484 0, SPA_MAXBLOCKSIZE); 1485 } 1486 error = dmu_tx_assign(tx, TXG_NOWAIT); 1487 if (error) { 1488 zfs_acl_ids_free(&acl_ids); 1489 zfs_dirent_unlock(dl); 1490 if (error == ERESTART) { 1491 dmu_tx_wait(tx); 1492 dmu_tx_abort(tx); 1493 goto top; 1494 } 1495 dmu_tx_abort(tx); 1496 ZFS_EXIT(zfsvfs); 1497 return (error); 1498 } 1499 zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, &acl_ids); 1500 1501 if (fuid_dirtied) 1502 zfs_fuid_sync(zfsvfs, tx); 1503 1504 (void) zfs_link_create(dl, zp, tx, ZNEW); 1505 1506 txtype = zfs_log_create_txtype(Z_FILE, vsecp, vap); 1507 if (flag & FIGNORECASE) 1508 txtype |= TX_CI; 1509 zfs_log_create(zilog, tx, txtype, dzp, zp, name, 1510 vsecp, acl_ids.z_fuidp, vap); 1511 zfs_acl_ids_free(&acl_ids); 1512 dmu_tx_commit(tx); 1513 } else { 1514 int aflags = (flag & FAPPEND) ? V_APPEND : 0; 1515 1516 /* 1517 * A directory entry already exists for this name. 1518 */ 1519 /* 1520 * Can't truncate an existing file if in exclusive mode. 1521 */ 1522 if (excl == EXCL) { 1523 error = EEXIST; 1524 goto out; 1525 } 1526 /* 1527 * Can't open a directory for writing. 1528 */ 1529 if ((ZTOV(zp)->v_type == VDIR) && (mode & S_IWRITE)) { 1530 error = EISDIR; 1531 goto out; 1532 } 1533 /* 1534 * Verify requested access to file. 1535 */ 1536 if (mode && (error = zfs_zaccess_rwx(zp, mode, aflags, cr))) { 1537 goto out; 1538 } 1539 1540 mutex_enter(&dzp->z_lock); 1541 dzp->z_seq++; 1542 mutex_exit(&dzp->z_lock); 1543 1544 /* 1545 * Truncate regular files if requested. 1546 */ 1547 if ((ZTOV(zp)->v_type == VREG) && 1548 (vap->va_mask & AT_SIZE) && (vap->va_size == 0)) { 1549 /* we can't hold any locks when calling zfs_freesp() */ 1550 zfs_dirent_unlock(dl); 1551 dl = NULL; 1552 error = zfs_freesp(zp, 0, 0, mode, TRUE); 1553 if (error == 0) { 1554 vnevent_create(ZTOV(zp), ct); 1555 } 1556 } 1557 } 1558out: 1559 if (dl) 1560 zfs_dirent_unlock(dl); 1561 1562 if (error) { 1563 if (zp) 1564 VN_RELE(ZTOV(zp)); 1565 } else { 1566 *vpp = ZTOV(zp); 1567 /* 1568 * If vnode is for a device return a specfs vnode instead. 1569 */ 1570 if (IS_DEVVP(*vpp)) { 1571 struct vnode *svp; 1572 1573 svp = specvp(*vpp, (*vpp)->v_rdev, (*vpp)->v_type, cr); 1574 VN_RELE(*vpp); 1575 if (svp == NULL) { 1576 error = ENOSYS; 1577 } 1578 *vpp = svp; 1579 } 1580 } 1581 1582 ZFS_EXIT(zfsvfs); 1583 return (error); 1584} 1585 1586/* 1587 * Remove an entry from a directory. 1588 * 1589 * IN: dvp - vnode of directory to remove entry from. 1590 * name - name of entry to remove. 1591 * cr - credentials of caller. 1592 * ct - caller context 1593 * flags - case flags 1594 * 1595 * RETURN: 0 if success 1596 * error code if failure 1597 * 1598 * Timestamps: 1599 * dvp - ctime|mtime 1600 * vp - ctime (if nlink > 0) 1601 */ 1602/*ARGSUSED*/ 1603static int 1604zfs_remove(vnode_t *dvp, char *name, cred_t *cr, caller_context_t *ct, 1605 int flags) 1606{ 1607 znode_t *zp, *dzp = VTOZ(dvp); 1608 znode_t *xzp = NULL; 1609 vnode_t *vp; 1610 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 1611 zilog_t *zilog; 1612 uint64_t acl_obj, xattr_obj; 1613 zfs_dirlock_t *dl; 1614 dmu_tx_t *tx; 1615 boolean_t may_delete_now, delete_now = FALSE; 1616 boolean_t unlinked, toobig = FALSE; 1617 uint64_t txtype; 1618 pathname_t *realnmp = NULL; 1619 pathname_t realnm; 1620 int error; 1621 int zflg = ZEXISTS; 1622 1623 ZFS_ENTER(zfsvfs); 1624 ZFS_VERIFY_ZP(dzp); 1625 zilog = zfsvfs->z_log; 1626 1627 if (flags & FIGNORECASE) { 1628 zflg |= ZCILOOK; 1629 pn_alloc(&realnm); 1630 realnmp = &realnm; 1631 } 1632 1633top: 1634 /* 1635 * Attempt to lock directory; fail if entry doesn't exist. 1636 */ 1637 if (error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, 1638 NULL, realnmp)) { 1639 if (realnmp) 1640 pn_free(realnmp); 1641 ZFS_EXIT(zfsvfs); 1642 return (error); 1643 } 1644 1645 vp = ZTOV(zp); 1646 1647 if (error = zfs_zaccess_delete(dzp, zp, cr)) { 1648 goto out; 1649 } 1650 1651 /* 1652 * Need to use rmdir for removing directories. 1653 */ 1654 if (vp->v_type == VDIR) { 1655 error = EPERM; 1656 goto out; 1657 } 1658 1659 vnevent_remove(vp, dvp, name, ct); 1660 1661 if (realnmp) 1662 dnlc_remove(dvp, realnmp->pn_buf); 1663 else 1664 dnlc_remove(dvp, name); 1665 1666 may_delete_now = FALSE; 1667 1668 /* 1669 * We may delete the znode now, or we may put it in the unlinked set; 1670 * it depends on whether we're the last link, and on whether there are 1671 * other holds on the vnode. So we dmu_tx_hold() the right things to 1672 * allow for either case. 1673 */ 1674 tx = dmu_tx_create(zfsvfs->z_os); 1675 dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); 1676 dmu_tx_hold_bonus(tx, zp->z_id); 1677 if (may_delete_now) { 1678 toobig = 1679 zp->z_phys->zp_size > zp->z_blksz * DMU_MAX_DELETEBLKCNT; 1680 /* if the file is too big, only hold_free a token amount */ 1681 dmu_tx_hold_free(tx, zp->z_id, 0, 1682 (toobig ? DMU_MAX_ACCESS : DMU_OBJECT_END)); 1683 } 1684 1685 /* are there any extended attributes? */ 1686 if ((xattr_obj = zp->z_phys->zp_xattr) != 0) { 1687 /* XXX - do we need this if we are deleting? */ 1688 dmu_tx_hold_bonus(tx, xattr_obj); 1689 } 1690 1691 /* are there any additional acls */ 1692 if ((acl_obj = zp->z_phys->zp_acl.z_acl_extern_obj) != 0 && 1693 may_delete_now) 1694 dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END); 1695 1696 /* charge as an update -- would be nice not to charge at all */ 1697 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); 1698 1699 error = dmu_tx_assign(tx, TXG_NOWAIT); 1700 if (error) { 1701 zfs_dirent_unlock(dl); 1702 VN_RELE(vp); 1703 if (error == ERESTART) { 1704 dmu_tx_wait(tx); 1705 dmu_tx_abort(tx); 1706 goto top; 1707 } 1708 if (realnmp) 1709 pn_free(realnmp); 1710 dmu_tx_abort(tx); 1711 ZFS_EXIT(zfsvfs); 1712 return (error); 1713 } 1714 1715 /* 1716 * Remove the directory entry. 1717 */ 1718 error = zfs_link_destroy(dl, zp, tx, zflg, &unlinked); 1719 1720 if (error) { 1721 dmu_tx_commit(tx); 1722 goto out; 1723 } 1724 1725 if (0 && unlinked) { 1726 VI_LOCK(vp); 1727 delete_now = may_delete_now && !toobig && 1728 vp->v_count == 1 && !vn_has_cached_data(vp) && 1729 zp->z_phys->zp_xattr == xattr_obj && 1730 zp->z_phys->zp_acl.z_acl_extern_obj == acl_obj; 1731 VI_UNLOCK(vp); 1732 } 1733 1734 if (delete_now) { 1735 if (zp->z_phys->zp_xattr) { 1736 error = zfs_zget(zfsvfs, zp->z_phys->zp_xattr, &xzp); 1737 ASSERT3U(error, ==, 0); 1738 ASSERT3U(xzp->z_phys->zp_links, ==, 2); 1739 dmu_buf_will_dirty(xzp->z_dbuf, tx); 1740 mutex_enter(&xzp->z_lock); 1741 xzp->z_unlinked = 1; 1742 xzp->z_phys->zp_links = 0; 1743 mutex_exit(&xzp->z_lock); 1744 zfs_unlinked_add(xzp, tx); 1745 zp->z_phys->zp_xattr = 0; /* probably unnecessary */ 1746 } 1747 mutex_enter(&zp->z_lock); 1748 VI_LOCK(vp); 1749 vp->v_count--; 1750 ASSERT3U(vp->v_count, ==, 0); 1751 VI_UNLOCK(vp); 1752 mutex_exit(&zp->z_lock); 1753 zfs_znode_delete(zp, tx); 1754 } else if (unlinked) { 1755 zfs_unlinked_add(zp, tx); 1756 } 1757 1758 txtype = TX_REMOVE; 1759 if (flags & FIGNORECASE) 1760 txtype |= TX_CI; 1761 zfs_log_remove(zilog, tx, txtype, dzp, name); 1762 1763 dmu_tx_commit(tx); 1764out: 1765 if (realnmp) 1766 pn_free(realnmp); 1767 1768 zfs_dirent_unlock(dl); 1769 1770 if (!delete_now) { 1771 VN_RELE(vp); 1772 } else if (xzp) { 1773 /* this rele is delayed to prevent nesting transactions */ 1774 VN_RELE(ZTOV(xzp)); 1775 } 1776 1777 ZFS_EXIT(zfsvfs); 1778 return (error); 1779} 1780 1781/* 1782 * Create a new directory and insert it into dvp using the name 1783 * provided. Return a pointer to the inserted directory. 1784 * 1785 * IN: dvp - vnode of directory to add subdir to. 1786 * dirname - name of new directory. 1787 * vap - attributes of new directory. 1788 * cr - credentials of caller. 1789 * ct - caller context 1790 * vsecp - ACL to be set 1791 * 1792 * OUT: vpp - vnode of created directory. 1793 * 1794 * RETURN: 0 if success 1795 * error code if failure 1796 * 1797 * Timestamps: 1798 * dvp - ctime|mtime updated 1799 * vp - ctime|mtime|atime updated 1800 */ 1801/*ARGSUSED*/ 1802static int 1803zfs_mkdir(vnode_t *dvp, char *dirname, vattr_t *vap, vnode_t **vpp, cred_t *cr, 1804 caller_context_t *ct, int flags, vsecattr_t *vsecp) 1805{ 1806 znode_t *zp, *dzp = VTOZ(dvp); 1807 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 1808 zilog_t *zilog; 1809 zfs_dirlock_t *dl; 1810 uint64_t txtype; 1811 dmu_tx_t *tx; 1812 int error; 1813 int zf = ZNEW; 1814 ksid_t *ksid; 1815 uid_t uid; 1816 gid_t gid = crgetgid(cr); 1817 zfs_acl_ids_t acl_ids; 1818 boolean_t fuid_dirtied; 1819 1820 ASSERT(vap->va_type == VDIR); 1821 1822 /* 1823 * If we have an ephemeral id, ACL, or XVATTR then 1824 * make sure file system is at proper version 1825 */ 1826 1827 ksid = crgetsid(cr, KSID_OWNER); 1828 if (ksid) 1829 uid = ksid_getid(ksid); 1830 else 1831 uid = crgetuid(cr); 1832 if (zfsvfs->z_use_fuids == B_FALSE && 1833 (vsecp || (vap->va_mask & AT_XVATTR) || IS_EPHEMERAL(crgetuid(cr))|| 1834 IS_EPHEMERAL(crgetgid(cr)))) 1835 return (EINVAL); 1836 1837 ZFS_ENTER(zfsvfs); 1838 ZFS_VERIFY_ZP(dzp); 1839 zilog = zfsvfs->z_log; 1840 1841 if (dzp->z_phys->zp_flags & ZFS_XATTR) { 1842 ZFS_EXIT(zfsvfs); 1843 return (EINVAL); 1844 } 1845 1846 if (zfsvfs->z_utf8 && u8_validate(dirname, 1847 strlen(dirname), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { 1848 ZFS_EXIT(zfsvfs); 1849 return (EILSEQ); 1850 } 1851 if (flags & FIGNORECASE) 1852 zf |= ZCILOOK; 1853 1854 if (vap->va_mask & AT_XVATTR) 1855 if ((error = secpolicy_xvattr(dvp, (xvattr_t *)vap, 1856 crgetuid(cr), cr, vap->va_type)) != 0) { 1857 ZFS_EXIT(zfsvfs); 1858 return (error); 1859 } 1860 1861 /* 1862 * First make sure the new directory doesn't exist. 1863 */ 1864top: 1865 *vpp = NULL; 1866 1867 if (error = zfs_dirent_lock(&dl, dzp, dirname, &zp, zf, 1868 NULL, NULL)) { 1869 ZFS_EXIT(zfsvfs); 1870 return (error); 1871 } 1872 1873 if (error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, 0, B_FALSE, cr)) { 1874 zfs_dirent_unlock(dl); 1875 ZFS_EXIT(zfsvfs); 1876 return (error); 1877 } 1878 1879 if ((error = zfs_acl_ids_create(dzp, 0, vap, cr, vsecp, 1880 &acl_ids)) != 0) { 1881 zfs_dirent_unlock(dl); 1882 ZFS_EXIT(zfsvfs); 1883 return (error); 1884 } 1885 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) { 1886 zfs_dirent_unlock(dl); 1887 ZFS_EXIT(zfsvfs); 1888 return (EDQUOT); 1889 } 1890 1891 /* 1892 * Add a new entry to the directory. 1893 */ 1894 tx = dmu_tx_create(zfsvfs->z_os); 1895 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname); 1896 dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); 1897 fuid_dirtied = zfsvfs->z_fuid_dirty; 1898 if (fuid_dirtied) 1899 zfs_fuid_txhold(zfsvfs, tx); 1900 if (acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) 1901 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 1902 0, SPA_MAXBLOCKSIZE); 1903 error = dmu_tx_assign(tx, TXG_NOWAIT); 1904 if (error) { 1905 zfs_acl_ids_free(&acl_ids); 1906 zfs_dirent_unlock(dl); 1907 if (error == ERESTART) { 1908 dmu_tx_wait(tx); 1909 dmu_tx_abort(tx); 1910 goto top; 1911 } 1912 dmu_tx_abort(tx); 1913 ZFS_EXIT(zfsvfs); 1914 return (error); 1915 } 1916 1917 /* 1918 * Create new node. 1919 */ 1920 zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, &acl_ids); 1921 1922 if (fuid_dirtied) 1923 zfs_fuid_sync(zfsvfs, tx); 1924 /* 1925 * Now put new name in parent dir. 1926 */ 1927 (void) zfs_link_create(dl, zp, tx, ZNEW); 1928 1929 *vpp = ZTOV(zp); 1930 1931 txtype = zfs_log_create_txtype(Z_DIR, vsecp, vap); 1932 if (flags & FIGNORECASE) 1933 txtype |= TX_CI; 1934 zfs_log_create(zilog, tx, txtype, dzp, zp, dirname, vsecp, 1935 acl_ids.z_fuidp, vap); 1936 1937 zfs_acl_ids_free(&acl_ids); 1938 dmu_tx_commit(tx); 1939 1940 zfs_dirent_unlock(dl); 1941 1942 ZFS_EXIT(zfsvfs); 1943 return (0); 1944} 1945 1946/* 1947 * Remove a directory subdir entry. If the current working 1948 * directory is the same as the subdir to be removed, the 1949 * remove will fail. 1950 * 1951 * IN: dvp - vnode of directory to remove from. 1952 * name - name of directory to be removed. 1953 * cwd - vnode of current working directory. 1954 * cr - credentials of caller. 1955 * ct - caller context 1956 * flags - case flags 1957 * 1958 * RETURN: 0 if success 1959 * error code if failure 1960 * 1961 * Timestamps: 1962 * dvp - ctime|mtime updated 1963 */ 1964/*ARGSUSED*/ 1965static int 1966zfs_rmdir(vnode_t *dvp, char *name, vnode_t *cwd, cred_t *cr, 1967 caller_context_t *ct, int flags) 1968{ 1969 znode_t *dzp = VTOZ(dvp); 1970 znode_t *zp; 1971 vnode_t *vp; 1972 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 1973 zilog_t *zilog; 1974 zfs_dirlock_t *dl; 1975 dmu_tx_t *tx; 1976 int error; 1977 int zflg = ZEXISTS; 1978 1979 ZFS_ENTER(zfsvfs); 1980 ZFS_VERIFY_ZP(dzp); 1981 zilog = zfsvfs->z_log; 1982 1983 if (flags & FIGNORECASE) 1984 zflg |= ZCILOOK; 1985top: 1986 zp = NULL; 1987 1988 /* 1989 * Attempt to lock directory; fail if entry doesn't exist. 1990 */ 1991 if (error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, 1992 NULL, NULL)) { 1993 ZFS_EXIT(zfsvfs); 1994 return (error); 1995 } 1996 1997 vp = ZTOV(zp); 1998 1999 if (error = zfs_zaccess_delete(dzp, zp, cr)) { 2000 goto out; 2001 } 2002 2003 if (vp->v_type != VDIR) { 2004 error = ENOTDIR; 2005 goto out; 2006 } 2007 2008 if (vp == cwd) { 2009 error = EINVAL; 2010 goto out; 2011 } 2012 2013 vnevent_rmdir(vp, dvp, name, ct); 2014 2015 /* 2016 * Grab a lock on the directory to make sure that noone is 2017 * trying to add (or lookup) entries while we are removing it. 2018 */ 2019 rw_enter(&zp->z_name_lock, RW_WRITER); 2020 2021 /* 2022 * Grab a lock on the parent pointer to make sure we play well 2023 * with the treewalk and directory rename code. 2024 */ 2025 rw_enter(&zp->z_parent_lock, RW_WRITER); 2026 2027 tx = dmu_tx_create(zfsvfs->z_os); 2028 dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); 2029 dmu_tx_hold_bonus(tx, zp->z_id); 2030 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); 2031 error = dmu_tx_assign(tx, TXG_NOWAIT); 2032 if (error) { 2033 rw_exit(&zp->z_parent_lock); 2034 rw_exit(&zp->z_name_lock); 2035 zfs_dirent_unlock(dl); 2036 VN_RELE(vp); 2037 if (error == ERESTART) { 2038 dmu_tx_wait(tx); 2039 dmu_tx_abort(tx); 2040 goto top; 2041 } 2042 dmu_tx_abort(tx); 2043 ZFS_EXIT(zfsvfs); 2044 return (error); 2045 } 2046 2047#ifdef FREEBSD_NAMECACHE 2048 cache_purge(dvp); 2049#endif 2050 2051 error = zfs_link_destroy(dl, zp, tx, zflg, NULL); 2052 2053 if (error == 0) { 2054 uint64_t txtype = TX_RMDIR; 2055 if (flags & FIGNORECASE) 2056 txtype |= TX_CI; 2057 zfs_log_remove(zilog, tx, txtype, dzp, name); 2058 } 2059 2060 dmu_tx_commit(tx); 2061 2062 rw_exit(&zp->z_parent_lock); 2063 rw_exit(&zp->z_name_lock); 2064#ifdef FREEBSD_NAMECACHE 2065 cache_purge(vp); 2066#endif 2067out: 2068 zfs_dirent_unlock(dl); 2069 2070 VN_RELE(vp); 2071 2072 ZFS_EXIT(zfsvfs); 2073 return (error); 2074} 2075 2076/* 2077 * Read as many directory entries as will fit into the provided 2078 * buffer from the given directory cursor position (specified in 2079 * the uio structure. 2080 * 2081 * IN: vp - vnode of directory to read. 2082 * uio - structure supplying read location, range info, 2083 * and return buffer. 2084 * cr - credentials of caller. 2085 * ct - caller context 2086 * flags - case flags 2087 * 2088 * OUT: uio - updated offset and range, buffer filled. 2089 * eofp - set to true if end-of-file detected. 2090 * 2091 * RETURN: 0 if success 2092 * error code if failure 2093 * 2094 * Timestamps: 2095 * vp - atime updated 2096 * 2097 * Note that the low 4 bits of the cookie returned by zap is always zero. 2098 * This allows us to use the low range for "special" directory entries: 2099 * We use 0 for '.', and 1 for '..'. If this is the root of the filesystem, 2100 * we use the offset 2 for the '.zfs' directory. 2101 */ 2102/* ARGSUSED */ 2103static int 2104zfs_readdir(vnode_t *vp, uio_t *uio, cred_t *cr, int *eofp, int *ncookies, u_long **cookies) 2105{ 2106 znode_t *zp = VTOZ(vp); 2107 iovec_t *iovp; 2108 edirent_t *eodp; 2109 dirent64_t *odp; 2110 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 2111 objset_t *os; 2112 caddr_t outbuf; 2113 size_t bufsize; 2114 zap_cursor_t zc; 2115 zap_attribute_t zap; 2116 uint_t bytes_wanted; 2117 uint64_t offset; /* must be unsigned; checks for < 1 */ 2118 int local_eof; 2119 int outcount; 2120 int error; 2121 uint8_t prefetch; 2122 boolean_t check_sysattrs; 2123 uint8_t type; 2124 int ncooks; 2125 u_long *cooks = NULL; 2126 int flags = 0; 2127 2128 ZFS_ENTER(zfsvfs); 2129 ZFS_VERIFY_ZP(zp); 2130 2131 /* 2132 * If we are not given an eof variable, 2133 * use a local one. 2134 */ 2135 if (eofp == NULL) 2136 eofp = &local_eof; 2137 2138 /* 2139 * Check for valid iov_len. 2140 */ 2141 if (uio->uio_iov->iov_len <= 0) { 2142 ZFS_EXIT(zfsvfs); 2143 return (EINVAL); 2144 } 2145 2146 /* 2147 * Quit if directory has been removed (posix) 2148 */ 2149 if ((*eofp = zp->z_unlinked) != 0) { 2150 ZFS_EXIT(zfsvfs); 2151 return (0); 2152 } 2153 2154 error = 0; 2155 os = zfsvfs->z_os; 2156 offset = uio->uio_loffset; 2157 prefetch = zp->z_zn_prefetch; 2158 2159 /* 2160 * Initialize the iterator cursor. 2161 */ 2162 if (offset <= 3) { 2163 /* 2164 * Start iteration from the beginning of the directory. 2165 */ 2166 zap_cursor_init(&zc, os, zp->z_id); 2167 } else { 2168 /* 2169 * The offset is a serialized cursor. 2170 */ 2171 zap_cursor_init_serialized(&zc, os, zp->z_id, offset); 2172 } 2173 2174 /* 2175 * Get space to change directory entries into fs independent format. 2176 */ 2177 iovp = uio->uio_iov; 2178 bytes_wanted = iovp->iov_len; 2179 if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1) { 2180 bufsize = bytes_wanted; 2181 outbuf = kmem_alloc(bufsize, KM_SLEEP); 2182 odp = (struct dirent64 *)outbuf; 2183 } else { 2184 bufsize = bytes_wanted; 2185 odp = (struct dirent64 *)iovp->iov_base; 2186 } 2187 eodp = (struct edirent *)odp; 2188 2189 if (ncookies != NULL) { 2190 /* 2191 * Minimum entry size is dirent size and 1 byte for a file name. 2192 */ 2193 ncooks = uio->uio_resid / (sizeof(struct dirent) - sizeof(((struct dirent *)NULL)->d_name) + 1); 2194 cooks = malloc(ncooks * sizeof(u_long), M_TEMP, M_WAITOK); 2195 *cookies = cooks; 2196 *ncookies = ncooks; 2197 } 2198 /* 2199 * If this VFS supports the system attribute view interface; and 2200 * we're looking at an extended attribute directory; and we care 2201 * about normalization conflicts on this vfs; then we must check 2202 * for normalization conflicts with the sysattr name space. 2203 */ 2204#ifdef TODO 2205 check_sysattrs = vfs_has_feature(vp->v_vfsp, VFSFT_SYSATTR_VIEWS) && 2206 (vp->v_flag & V_XATTRDIR) && zfsvfs->z_norm && 2207 (flags & V_RDDIR_ENTFLAGS); 2208#else 2209 check_sysattrs = 0; 2210#endif 2211 2212 /* 2213 * Transform to file-system independent format 2214 */ 2215 outcount = 0; 2216 while (outcount < bytes_wanted) { 2217 ino64_t objnum; 2218 ushort_t reclen; 2219 off64_t *next; 2220 2221 /* 2222 * Special case `.', `..', and `.zfs'. 2223 */ 2224 if (offset == 0) { 2225 (void) strcpy(zap.za_name, "."); 2226 zap.za_normalization_conflict = 0; 2227 objnum = zp->z_id; 2228 type = DT_DIR; 2229 } else if (offset == 1) { 2230 (void) strcpy(zap.za_name, ".."); 2231 zap.za_normalization_conflict = 0; 2232 objnum = zp->z_phys->zp_parent; 2233 type = DT_DIR; 2234 } else if (offset == 2 && zfs_show_ctldir(zp)) { 2235 (void) strcpy(zap.za_name, ZFS_CTLDIR_NAME); 2236 zap.za_normalization_conflict = 0; 2237 objnum = ZFSCTL_INO_ROOT; 2238 type = DT_DIR; 2239 } else { 2240 /* 2241 * Grab next entry. 2242 */ 2243 if (error = zap_cursor_retrieve(&zc, &zap)) { 2244 if ((*eofp = (error == ENOENT)) != 0) 2245 break; 2246 else 2247 goto update; 2248 } 2249 2250 if (zap.za_integer_length != 8 || 2251 zap.za_num_integers != 1) { 2252 cmn_err(CE_WARN, "zap_readdir: bad directory " 2253 "entry, obj = %lld, offset = %lld\n", 2254 (u_longlong_t)zp->z_id, 2255 (u_longlong_t)offset); 2256 error = ENXIO; 2257 goto update; 2258 } 2259 2260 objnum = ZFS_DIRENT_OBJ(zap.za_first_integer); 2261 /* 2262 * MacOS X can extract the object type here such as: 2263 * uint8_t type = ZFS_DIRENT_TYPE(zap.za_first_integer); 2264 */ 2265 type = ZFS_DIRENT_TYPE(zap.za_first_integer); 2266 2267 if (check_sysattrs && !zap.za_normalization_conflict) { 2268#ifdef TODO 2269 zap.za_normalization_conflict = 2270 xattr_sysattr_casechk(zap.za_name); 2271#else 2272 panic("%s:%u: TODO", __func__, __LINE__); 2273#endif 2274 } 2275 } 2276 2277 if (flags & V_RDDIR_ENTFLAGS) 2278 reclen = EDIRENT_RECLEN(strlen(zap.za_name)); 2279 else 2280 reclen = DIRENT64_RECLEN(strlen(zap.za_name)); 2281 2282 /* 2283 * Will this entry fit in the buffer? 2284 */ 2285 if (outcount + reclen > bufsize) { 2286 /* 2287 * Did we manage to fit anything in the buffer? 2288 */ 2289 if (!outcount) { 2290 error = EINVAL; 2291 goto update; 2292 } 2293 break; 2294 } 2295 if (flags & V_RDDIR_ENTFLAGS) { 2296 /* 2297 * Add extended flag entry: 2298 */ 2299 eodp->ed_ino = objnum; 2300 eodp->ed_reclen = reclen; 2301 /* NOTE: ed_off is the offset for the *next* entry */ 2302 next = &(eodp->ed_off); 2303 eodp->ed_eflags = zap.za_normalization_conflict ? 2304 ED_CASE_CONFLICT : 0; 2305 (void) strncpy(eodp->ed_name, zap.za_name, 2306 EDIRENT_NAMELEN(reclen)); 2307 eodp = (edirent_t *)((intptr_t)eodp + reclen); 2308 } else { 2309 /* 2310 * Add normal entry: 2311 */ 2312 odp->d_ino = objnum; 2313 odp->d_reclen = reclen; 2314 odp->d_namlen = strlen(zap.za_name); 2315 (void) strlcpy(odp->d_name, zap.za_name, odp->d_namlen + 1); 2316 odp->d_type = type; 2317 odp = (dirent64_t *)((intptr_t)odp + reclen); 2318 } 2319 outcount += reclen; 2320 2321 ASSERT(outcount <= bufsize); 2322 2323 /* Prefetch znode */ 2324 if (prefetch) 2325 dmu_prefetch(os, objnum, 0, 0); 2326 2327 /* 2328 * Move to the next entry, fill in the previous offset. 2329 */ 2330 if (offset > 2 || (offset == 2 && !zfs_show_ctldir(zp))) { 2331 zap_cursor_advance(&zc); 2332 offset = zap_cursor_serialize(&zc); 2333 } else { 2334 offset += 1; 2335 } 2336 2337 if (cooks != NULL) { 2338 *cooks++ = offset; 2339 ncooks--; 2340 KASSERT(ncooks >= 0, ("ncookies=%d", ncooks)); 2341 } 2342 } 2343 zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */ 2344 2345 /* Subtract unused cookies */ 2346 if (ncookies != NULL) 2347 *ncookies -= ncooks; 2348 2349 if (uio->uio_segflg == UIO_SYSSPACE && uio->uio_iovcnt == 1) { 2350 iovp->iov_base += outcount; 2351 iovp->iov_len -= outcount; 2352 uio->uio_resid -= outcount; 2353 } else if (error = uiomove(outbuf, (long)outcount, UIO_READ, uio)) { 2354 /* 2355 * Reset the pointer. 2356 */ 2357 offset = uio->uio_loffset; 2358 } 2359 2360update: 2361 zap_cursor_fini(&zc); 2362 if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1) 2363 kmem_free(outbuf, bufsize); 2364 2365 if (error == ENOENT) 2366 error = 0; 2367 2368 ZFS_ACCESSTIME_STAMP(zfsvfs, zp); 2369 2370 uio->uio_loffset = offset; 2371 ZFS_EXIT(zfsvfs); 2372 if (error != 0 && cookies != NULL) { 2373 free(*cookies, M_TEMP); 2374 *cookies = NULL; 2375 *ncookies = 0; 2376 } 2377 return (error); 2378} 2379 2380ulong_t zfs_fsync_sync_cnt = 4; 2381 2382static int 2383zfs_fsync(vnode_t *vp, int syncflag, cred_t *cr, caller_context_t *ct) 2384{ 2385 znode_t *zp = VTOZ(vp); 2386 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 2387 2388 (void) tsd_set(zfs_fsyncer_key, (void *)zfs_fsync_sync_cnt); 2389 2390 ZFS_ENTER(zfsvfs); 2391 ZFS_VERIFY_ZP(zp); 2392 zil_commit(zfsvfs->z_log, zp->z_last_itx, zp->z_id); 2393 ZFS_EXIT(zfsvfs); 2394 return (0); 2395} 2396 2397 2398/* 2399 * Get the requested file attributes and place them in the provided 2400 * vattr structure. 2401 * 2402 * IN: vp - vnode of file. 2403 * vap - va_mask identifies requested attributes. 2404 * If AT_XVATTR set, then optional attrs are requested 2405 * flags - ATTR_NOACLCHECK (CIFS server context) 2406 * cr - credentials of caller. 2407 * ct - caller context 2408 * 2409 * OUT: vap - attribute values. 2410 * 2411 * RETURN: 0 (always succeeds) 2412 */ 2413/* ARGSUSED */ 2414static int 2415zfs_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr, 2416 caller_context_t *ct) 2417{ 2418 znode_t *zp = VTOZ(vp); 2419 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 2420 znode_phys_t *pzp; 2421 int error = 0; 2422 uint32_t blksize; 2423 u_longlong_t nblocks; 2424 uint64_t links; 2425 xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */ 2426 xoptattr_t *xoap = NULL; 2427 boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; 2428 2429 ZFS_ENTER(zfsvfs); 2430 ZFS_VERIFY_ZP(zp); 2431 pzp = zp->z_phys; 2432 2433 /* 2434 * If ACL is trivial don't bother looking for ACE_READ_ATTRIBUTES. 2435 * Also, if we are the owner don't bother, since owner should 2436 * always be allowed to read basic attributes of file. 2437 */ 2438 if (!(pzp->zp_flags & ZFS_ACL_TRIVIAL) && 2439 (pzp->zp_uid != crgetuid(cr))) { 2440 if (error = zfs_zaccess(zp, ACE_READ_ATTRIBUTES, 0, 2441 skipaclchk, cr)) { 2442 ZFS_EXIT(zfsvfs); 2443 return (error); 2444 } 2445 } 2446 2447 /* 2448 * Return all attributes. It's cheaper to provide the answer 2449 * than to determine whether we were asked the question. 2450 */ 2451 2452 mutex_enter(&zp->z_lock); 2453 vap->va_type = IFTOVT(pzp->zp_mode); 2454 vap->va_mode = pzp->zp_mode & ~S_IFMT; 2455 zfs_fuid_map_ids(zp, cr, &vap->va_uid, &vap->va_gid); 2456// vap->va_fsid = zp->z_zfsvfs->z_vfs->vfs_dev; 2457 vap->va_nodeid = zp->z_id; 2458 if ((vp->v_flag & VROOT) && zfs_show_ctldir(zp)) 2459 links = pzp->zp_links + 1; 2460 else 2461 links = pzp->zp_links; 2462 vap->va_nlink = MIN(links, UINT32_MAX); /* nlink_t limit! */ 2463 vap->va_size = pzp->zp_size; 2464 vap->va_fsid = vp->v_mount->mnt_stat.f_fsid.val[0]; 2465 vap->va_rdev = zfs_cmpldev(pzp->zp_rdev); 2466 vap->va_seq = zp->z_seq; 2467 vap->va_flags = 0; /* FreeBSD: Reset chflags(2) flags. */ 2468 2469 /* 2470 * Add in any requested optional attributes and the create time. 2471 * Also set the corresponding bits in the returned attribute bitmap. 2472 */ 2473 if ((xoap = xva_getxoptattr(xvap)) != NULL && zfsvfs->z_use_fuids) { 2474 if (XVA_ISSET_REQ(xvap, XAT_ARCHIVE)) { 2475 xoap->xoa_archive = 2476 ((pzp->zp_flags & ZFS_ARCHIVE) != 0); 2477 XVA_SET_RTN(xvap, XAT_ARCHIVE); 2478 } 2479 2480 if (XVA_ISSET_REQ(xvap, XAT_READONLY)) { 2481 xoap->xoa_readonly = 2482 ((pzp->zp_flags & ZFS_READONLY) != 0); 2483 XVA_SET_RTN(xvap, XAT_READONLY); 2484 } 2485 2486 if (XVA_ISSET_REQ(xvap, XAT_SYSTEM)) { 2487 xoap->xoa_system = 2488 ((pzp->zp_flags & ZFS_SYSTEM) != 0); 2489 XVA_SET_RTN(xvap, XAT_SYSTEM); 2490 } 2491 2492 if (XVA_ISSET_REQ(xvap, XAT_HIDDEN)) { 2493 xoap->xoa_hidden = 2494 ((pzp->zp_flags & ZFS_HIDDEN) != 0); 2495 XVA_SET_RTN(xvap, XAT_HIDDEN); 2496 } 2497 2498 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { 2499 xoap->xoa_nounlink = 2500 ((pzp->zp_flags & ZFS_NOUNLINK) != 0); 2501 XVA_SET_RTN(xvap, XAT_NOUNLINK); 2502 } 2503 2504 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { 2505 xoap->xoa_immutable = 2506 ((pzp->zp_flags & ZFS_IMMUTABLE) != 0); 2507 XVA_SET_RTN(xvap, XAT_IMMUTABLE); 2508 } 2509 2510 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { 2511 xoap->xoa_appendonly = 2512 ((pzp->zp_flags & ZFS_APPENDONLY) != 0); 2513 XVA_SET_RTN(xvap, XAT_APPENDONLY); 2514 } 2515 2516 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { 2517 xoap->xoa_nodump = 2518 ((pzp->zp_flags & ZFS_NODUMP) != 0); 2519 XVA_SET_RTN(xvap, XAT_NODUMP); 2520 } 2521 2522 if (XVA_ISSET_REQ(xvap, XAT_OPAQUE)) { 2523 xoap->xoa_opaque = 2524 ((pzp->zp_flags & ZFS_OPAQUE) != 0); 2525 XVA_SET_RTN(xvap, XAT_OPAQUE); 2526 } 2527 2528 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { 2529 xoap->xoa_av_quarantined = 2530 ((pzp->zp_flags & ZFS_AV_QUARANTINED) != 0); 2531 XVA_SET_RTN(xvap, XAT_AV_QUARANTINED); 2532 } 2533 2534 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { 2535 xoap->xoa_av_modified = 2536 ((pzp->zp_flags & ZFS_AV_MODIFIED) != 0); 2537 XVA_SET_RTN(xvap, XAT_AV_MODIFIED); 2538 } 2539 2540 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) && 2541 vp->v_type == VREG && 2542 (pzp->zp_flags & ZFS_BONUS_SCANSTAMP)) { 2543 size_t len; 2544 dmu_object_info_t doi; 2545 2546 /* 2547 * Only VREG files have anti-virus scanstamps, so we 2548 * won't conflict with symlinks in the bonus buffer. 2549 */ 2550 dmu_object_info_from_db(zp->z_dbuf, &doi); 2551 len = sizeof (xoap->xoa_av_scanstamp) + 2552 sizeof (znode_phys_t); 2553 if (len <= doi.doi_bonus_size) { 2554 /* 2555 * pzp points to the start of the 2556 * znode_phys_t. pzp + 1 points to the 2557 * first byte after the znode_phys_t. 2558 */ 2559 (void) memcpy(xoap->xoa_av_scanstamp, 2560 pzp + 1, 2561 sizeof (xoap->xoa_av_scanstamp)); 2562 XVA_SET_RTN(xvap, XAT_AV_SCANSTAMP); 2563 } 2564 } 2565 2566 if (XVA_ISSET_REQ(xvap, XAT_CREATETIME)) { 2567 ZFS_TIME_DECODE(&xoap->xoa_createtime, pzp->zp_crtime); 2568 XVA_SET_RTN(xvap, XAT_CREATETIME); 2569 } 2570 } 2571 2572 ZFS_TIME_DECODE(&vap->va_atime, pzp->zp_atime); 2573 ZFS_TIME_DECODE(&vap->va_mtime, pzp->zp_mtime); 2574 ZFS_TIME_DECODE(&vap->va_ctime, pzp->zp_ctime); 2575 ZFS_TIME_DECODE(&vap->va_birthtime, pzp->zp_crtime); 2576 2577 mutex_exit(&zp->z_lock); 2578 2579 dmu_object_size_from_db(zp->z_dbuf, &blksize, &nblocks); 2580 vap->va_blksize = blksize; 2581 vap->va_bytes = nblocks << 9; /* nblocks * 512 */ 2582 2583 if (zp->z_blksz == 0) { 2584 /* 2585 * Block size hasn't been set; suggest maximal I/O transfers. 2586 */ 2587 vap->va_blksize = zfsvfs->z_max_blksz; 2588 } 2589 2590 ZFS_EXIT(zfsvfs); 2591 return (0); 2592} 2593 2594/* 2595 * Set the file attributes to the values contained in the 2596 * vattr structure. 2597 * 2598 * IN: vp - vnode of file to be modified. 2599 * vap - new attribute values. 2600 * If AT_XVATTR set, then optional attrs are being set 2601 * flags - ATTR_UTIME set if non-default time values provided. 2602 * - ATTR_NOACLCHECK (CIFS context only). 2603 * cr - credentials of caller. 2604 * ct - caller context 2605 * 2606 * RETURN: 0 if success 2607 * error code if failure 2608 * 2609 * Timestamps: 2610 * vp - ctime updated, mtime updated if size changed. 2611 */ 2612/* ARGSUSED */ 2613static int 2614zfs_setattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr, 2615 caller_context_t *ct) 2616{ 2617 znode_t *zp = VTOZ(vp); 2618 znode_phys_t *pzp; 2619 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 2620 zilog_t *zilog; 2621 dmu_tx_t *tx; 2622 vattr_t oldva; 2623 xvattr_t tmpxvattr; 2624 uint_t mask = vap->va_mask; 2625 uint_t saved_mask; 2626 uint64_t saved_mode; 2627 int trim_mask = 0; 2628 uint64_t new_mode; 2629 uint64_t new_uid, new_gid; 2630 znode_t *attrzp; 2631 int need_policy = FALSE; 2632 int err; 2633 zfs_fuid_info_t *fuidp = NULL; 2634 xvattr_t *xvap = (xvattr_t *)vap; /* vap may be an xvattr_t * */ 2635 xoptattr_t *xoap; 2636 zfs_acl_t *aclp = NULL; 2637 boolean_t skipaclchk = (flags & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; 2638 boolean_t fuid_dirtied = B_FALSE; 2639 2640 if (mask == 0) 2641 return (0); 2642 2643 if (mask & AT_NOSET) 2644 return (EINVAL); 2645 2646 ZFS_ENTER(zfsvfs); 2647 ZFS_VERIFY_ZP(zp); 2648 2649 pzp = zp->z_phys; 2650 zilog = zfsvfs->z_log; 2651 2652 /* 2653 * Make sure that if we have ephemeral uid/gid or xvattr specified 2654 * that file system is at proper version level 2655 */ 2656 2657 if (zfsvfs->z_use_fuids == B_FALSE && 2658 (((mask & AT_UID) && IS_EPHEMERAL(vap->va_uid)) || 2659 ((mask & AT_GID) && IS_EPHEMERAL(vap->va_gid)) || 2660 (mask & AT_XVATTR))) { 2661 ZFS_EXIT(zfsvfs); 2662 return (EINVAL); 2663 } 2664 2665 if (mask & AT_SIZE && vp->v_type == VDIR) { 2666 ZFS_EXIT(zfsvfs); 2667 return (EISDIR); 2668 } 2669 2670 if (mask & AT_SIZE && vp->v_type != VREG && vp->v_type != VFIFO) { 2671 ZFS_EXIT(zfsvfs); 2672 return (EINVAL); 2673 } 2674 2675 /* 2676 * If this is an xvattr_t, then get a pointer to the structure of 2677 * optional attributes. If this is NULL, then we have a vattr_t. 2678 */ 2679 xoap = xva_getxoptattr(xvap); 2680 2681 xva_init(&tmpxvattr); 2682 2683 /* 2684 * Immutable files can only alter immutable bit and atime 2685 */ 2686 if ((pzp->zp_flags & ZFS_IMMUTABLE) && 2687 ((mask & (AT_SIZE|AT_UID|AT_GID|AT_MTIME|AT_MODE)) || 2688 ((mask & AT_XVATTR) && XVA_ISSET_REQ(xvap, XAT_CREATETIME)))) { 2689 ZFS_EXIT(zfsvfs); 2690 return (EPERM); 2691 } 2692 2693 if ((mask & AT_SIZE) && (pzp->zp_flags & ZFS_READONLY)) { 2694 ZFS_EXIT(zfsvfs); 2695 return (EPERM); 2696 } 2697 2698 /* 2699 * Verify timestamps doesn't overflow 32 bits. 2700 * ZFS can handle large timestamps, but 32bit syscalls can't 2701 * handle times greater than 2039. This check should be removed 2702 * once large timestamps are fully supported. 2703 */ 2704 if (mask & (AT_ATIME | AT_MTIME)) { 2705 if (((mask & AT_ATIME) && TIMESPEC_OVERFLOW(&vap->va_atime)) || 2706 ((mask & AT_MTIME) && TIMESPEC_OVERFLOW(&vap->va_mtime))) { 2707 ZFS_EXIT(zfsvfs); 2708 return (EOVERFLOW); 2709 } 2710 } 2711 2712top: 2713 attrzp = NULL; 2714 2715 if (zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) { 2716 ZFS_EXIT(zfsvfs); 2717 return (EROFS); 2718 } 2719 2720 /* 2721 * First validate permissions 2722 */ 2723 2724 if (mask & AT_SIZE) { 2725 err = zfs_zaccess(zp, ACE_WRITE_DATA, 0, skipaclchk, cr); 2726 if (err) { 2727 ZFS_EXIT(zfsvfs); 2728 return (err); 2729 } 2730 /* 2731 * XXX - Note, we are not providing any open 2732 * mode flags here (like FNDELAY), so we may 2733 * block if there are locks present... this 2734 * should be addressed in openat(). 2735 */ 2736 /* XXX - would it be OK to generate a log record here? */ 2737 err = zfs_freesp(zp, vap->va_size, 0, 0, FALSE); 2738 if (err) { 2739 ZFS_EXIT(zfsvfs); 2740 return (err); 2741 } 2742 } 2743 2744 if (mask & (AT_ATIME|AT_MTIME) || 2745 ((mask & AT_XVATTR) && (XVA_ISSET_REQ(xvap, XAT_HIDDEN) || 2746 XVA_ISSET_REQ(xvap, XAT_READONLY) || 2747 XVA_ISSET_REQ(xvap, XAT_ARCHIVE) || 2748 XVA_ISSET_REQ(xvap, XAT_CREATETIME) || 2749 XVA_ISSET_REQ(xvap, XAT_SYSTEM)))) 2750 need_policy = zfs_zaccess(zp, ACE_WRITE_ATTRIBUTES, 0, 2751 skipaclchk, cr); 2752 2753 if (mask & (AT_UID|AT_GID)) { 2754 int idmask = (mask & (AT_UID|AT_GID)); 2755 int take_owner; 2756 int take_group; 2757 2758 /* 2759 * NOTE: even if a new mode is being set, 2760 * we may clear S_ISUID/S_ISGID bits. 2761 */ 2762 2763 if (!(mask & AT_MODE)) 2764 vap->va_mode = pzp->zp_mode; 2765 2766 /* 2767 * Take ownership or chgrp to group we are a member of 2768 */ 2769 2770 take_owner = (mask & AT_UID) && (vap->va_uid == crgetuid(cr)); 2771 take_group = (mask & AT_GID) && 2772 zfs_groupmember(zfsvfs, vap->va_gid, cr); 2773 2774 /* 2775 * If both AT_UID and AT_GID are set then take_owner and 2776 * take_group must both be set in order to allow taking 2777 * ownership. 2778 * 2779 * Otherwise, send the check through secpolicy_vnode_setattr() 2780 * 2781 */ 2782 2783 if (((idmask == (AT_UID|AT_GID)) && take_owner && take_group) || 2784 ((idmask == AT_UID) && take_owner) || 2785 ((idmask == AT_GID) && take_group)) { 2786 if (zfs_zaccess(zp, ACE_WRITE_OWNER, 0, 2787 skipaclchk, cr) == 0) { 2788 /* 2789 * Remove setuid/setgid for non-privileged users 2790 */ 2791 secpolicy_setid_clear(vap, vp, cr); 2792 trim_mask = (mask & (AT_UID|AT_GID)); 2793 } else { 2794 need_policy = TRUE; 2795 } 2796 } else { 2797 need_policy = TRUE; 2798 } 2799 } 2800 2801 mutex_enter(&zp->z_lock); 2802 oldva.va_mode = pzp->zp_mode; 2803 zfs_fuid_map_ids(zp, cr, &oldva.va_uid, &oldva.va_gid); 2804 if (mask & AT_XVATTR) { 2805 /* 2806 * Update xvattr mask to include only those attributes 2807 * that are actually changing. 2808 * 2809 * the bits will be restored prior to actually setting 2810 * the attributes so the caller thinks they were set. 2811 */ 2812 if (XVA_ISSET_REQ(xvap, XAT_APPENDONLY)) { 2813 if (xoap->xoa_appendonly != 2814 ((pzp->zp_flags & ZFS_APPENDONLY) != 0)) { 2815 need_policy = TRUE; 2816 } else { 2817 XVA_CLR_REQ(xvap, XAT_APPENDONLY); 2818 XVA_SET_REQ(&tmpxvattr, XAT_APPENDONLY); 2819 } 2820 } 2821 2822 if (XVA_ISSET_REQ(xvap, XAT_NOUNLINK)) { 2823 if (xoap->xoa_nounlink != 2824 ((pzp->zp_flags & ZFS_NOUNLINK) != 0)) { 2825 need_policy = TRUE; 2826 } else { 2827 XVA_CLR_REQ(xvap, XAT_NOUNLINK); 2828 XVA_SET_REQ(&tmpxvattr, XAT_NOUNLINK); 2829 } 2830 } 2831 2832 if (XVA_ISSET_REQ(xvap, XAT_IMMUTABLE)) { 2833 if (xoap->xoa_immutable != 2834 ((pzp->zp_flags & ZFS_IMMUTABLE) != 0)) { 2835 need_policy = TRUE; 2836 } else { 2837 XVA_CLR_REQ(xvap, XAT_IMMUTABLE); 2838 XVA_SET_REQ(&tmpxvattr, XAT_IMMUTABLE); 2839 } 2840 } 2841 2842 if (XVA_ISSET_REQ(xvap, XAT_NODUMP)) { 2843 if (xoap->xoa_nodump != 2844 ((pzp->zp_flags & ZFS_NODUMP) != 0)) { 2845 need_policy = TRUE; 2846 } else { 2847 XVA_CLR_REQ(xvap, XAT_NODUMP); 2848 XVA_SET_REQ(&tmpxvattr, XAT_NODUMP); 2849 } 2850 } 2851 2852 if (XVA_ISSET_REQ(xvap, XAT_AV_MODIFIED)) { 2853 if (xoap->xoa_av_modified != 2854 ((pzp->zp_flags & ZFS_AV_MODIFIED) != 0)) { 2855 need_policy = TRUE; 2856 } else { 2857 XVA_CLR_REQ(xvap, XAT_AV_MODIFIED); 2858 XVA_SET_REQ(&tmpxvattr, XAT_AV_MODIFIED); 2859 } 2860 } 2861 2862 if (XVA_ISSET_REQ(xvap, XAT_AV_QUARANTINED)) { 2863 if ((vp->v_type != VREG && 2864 xoap->xoa_av_quarantined) || 2865 xoap->xoa_av_quarantined != 2866 ((pzp->zp_flags & ZFS_AV_QUARANTINED) != 0)) { 2867 need_policy = TRUE; 2868 } else { 2869 XVA_CLR_REQ(xvap, XAT_AV_QUARANTINED); 2870 XVA_SET_REQ(&tmpxvattr, XAT_AV_QUARANTINED); 2871 } 2872 } 2873 2874 if (need_policy == FALSE && 2875 (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP) || 2876 XVA_ISSET_REQ(xvap, XAT_OPAQUE))) { 2877 need_policy = TRUE; 2878 } 2879 } 2880 2881 mutex_exit(&zp->z_lock); 2882 2883 if (mask & AT_MODE) { 2884 if (zfs_zaccess(zp, ACE_WRITE_ACL, 0, skipaclchk, cr) == 0) { 2885 err = secpolicy_setid_setsticky_clear(vp, vap, 2886 &oldva, cr); 2887 if (err) { 2888 ZFS_EXIT(zfsvfs); 2889 return (err); 2890 } 2891 trim_mask |= AT_MODE; 2892 } else { 2893 need_policy = TRUE; 2894 } 2895 } 2896 2897 if (need_policy) { 2898 /* 2899 * If trim_mask is set then take ownership 2900 * has been granted or write_acl is present and user 2901 * has the ability to modify mode. In that case remove 2902 * UID|GID and or MODE from mask so that 2903 * secpolicy_vnode_setattr() doesn't revoke it. 2904 */ 2905 2906 if (trim_mask) { 2907 saved_mask = vap->va_mask; 2908 vap->va_mask &= ~trim_mask; 2909 if (trim_mask & AT_MODE) { 2910 /* 2911 * Save the mode, as secpolicy_vnode_setattr() 2912 * will overwrite it with ova.va_mode. 2913 */ 2914 saved_mode = vap->va_mode; 2915 } 2916 } 2917 err = secpolicy_vnode_setattr(cr, vp, vap, &oldva, flags, 2918 (int (*)(void *, int, cred_t *))zfs_zaccess_unix, zp); 2919 if (err) { 2920 ZFS_EXIT(zfsvfs); 2921 return (err); 2922 } 2923 2924 if (trim_mask) { 2925 vap->va_mask |= saved_mask; 2926 if (trim_mask & AT_MODE) { 2927 /* 2928 * Recover the mode after 2929 * secpolicy_vnode_setattr(). 2930 */ 2931 vap->va_mode = saved_mode; 2932 } 2933 } 2934 } 2935 2936 /* 2937 * secpolicy_vnode_setattr, or take ownership may have 2938 * changed va_mask 2939 */ 2940 mask = vap->va_mask; 2941 2942 tx = dmu_tx_create(zfsvfs->z_os); 2943 dmu_tx_hold_bonus(tx, zp->z_id); 2944 2945 if (mask & AT_MODE) { 2946 uint64_t pmode = pzp->zp_mode; 2947 2948 new_mode = (pmode & S_IFMT) | (vap->va_mode & ~S_IFMT); 2949 2950 if (err = zfs_acl_chmod_setattr(zp, &aclp, new_mode)) 2951 goto out; 2952 if (pzp->zp_acl.z_acl_extern_obj) { 2953 /* Are we upgrading ACL from old V0 format to new V1 */ 2954 if (zfsvfs->z_version <= ZPL_VERSION_FUID && 2955 pzp->zp_acl.z_acl_version == 2956 ZFS_ACL_VERSION_INITIAL) { 2957 dmu_tx_hold_free(tx, 2958 pzp->zp_acl.z_acl_extern_obj, 0, 2959 DMU_OBJECT_END); 2960 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 2961 0, aclp->z_acl_bytes); 2962 } else { 2963 dmu_tx_hold_write(tx, 2964 pzp->zp_acl.z_acl_extern_obj, 0, 2965 aclp->z_acl_bytes); 2966 } 2967 } else if (aclp->z_acl_bytes > ZFS_ACE_SPACE) { 2968 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 2969 0, aclp->z_acl_bytes); 2970 } 2971 } 2972 2973 if (mask & (AT_UID | AT_GID)) { 2974 if (pzp->zp_xattr) { 2975 err = zfs_zget(zp->z_zfsvfs, pzp->zp_xattr, &attrzp); 2976 if (err) 2977 goto out; 2978 dmu_tx_hold_bonus(tx, attrzp->z_id); 2979 } 2980 if (mask & AT_UID) { 2981 new_uid = zfs_fuid_create(zfsvfs, 2982 (uint64_t)vap->va_uid, cr, ZFS_OWNER, &fuidp); 2983 if (new_uid != pzp->zp_uid && 2984 zfs_usergroup_overquota(zfsvfs, B_FALSE, new_uid)) { 2985 err = EDQUOT; 2986 goto out; 2987 } 2988 } 2989 2990 if (mask & AT_GID) { 2991 new_gid = zfs_fuid_create(zfsvfs, (uint64_t)vap->va_gid, 2992 cr, ZFS_GROUP, &fuidp); 2993 if (new_gid != pzp->zp_gid && 2994 zfs_usergroup_overquota(zfsvfs, B_TRUE, new_gid)) { 2995 err = EDQUOT; 2996 goto out; 2997 } 2998 } 2999 fuid_dirtied = zfsvfs->z_fuid_dirty; 3000 if (fuid_dirtied) { 3001 if (zfsvfs->z_fuid_obj == 0) { 3002 dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT); 3003 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, 3004 FUID_SIZE_ESTIMATE(zfsvfs)); 3005 dmu_tx_hold_zap(tx, MASTER_NODE_OBJ, 3006 FALSE, NULL); 3007 } else { 3008 dmu_tx_hold_bonus(tx, zfsvfs->z_fuid_obj); 3009 dmu_tx_hold_write(tx, zfsvfs->z_fuid_obj, 0, 3010 FUID_SIZE_ESTIMATE(zfsvfs)); 3011 } 3012 } 3013 } 3014 3015 err = dmu_tx_assign(tx, TXG_NOWAIT); 3016 if (err) { 3017 if (err == ERESTART) 3018 dmu_tx_wait(tx); 3019 goto out; 3020 } 3021 3022 dmu_buf_will_dirty(zp->z_dbuf, tx); 3023 3024 /* 3025 * Set each attribute requested. 3026 * We group settings according to the locks they need to acquire. 3027 * 3028 * Note: you cannot set ctime directly, although it will be 3029 * updated as a side-effect of calling this function. 3030 */ 3031 3032 mutex_enter(&zp->z_lock); 3033 3034 if (mask & AT_MODE) { 3035 mutex_enter(&zp->z_acl_lock); 3036 zp->z_phys->zp_mode = new_mode; 3037 err = zfs_aclset_common(zp, aclp, cr, tx); 3038 ASSERT3U(err, ==, 0); 3039 mutex_exit(&zp->z_acl_lock); 3040 } 3041 3042 if (attrzp) 3043 mutex_enter(&attrzp->z_lock); 3044 3045 if (mask & AT_UID) { 3046 pzp->zp_uid = new_uid; 3047 if (attrzp) 3048 attrzp->z_phys->zp_uid = new_uid; 3049 } 3050 3051 if (mask & AT_GID) { 3052 pzp->zp_gid = new_gid; 3053 if (attrzp) 3054 attrzp->z_phys->zp_gid = new_gid; 3055 } 3056 3057 if (attrzp) 3058 mutex_exit(&attrzp->z_lock); 3059 3060 if (mask & AT_ATIME) 3061 ZFS_TIME_ENCODE(&vap->va_atime, pzp->zp_atime); 3062 3063 if (mask & AT_MTIME) 3064 ZFS_TIME_ENCODE(&vap->va_mtime, pzp->zp_mtime); 3065 3066 /* XXX - shouldn't this be done *before* the ATIME/MTIME checks? */ 3067 if (mask & AT_SIZE) 3068 zfs_time_stamper_locked(zp, CONTENT_MODIFIED, tx); 3069 else if (mask != 0) 3070 zfs_time_stamper_locked(zp, STATE_CHANGED, tx); 3071 /* 3072 * Do this after setting timestamps to prevent timestamp 3073 * update from toggling bit 3074 */ 3075 3076 if (xoap && (mask & AT_XVATTR)) { 3077 3078 /* 3079 * restore trimmed off masks 3080 * so that return masks can be set for caller. 3081 */ 3082 3083 if (XVA_ISSET_REQ(&tmpxvattr, XAT_APPENDONLY)) { 3084 XVA_SET_REQ(xvap, XAT_APPENDONLY); 3085 } 3086 if (XVA_ISSET_REQ(&tmpxvattr, XAT_NOUNLINK)) { 3087 XVA_SET_REQ(xvap, XAT_NOUNLINK); 3088 } 3089 if (XVA_ISSET_REQ(&tmpxvattr, XAT_IMMUTABLE)) { 3090 XVA_SET_REQ(xvap, XAT_IMMUTABLE); 3091 } 3092 if (XVA_ISSET_REQ(&tmpxvattr, XAT_NODUMP)) { 3093 XVA_SET_REQ(xvap, XAT_NODUMP); 3094 } 3095 if (XVA_ISSET_REQ(&tmpxvattr, XAT_AV_MODIFIED)) { 3096 XVA_SET_REQ(xvap, XAT_AV_MODIFIED); 3097 } 3098 if (XVA_ISSET_REQ(&tmpxvattr, XAT_AV_QUARANTINED)) { 3099 XVA_SET_REQ(xvap, XAT_AV_QUARANTINED); 3100 } 3101 3102 if (XVA_ISSET_REQ(xvap, XAT_AV_SCANSTAMP)) { 3103 size_t len; 3104 dmu_object_info_t doi; 3105 3106 ASSERT(vp->v_type == VREG); 3107 3108 /* Grow the bonus buffer if necessary. */ 3109 dmu_object_info_from_db(zp->z_dbuf, &doi); 3110 len = sizeof (xoap->xoa_av_scanstamp) + 3111 sizeof (znode_phys_t); 3112 if (len > doi.doi_bonus_size) 3113 VERIFY(dmu_set_bonus(zp->z_dbuf, len, tx) == 0); 3114 } 3115 zfs_xvattr_set(zp, xvap); 3116 } 3117 3118 if (fuid_dirtied) 3119 zfs_fuid_sync(zfsvfs, tx); 3120 3121 if (mask != 0) 3122 zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask, fuidp); 3123 3124 mutex_exit(&zp->z_lock); 3125 3126out: 3127 if (attrzp) 3128 VN_RELE(ZTOV(attrzp)); 3129 3130 if (aclp) { 3131 zfs_acl_free(aclp); 3132 aclp = NULL; 3133 } 3134 3135 if (fuidp) { 3136 zfs_fuid_info_free(fuidp); 3137 fuidp = NULL; 3138 } 3139 3140 if (err) 3141 dmu_tx_abort(tx); 3142 else 3143 dmu_tx_commit(tx); 3144 3145 if (err == ERESTART) 3146 goto top; 3147 3148 ZFS_EXIT(zfsvfs); 3149 return (err); 3150} 3151 3152typedef struct zfs_zlock { 3153 krwlock_t *zl_rwlock; /* lock we acquired */ 3154 znode_t *zl_znode; /* znode we held */ 3155 struct zfs_zlock *zl_next; /* next in list */ 3156} zfs_zlock_t; 3157 3158/* 3159 * Drop locks and release vnodes that were held by zfs_rename_lock(). 3160 */ 3161static void 3162zfs_rename_unlock(zfs_zlock_t **zlpp) 3163{ 3164 zfs_zlock_t *zl; 3165 3166 while ((zl = *zlpp) != NULL) { 3167 if (zl->zl_znode != NULL) 3168 VN_RELE(ZTOV(zl->zl_znode)); 3169 rw_exit(zl->zl_rwlock); 3170 *zlpp = zl->zl_next; 3171 kmem_free(zl, sizeof (*zl)); 3172 } 3173} 3174 3175/* 3176 * Search back through the directory tree, using the ".." entries. 3177 * Lock each directory in the chain to prevent concurrent renames. 3178 * Fail any attempt to move a directory into one of its own descendants. 3179 * XXX - z_parent_lock can overlap with map or grow locks 3180 */ 3181static int 3182zfs_rename_lock(znode_t *szp, znode_t *tdzp, znode_t *sdzp, zfs_zlock_t **zlpp) 3183{ 3184 zfs_zlock_t *zl; 3185 znode_t *zp = tdzp; 3186 uint64_t rootid = zp->z_zfsvfs->z_root; 3187 uint64_t *oidp = &zp->z_id; 3188 krwlock_t *rwlp = &szp->z_parent_lock; 3189 krw_t rw = RW_WRITER; 3190 3191 /* 3192 * First pass write-locks szp and compares to zp->z_id. 3193 * Later passes read-lock zp and compare to zp->z_parent. 3194 */ 3195 do { 3196 if (!rw_tryenter(rwlp, rw)) { 3197 /* 3198 * Another thread is renaming in this path. 3199 * Note that if we are a WRITER, we don't have any 3200 * parent_locks held yet. 3201 */ 3202 if (rw == RW_READER && zp->z_id > szp->z_id) { 3203 /* 3204 * Drop our locks and restart 3205 */ 3206 zfs_rename_unlock(&zl); 3207 *zlpp = NULL; 3208 zp = tdzp; 3209 oidp = &zp->z_id; 3210 rwlp = &szp->z_parent_lock; 3211 rw = RW_WRITER; 3212 continue; 3213 } else { 3214 /* 3215 * Wait for other thread to drop its locks 3216 */ 3217 rw_enter(rwlp, rw); 3218 } 3219 } 3220 3221 zl = kmem_alloc(sizeof (*zl), KM_SLEEP); 3222 zl->zl_rwlock = rwlp; 3223 zl->zl_znode = NULL; 3224 zl->zl_next = *zlpp; 3225 *zlpp = zl; 3226 3227 if (*oidp == szp->z_id) /* We're a descendant of szp */ 3228 return (EINVAL); 3229 3230 if (*oidp == rootid) /* We've hit the top */ 3231 return (0); 3232 3233 if (rw == RW_READER) { /* i.e. not the first pass */ 3234 int error = zfs_zget(zp->z_zfsvfs, *oidp, &zp); 3235 if (error) 3236 return (error); 3237 zl->zl_znode = zp; 3238 } 3239 oidp = &zp->z_phys->zp_parent; 3240 rwlp = &zp->z_parent_lock; 3241 rw = RW_READER; 3242 3243 } while (zp->z_id != sdzp->z_id); 3244 3245 return (0); 3246} 3247 3248/* 3249 * Move an entry from the provided source directory to the target 3250 * directory. Change the entry name as indicated. 3251 * 3252 * IN: sdvp - Source directory containing the "old entry". 3253 * snm - Old entry name. 3254 * tdvp - Target directory to contain the "new entry". 3255 * tnm - New entry name. 3256 * cr - credentials of caller. 3257 * ct - caller context 3258 * flags - case flags 3259 * 3260 * RETURN: 0 if success 3261 * error code if failure 3262 * 3263 * Timestamps: 3264 * sdvp,tdvp - ctime|mtime updated 3265 */ 3266/*ARGSUSED*/ 3267static int 3268zfs_rename(vnode_t *sdvp, char *snm, vnode_t *tdvp, char *tnm, cred_t *cr, 3269 caller_context_t *ct, int flags) 3270{ 3271 znode_t *tdzp, *szp, *tzp; 3272 znode_t *sdzp = VTOZ(sdvp); 3273 zfsvfs_t *zfsvfs = sdzp->z_zfsvfs; 3274 zilog_t *zilog; 3275 vnode_t *realvp; 3276 zfs_dirlock_t *sdl, *tdl; 3277 dmu_tx_t *tx; 3278 zfs_zlock_t *zl; 3279 int cmp, serr, terr; 3280 int error = 0; 3281 int zflg = 0; 3282 3283 ZFS_ENTER(zfsvfs); 3284 ZFS_VERIFY_ZP(sdzp); 3285 zilog = zfsvfs->z_log; 3286 3287 /* 3288 * Make sure we have the real vp for the target directory. 3289 */ 3290 if (VOP_REALVP(tdvp, &realvp, ct) == 0) 3291 tdvp = realvp; 3292 3293 if (tdvp->v_vfsp != sdvp->v_vfsp) { 3294 ZFS_EXIT(zfsvfs); 3295 return (EXDEV); 3296 } 3297 3298 tdzp = VTOZ(tdvp); 3299 ZFS_VERIFY_ZP(tdzp); 3300 if (zfsvfs->z_utf8 && u8_validate(tnm, 3301 strlen(tnm), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { 3302 ZFS_EXIT(zfsvfs); 3303 return (EILSEQ); 3304 } 3305 3306 if (flags & FIGNORECASE) 3307 zflg |= ZCILOOK; 3308 3309top: 3310 szp = NULL; 3311 tzp = NULL; 3312 zl = NULL; 3313 3314 /* 3315 * This is to prevent the creation of links into attribute space 3316 * by renaming a linked file into/outof an attribute directory. 3317 * See the comment in zfs_link() for why this is considered bad. 3318 */ 3319 if ((tdzp->z_phys->zp_flags & ZFS_XATTR) != 3320 (sdzp->z_phys->zp_flags & ZFS_XATTR)) { 3321 ZFS_EXIT(zfsvfs); 3322 return (EINVAL); 3323 } 3324 3325 /* 3326 * Lock source and target directory entries. To prevent deadlock, 3327 * a lock ordering must be defined. We lock the directory with 3328 * the smallest object id first, or if it's a tie, the one with 3329 * the lexically first name. 3330 */ 3331 if (sdzp->z_id < tdzp->z_id) { 3332 cmp = -1; 3333 } else if (sdzp->z_id > tdzp->z_id) { 3334 cmp = 1; 3335 } else { 3336 /* 3337 * First compare the two name arguments without 3338 * considering any case folding. 3339 */ 3340 int nofold = (zfsvfs->z_norm & ~U8_TEXTPREP_TOUPPER); 3341 3342 cmp = u8_strcmp(snm, tnm, 0, nofold, U8_UNICODE_LATEST, &error); 3343 ASSERT(error == 0 || !zfsvfs->z_utf8); 3344 if (cmp == 0) { 3345 /* 3346 * POSIX: "If the old argument and the new argument 3347 * both refer to links to the same existing file, 3348 * the rename() function shall return successfully 3349 * and perform no other action." 3350 */ 3351 ZFS_EXIT(zfsvfs); 3352 return (0); 3353 } 3354 /* 3355 * If the file system is case-folding, then we may 3356 * have some more checking to do. A case-folding file 3357 * system is either supporting mixed case sensitivity 3358 * access or is completely case-insensitive. Note 3359 * that the file system is always case preserving. 3360 * 3361 * In mixed sensitivity mode case sensitive behavior 3362 * is the default. FIGNORECASE must be used to 3363 * explicitly request case insensitive behavior. 3364 * 3365 * If the source and target names provided differ only 3366 * by case (e.g., a request to rename 'tim' to 'Tim'), 3367 * we will treat this as a special case in the 3368 * case-insensitive mode: as long as the source name 3369 * is an exact match, we will allow this to proceed as 3370 * a name-change request. 3371 */ 3372 if ((zfsvfs->z_case == ZFS_CASE_INSENSITIVE || 3373 (zfsvfs->z_case == ZFS_CASE_MIXED && 3374 flags & FIGNORECASE)) && 3375 u8_strcmp(snm, tnm, 0, zfsvfs->z_norm, U8_UNICODE_LATEST, 3376 &error) == 0) { 3377 /* 3378 * case preserving rename request, require exact 3379 * name matches 3380 */ 3381 zflg |= ZCIEXACT; 3382 zflg &= ~ZCILOOK; 3383 } 3384 } 3385 3386 /* 3387 * If the source and destination directories are the same, we should 3388 * grab the z_name_lock of that directory only once. 3389 */ 3390 if (sdzp == tdzp) { 3391 zflg |= ZHAVELOCK; 3392 rw_enter(&sdzp->z_name_lock, RW_READER); 3393 } 3394 3395 if (cmp < 0) { 3396 serr = zfs_dirent_lock(&sdl, sdzp, snm, &szp, 3397 ZEXISTS | zflg, NULL, NULL); 3398 terr = zfs_dirent_lock(&tdl, 3399 tdzp, tnm, &tzp, ZRENAMING | zflg, NULL, NULL); 3400 } else { 3401 terr = zfs_dirent_lock(&tdl, 3402 tdzp, tnm, &tzp, zflg, NULL, NULL); 3403 serr = zfs_dirent_lock(&sdl, 3404 sdzp, snm, &szp, ZEXISTS | ZRENAMING | zflg, 3405 NULL, NULL); 3406 } 3407 3408 if (serr) { 3409 /* 3410 * Source entry invalid or not there. 3411 */ 3412 if (!terr) { 3413 zfs_dirent_unlock(tdl); 3414 if (tzp) 3415 VN_RELE(ZTOV(tzp)); 3416 } 3417 3418 if (sdzp == tdzp) 3419 rw_exit(&sdzp->z_name_lock); 3420 3421 if (strcmp(snm, ".") == 0 || strcmp(snm, "..") == 0) 3422 serr = EINVAL; 3423 ZFS_EXIT(zfsvfs); 3424 return (serr); 3425 } 3426 if (terr) { 3427 zfs_dirent_unlock(sdl); 3428 VN_RELE(ZTOV(szp)); 3429 3430 if (sdzp == tdzp) 3431 rw_exit(&sdzp->z_name_lock); 3432 3433 if (strcmp(tnm, "..") == 0) 3434 terr = EINVAL; 3435 ZFS_EXIT(zfsvfs); 3436 return (terr); 3437 } 3438 3439 /* 3440 * Must have write access at the source to remove the old entry 3441 * and write access at the target to create the new entry. 3442 * Note that if target and source are the same, this can be 3443 * done in a single check. 3444 */ 3445 3446 if (error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr)) 3447 goto out; 3448 3449 if (ZTOV(szp)->v_type == VDIR) { 3450 /* 3451 * Check to make sure rename is valid. 3452 * Can't do a move like this: /usr/a/b to /usr/a/b/c/d 3453 */ 3454 if (error = zfs_rename_lock(szp, tdzp, sdzp, &zl)) 3455 goto out; 3456 } 3457 3458 /* 3459 * Does target exist? 3460 */ 3461 if (tzp) { 3462 /* 3463 * Source and target must be the same type. 3464 */ 3465 if (ZTOV(szp)->v_type == VDIR) { 3466 if (ZTOV(tzp)->v_type != VDIR) { 3467 error = ENOTDIR; 3468 goto out; 3469 } 3470 } else { 3471 if (ZTOV(tzp)->v_type == VDIR) { 3472 error = EISDIR; 3473 goto out; 3474 } 3475 } 3476 /* 3477 * POSIX dictates that when the source and target 3478 * entries refer to the same file object, rename 3479 * must do nothing and exit without error. 3480 */ 3481 if (szp->z_id == tzp->z_id) { 3482 error = 0; 3483 goto out; 3484 } 3485 } 3486 3487 vnevent_rename_src(ZTOV(szp), sdvp, snm, ct); 3488 if (tzp) 3489 vnevent_rename_dest(ZTOV(tzp), tdvp, tnm, ct); 3490 3491 /* 3492 * notify the target directory if it is not the same 3493 * as source directory. 3494 */ 3495 if (tdvp != sdvp) { 3496 vnevent_rename_dest_dir(tdvp, ct); 3497 } 3498 3499 tx = dmu_tx_create(zfsvfs->z_os); 3500 dmu_tx_hold_bonus(tx, szp->z_id); /* nlink changes */ 3501 dmu_tx_hold_bonus(tx, sdzp->z_id); /* nlink changes */ 3502 dmu_tx_hold_zap(tx, sdzp->z_id, FALSE, snm); 3503 dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, tnm); 3504 if (sdzp != tdzp) 3505 dmu_tx_hold_bonus(tx, tdzp->z_id); /* nlink changes */ 3506 if (tzp) 3507 dmu_tx_hold_bonus(tx, tzp->z_id); /* parent changes */ 3508 dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); 3509 error = dmu_tx_assign(tx, TXG_NOWAIT); 3510 if (error) { 3511 if (zl != NULL) 3512 zfs_rename_unlock(&zl); 3513 zfs_dirent_unlock(sdl); 3514 zfs_dirent_unlock(tdl); 3515 3516 if (sdzp == tdzp) 3517 rw_exit(&sdzp->z_name_lock); 3518 3519 VN_RELE(ZTOV(szp)); 3520 if (tzp) 3521 VN_RELE(ZTOV(tzp)); 3522 if (error == ERESTART) { 3523 dmu_tx_wait(tx); 3524 dmu_tx_abort(tx); 3525 goto top; 3526 } 3527 dmu_tx_abort(tx); 3528 ZFS_EXIT(zfsvfs); 3529 return (error); 3530 } 3531 3532 if (tzp) /* Attempt to remove the existing target */ 3533 error = zfs_link_destroy(tdl, tzp, tx, zflg, NULL); 3534 3535 if (error == 0) { 3536 error = zfs_link_create(tdl, szp, tx, ZRENAMING); 3537 if (error == 0) { 3538 szp->z_phys->zp_flags |= ZFS_AV_MODIFIED; 3539 3540 error = zfs_link_destroy(sdl, szp, tx, ZRENAMING, NULL); 3541 ASSERT(error == 0); 3542 3543 zfs_log_rename(zilog, tx, 3544 TX_RENAME | (flags & FIGNORECASE ? TX_CI : 0), 3545 sdzp, sdl->dl_name, tdzp, tdl->dl_name, szp); 3546 3547 /* Update path information for the target vnode */ 3548 vn_renamepath(tdvp, ZTOV(szp), tnm, strlen(tnm)); 3549 } 3550#ifdef FREEBSD_NAMECACHE 3551 if (error == 0) { 3552 cache_purge(sdvp); 3553 cache_purge(tdvp); 3554 } 3555#endif 3556 } 3557 3558 dmu_tx_commit(tx); 3559out: 3560 if (zl != NULL) 3561 zfs_rename_unlock(&zl); 3562 3563 zfs_dirent_unlock(sdl); 3564 zfs_dirent_unlock(tdl); 3565 3566 if (sdzp == tdzp) 3567 rw_exit(&sdzp->z_name_lock); 3568 3569 VN_RELE(ZTOV(szp)); 3570 if (tzp) 3571 VN_RELE(ZTOV(tzp)); 3572 3573 ZFS_EXIT(zfsvfs); 3574 3575 return (error); 3576} 3577 3578/* 3579 * Insert the indicated symbolic reference entry into the directory. 3580 * 3581 * IN: dvp - Directory to contain new symbolic link. 3582 * link - Name for new symlink entry. 3583 * vap - Attributes of new entry. 3584 * target - Target path of new symlink. 3585 * cr - credentials of caller. 3586 * ct - caller context 3587 * flags - case flags 3588 * 3589 * RETURN: 0 if success 3590 * error code if failure 3591 * 3592 * Timestamps: 3593 * dvp - ctime|mtime updated 3594 */ 3595/*ARGSUSED*/ 3596static int 3597zfs_symlink(vnode_t *dvp, vnode_t **vpp, char *name, vattr_t *vap, char *link, 3598 cred_t *cr, kthread_t *td) 3599{ 3600 znode_t *zp, *dzp = VTOZ(dvp); 3601 zfs_dirlock_t *dl; 3602 dmu_tx_t *tx; 3603 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 3604 zilog_t *zilog; 3605 int len = strlen(link); 3606 int error; 3607 int zflg = ZNEW; 3608 zfs_acl_ids_t acl_ids; 3609 boolean_t fuid_dirtied; 3610 int flags = 0; 3611 3612 ASSERT(vap->va_type == VLNK); 3613 3614 ZFS_ENTER(zfsvfs); 3615 ZFS_VERIFY_ZP(dzp); 3616 zilog = zfsvfs->z_log; 3617 3618 if (zfsvfs->z_utf8 && u8_validate(name, strlen(name), 3619 NULL, U8_VALIDATE_ENTIRE, &error) < 0) { 3620 ZFS_EXIT(zfsvfs); 3621 return (EILSEQ); 3622 } 3623 if (flags & FIGNORECASE) 3624 zflg |= ZCILOOK; 3625top: 3626 if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) { 3627 ZFS_EXIT(zfsvfs); 3628 return (error); 3629 } 3630 3631 if (len > MAXPATHLEN) { 3632 ZFS_EXIT(zfsvfs); 3633 return (ENAMETOOLONG); 3634 } 3635 3636 /* 3637 * Attempt to lock directory; fail if entry already exists. 3638 */ 3639 error = zfs_dirent_lock(&dl, dzp, name, &zp, zflg, NULL, NULL); 3640 if (error) { 3641 ZFS_EXIT(zfsvfs); 3642 return (error); 3643 } 3644 3645 VERIFY(0 == zfs_acl_ids_create(dzp, 0, vap, cr, NULL, &acl_ids)); 3646 if (zfs_acl_ids_overquota(zfsvfs, &acl_ids)) { 3647 zfs_acl_ids_free(&acl_ids); 3648 zfs_dirent_unlock(dl); 3649 ZFS_EXIT(zfsvfs); 3650 return (EDQUOT); 3651 } 3652 tx = dmu_tx_create(zfsvfs->z_os); 3653 fuid_dirtied = zfsvfs->z_fuid_dirty; 3654 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len)); 3655 dmu_tx_hold_bonus(tx, dzp->z_id); 3656 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); 3657 if (acl_ids.z_aclp->z_acl_bytes > ZFS_ACE_SPACE) 3658 dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, SPA_MAXBLOCKSIZE); 3659 if (fuid_dirtied) 3660 zfs_fuid_txhold(zfsvfs, tx); 3661 error = dmu_tx_assign(tx, TXG_NOWAIT); 3662 if (error) { 3663 zfs_acl_ids_free(&acl_ids); 3664 zfs_dirent_unlock(dl); 3665 if (error == ERESTART) { 3666 dmu_tx_wait(tx); 3667 dmu_tx_abort(tx); 3668 goto top; 3669 } 3670 dmu_tx_abort(tx); 3671 ZFS_EXIT(zfsvfs); 3672 return (error); 3673 } 3674 3675 dmu_buf_will_dirty(dzp->z_dbuf, tx); 3676 3677 /* 3678 * Create a new object for the symlink. 3679 * Put the link content into bonus buffer if it will fit; 3680 * otherwise, store it just like any other file data. 3681 */ 3682 if (sizeof (znode_phys_t) + len <= dmu_bonus_max()) { 3683 zfs_mknode(dzp, vap, tx, cr, 0, &zp, len, &acl_ids); 3684 if (len != 0) 3685 bcopy(link, zp->z_phys + 1, len); 3686 } else { 3687 dmu_buf_t *dbp; 3688 3689 zfs_mknode(dzp, vap, tx, cr, 0, &zp, 0, &acl_ids); 3690 3691 if (fuid_dirtied) 3692 zfs_fuid_sync(zfsvfs, tx); 3693 /* 3694 * Nothing can access the znode yet so no locking needed 3695 * for growing the znode's blocksize. 3696 */ 3697 zfs_grow_blocksize(zp, len, tx); 3698 3699 VERIFY(0 == dmu_buf_hold(zfsvfs->z_os, 3700 zp->z_id, 0, FTAG, &dbp)); 3701 dmu_buf_will_dirty(dbp, tx); 3702 3703 ASSERT3U(len, <=, dbp->db_size); 3704 bcopy(link, dbp->db_data, len); 3705 dmu_buf_rele(dbp, FTAG); 3706 } 3707 zp->z_phys->zp_size = len; 3708 3709 /* 3710 * Insert the new object into the directory. 3711 */ 3712 (void) zfs_link_create(dl, zp, tx, ZNEW); 3713 if (error == 0) { 3714 uint64_t txtype = TX_SYMLINK; 3715 if (flags & FIGNORECASE) 3716 txtype |= TX_CI; 3717 zfs_log_symlink(zilog, tx, txtype, dzp, zp, name, link); 3718 *vpp = ZTOV(zp); 3719 } 3720 3721 zfs_acl_ids_free(&acl_ids); 3722 3723 dmu_tx_commit(tx); 3724 3725 zfs_dirent_unlock(dl); 3726 3727 ZFS_EXIT(zfsvfs); 3728 return (error); 3729} 3730 3731/* 3732 * Return, in the buffer contained in the provided uio structure, 3733 * the symbolic path referred to by vp. 3734 * 3735 * IN: vp - vnode of symbolic link. 3736 * uoip - structure to contain the link path. 3737 * cr - credentials of caller. 3738 * ct - caller context 3739 * 3740 * OUT: uio - structure to contain the link path. 3741 * 3742 * RETURN: 0 if success 3743 * error code if failure 3744 * 3745 * Timestamps: 3746 * vp - atime updated 3747 */ 3748/* ARGSUSED */ 3749static int 3750zfs_readlink(vnode_t *vp, uio_t *uio, cred_t *cr, caller_context_t *ct) 3751{ 3752 znode_t *zp = VTOZ(vp); 3753 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 3754 size_t bufsz; 3755 int error; 3756 3757 ZFS_ENTER(zfsvfs); 3758 ZFS_VERIFY_ZP(zp); 3759 3760 bufsz = (size_t)zp->z_phys->zp_size; 3761 if (bufsz + sizeof (znode_phys_t) <= zp->z_dbuf->db_size) { 3762 error = uiomove(zp->z_phys + 1, 3763 MIN((size_t)bufsz, uio->uio_resid), UIO_READ, uio); 3764 } else { 3765 dmu_buf_t *dbp; 3766 error = dmu_buf_hold(zfsvfs->z_os, zp->z_id, 0, FTAG, &dbp); 3767 if (error) { 3768 ZFS_EXIT(zfsvfs); 3769 return (error); 3770 } 3771 error = uiomove(dbp->db_data, 3772 MIN((size_t)bufsz, uio->uio_resid), UIO_READ, uio); 3773 dmu_buf_rele(dbp, FTAG); 3774 } 3775 3776 ZFS_ACCESSTIME_STAMP(zfsvfs, zp); 3777 ZFS_EXIT(zfsvfs); 3778 return (error); 3779} 3780 3781/* 3782 * Insert a new entry into directory tdvp referencing svp. 3783 * 3784 * IN: tdvp - Directory to contain new entry. 3785 * svp - vnode of new entry. 3786 * name - name of new entry. 3787 * cr - credentials of caller. 3788 * ct - caller context 3789 * 3790 * RETURN: 0 if success 3791 * error code if failure 3792 * 3793 * Timestamps: 3794 * tdvp - ctime|mtime updated 3795 * svp - ctime updated 3796 */ 3797/* ARGSUSED */ 3798static int 3799zfs_link(vnode_t *tdvp, vnode_t *svp, char *name, cred_t *cr, 3800 caller_context_t *ct, int flags) 3801{ 3802 znode_t *dzp = VTOZ(tdvp); 3803 znode_t *tzp, *szp; 3804 zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 3805 zilog_t *zilog; 3806 zfs_dirlock_t *dl; 3807 dmu_tx_t *tx; 3808 vnode_t *realvp; 3809 int error; 3810 int zf = ZNEW; 3811 uid_t owner; 3812 3813 ASSERT(tdvp->v_type == VDIR); 3814 3815 ZFS_ENTER(zfsvfs); 3816 ZFS_VERIFY_ZP(dzp); 3817 zilog = zfsvfs->z_log; 3818 3819 if (VOP_REALVP(svp, &realvp, ct) == 0) 3820 svp = realvp; 3821 3822 if (svp->v_vfsp != tdvp->v_vfsp) { 3823 ZFS_EXIT(zfsvfs); 3824 return (EXDEV); 3825 } 3826 szp = VTOZ(svp); 3827 ZFS_VERIFY_ZP(szp); 3828 3829 if (zfsvfs->z_utf8 && u8_validate(name, 3830 strlen(name), NULL, U8_VALIDATE_ENTIRE, &error) < 0) { 3831 ZFS_EXIT(zfsvfs); 3832 return (EILSEQ); 3833 } 3834 if (flags & FIGNORECASE) 3835 zf |= ZCILOOK; 3836 3837top: 3838 /* 3839 * We do not support links between attributes and non-attributes 3840 * because of the potential security risk of creating links 3841 * into "normal" file space in order to circumvent restrictions 3842 * imposed in attribute space. 3843 */ 3844 if ((szp->z_phys->zp_flags & ZFS_XATTR) != 3845 (dzp->z_phys->zp_flags & ZFS_XATTR)) { 3846 ZFS_EXIT(zfsvfs); 3847 return (EINVAL); 3848 } 3849 3850 /* 3851 * POSIX dictates that we return EPERM here. 3852 * Better choices include ENOTSUP or EISDIR. 3853 */ 3854 if (svp->v_type == VDIR) { 3855 ZFS_EXIT(zfsvfs); 3856 return (EPERM); 3857 } 3858 3859 owner = zfs_fuid_map_id(zfsvfs, szp->z_phys->zp_uid, cr, ZFS_OWNER); 3860 if (owner != crgetuid(cr) && 3861 secpolicy_basic_link(svp, cr) != 0) { 3862 ZFS_EXIT(zfsvfs); 3863 return (EPERM); 3864 } 3865 3866 if (error = zfs_zaccess(dzp, ACE_ADD_FILE, 0, B_FALSE, cr)) { 3867 ZFS_EXIT(zfsvfs); 3868 return (error); 3869 } 3870 3871 /* 3872 * Attempt to lock directory; fail if entry already exists. 3873 */ 3874 error = zfs_dirent_lock(&dl, dzp, name, &tzp, zf, NULL, NULL); 3875 if (error) { 3876 ZFS_EXIT(zfsvfs); 3877 return (error); 3878 } 3879 3880 tx = dmu_tx_create(zfsvfs->z_os); 3881 dmu_tx_hold_bonus(tx, szp->z_id); 3882 dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); 3883 error = dmu_tx_assign(tx, TXG_NOWAIT); 3884 if (error) { 3885 zfs_dirent_unlock(dl); 3886 if (error == ERESTART) { 3887 dmu_tx_wait(tx); 3888 dmu_tx_abort(tx); 3889 goto top; 3890 } 3891 dmu_tx_abort(tx); 3892 ZFS_EXIT(zfsvfs); 3893 return (error); 3894 } 3895 3896 error = zfs_link_create(dl, szp, tx, 0); 3897 3898 if (error == 0) { 3899 uint64_t txtype = TX_LINK; 3900 if (flags & FIGNORECASE) 3901 txtype |= TX_CI; 3902 zfs_log_link(zilog, tx, txtype, dzp, szp, name); 3903 } 3904 3905 dmu_tx_commit(tx); 3906 3907 zfs_dirent_unlock(dl); 3908 3909 if (error == 0) { 3910 vnevent_link(svp, ct); 3911 } 3912 3913 ZFS_EXIT(zfsvfs); 3914 return (error); 3915} 3916 3917/*ARGSUSED*/ 3918void 3919zfs_inactive(vnode_t *vp, cred_t *cr, caller_context_t *ct) 3920{ 3921 znode_t *zp = VTOZ(vp); 3922 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 3923 int error; 3924 3925 rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER); 3926 if (zp->z_dbuf == NULL) { 3927 /* 3928 * The fs has been unmounted, or we did a 3929 * suspend/resume and this file no longer exists. 3930 */ 3931 VI_LOCK(vp); 3932 vp->v_count = 0; /* count arrives as 1 */ 3933 VI_UNLOCK(vp); 3934 vrecycle(vp, curthread); 3935 rw_exit(&zfsvfs->z_teardown_inactive_lock); 3936 return; 3937 } 3938 3939 if (zp->z_atime_dirty && zp->z_unlinked == 0) { 3940 dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os); 3941 3942 dmu_tx_hold_bonus(tx, zp->z_id); 3943 error = dmu_tx_assign(tx, TXG_WAIT); 3944 if (error) { 3945 dmu_tx_abort(tx); 3946 } else { 3947 dmu_buf_will_dirty(zp->z_dbuf, tx); 3948 mutex_enter(&zp->z_lock); 3949 zp->z_atime_dirty = 0; 3950 mutex_exit(&zp->z_lock); 3951 dmu_tx_commit(tx); 3952 } 3953 } 3954 3955 zfs_zinactive(zp); 3956 rw_exit(&zfsvfs->z_teardown_inactive_lock); 3957} 3958 3959CTASSERT(sizeof(struct zfid_short) <= sizeof(struct fid)); 3960CTASSERT(sizeof(struct zfid_long) <= sizeof(struct fid)); 3961 3962/*ARGSUSED*/ 3963static int 3964zfs_fid(vnode_t *vp, fid_t *fidp, caller_context_t *ct) 3965{ 3966 znode_t *zp = VTOZ(vp); 3967 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 3968 uint32_t gen; 3969 uint64_t object = zp->z_id; 3970 zfid_short_t *zfid; 3971 int size, i; 3972 3973 ZFS_ENTER(zfsvfs); 3974 ZFS_VERIFY_ZP(zp); 3975 gen = (uint32_t)zp->z_gen; 3976 3977 size = (zfsvfs->z_parent != zfsvfs) ? LONG_FID_LEN : SHORT_FID_LEN; 3978 fidp->fid_len = size; 3979 3980 zfid = (zfid_short_t *)fidp; 3981 3982 zfid->zf_len = size; 3983 3984 for (i = 0; i < sizeof (zfid->zf_object); i++) 3985 zfid->zf_object[i] = (uint8_t)(object >> (8 * i)); 3986 3987 /* Must have a non-zero generation number to distinguish from .zfs */ 3988 if (gen == 0) 3989 gen = 1; 3990 for (i = 0; i < sizeof (zfid->zf_gen); i++) 3991 zfid->zf_gen[i] = (uint8_t)(gen >> (8 * i)); 3992 3993 if (size == LONG_FID_LEN) { 3994 uint64_t objsetid = dmu_objset_id(zfsvfs->z_os); 3995 zfid_long_t *zlfid; 3996 3997 zlfid = (zfid_long_t *)fidp; 3998 3999 for (i = 0; i < sizeof (zlfid->zf_setid); i++) 4000 zlfid->zf_setid[i] = (uint8_t)(objsetid >> (8 * i)); 4001 4002 /* XXX - this should be the generation number for the objset */ 4003 for (i = 0; i < sizeof (zlfid->zf_setgen); i++) 4004 zlfid->zf_setgen[i] = 0; 4005 } 4006 4007 ZFS_EXIT(zfsvfs); 4008 return (0); 4009} 4010 4011static int 4012zfs_pathconf(vnode_t *vp, int cmd, ulong_t *valp, cred_t *cr, 4013 caller_context_t *ct) 4014{ 4015 znode_t *zp, *xzp; 4016 zfsvfs_t *zfsvfs; 4017 zfs_dirlock_t *dl; 4018 int error; 4019 4020 switch (cmd) { 4021 case _PC_LINK_MAX: 4022 *valp = INT_MAX; 4023 return (0); 4024 4025 case _PC_FILESIZEBITS: 4026 *valp = 64; 4027 return (0); 4028 4029#if 0 4030 case _PC_XATTR_EXISTS: 4031 zp = VTOZ(vp); 4032 zfsvfs = zp->z_zfsvfs; 4033 ZFS_ENTER(zfsvfs); 4034 ZFS_VERIFY_ZP(zp); 4035 *valp = 0; 4036 error = zfs_dirent_lock(&dl, zp, "", &xzp, 4037 ZXATTR | ZEXISTS | ZSHARED, NULL, NULL); 4038 if (error == 0) { 4039 zfs_dirent_unlock(dl); 4040 if (!zfs_dirempty(xzp)) 4041 *valp = 1; 4042 VN_RELE(ZTOV(xzp)); 4043 } else if (error == ENOENT) { 4044 /* 4045 * If there aren't extended attributes, it's the 4046 * same as having zero of them. 4047 */ 4048 error = 0; 4049 } 4050 ZFS_EXIT(zfsvfs); 4051 return (error); 4052#endif 4053 4054 case _PC_ACL_EXTENDED: 4055 *valp = 0; 4056 return (0); 4057 4058 case _PC_ACL_NFS4: 4059 *valp = 1; 4060 return (0); 4061 4062 case _PC_ACL_PATH_MAX: 4063 *valp = ACL_MAX_ENTRIES; 4064 return (0); 4065 4066 case _PC_MIN_HOLE_SIZE: 4067 *valp = (int)SPA_MINBLOCKSIZE; 4068 return (0); 4069 4070 default: 4071 return (EOPNOTSUPP); 4072 } 4073} 4074 4075/*ARGSUSED*/ 4076static int 4077zfs_getsecattr(vnode_t *vp, vsecattr_t *vsecp, int flag, cred_t *cr, 4078 caller_context_t *ct) 4079{ 4080 znode_t *zp = VTOZ(vp); 4081 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 4082 int error; 4083 boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; 4084 4085 ZFS_ENTER(zfsvfs); 4086 ZFS_VERIFY_ZP(zp); 4087 error = zfs_getacl(zp, vsecp, skipaclchk, cr); 4088 ZFS_EXIT(zfsvfs); 4089 4090 return (error); 4091} 4092 4093/*ARGSUSED*/ 4094static int 4095zfs_setsecattr(vnode_t *vp, vsecattr_t *vsecp, int flag, cred_t *cr, 4096 caller_context_t *ct) 4097{ 4098 znode_t *zp = VTOZ(vp); 4099 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 4100 int error; 4101 boolean_t skipaclchk = (flag & ATTR_NOACLCHECK) ? B_TRUE : B_FALSE; 4102 4103 ZFS_ENTER(zfsvfs); 4104 ZFS_VERIFY_ZP(zp); 4105 error = zfs_setacl(zp, vsecp, skipaclchk, cr); 4106 ZFS_EXIT(zfsvfs); 4107 return (error); 4108} 4109 4110static int 4111zfs_freebsd_open(ap) 4112 struct vop_open_args /* { 4113 struct vnode *a_vp; 4114 int a_mode; 4115 struct ucred *a_cred; 4116 struct thread *a_td; 4117 } */ *ap; 4118{ 4119 vnode_t *vp = ap->a_vp; 4120 znode_t *zp = VTOZ(vp); 4121 int error; 4122 4123 error = zfs_open(&vp, ap->a_mode, ap->a_cred, NULL); 4124 if (error == 0) 4125 vnode_create_vobject(vp, zp->z_phys->zp_size, ap->a_td); 4126 return (error); 4127} 4128 4129static int 4130zfs_freebsd_close(ap) 4131 struct vop_close_args /* { 4132 struct vnode *a_vp; 4133 int a_fflag; 4134 struct ucred *a_cred; 4135 struct thread *a_td; 4136 } */ *ap; 4137{ 4138 4139 return (zfs_close(ap->a_vp, ap->a_fflag, 0, 0, ap->a_cred, NULL)); 4140} 4141 4142static int 4143zfs_freebsd_ioctl(ap) 4144 struct vop_ioctl_args /* { 4145 struct vnode *a_vp; 4146 u_long a_command; 4147 caddr_t a_data; 4148 int a_fflag; 4149 struct ucred *cred; 4150 struct thread *td; 4151 } */ *ap; 4152{ 4153 4154 return (zfs_ioctl(ap->a_vp, ap->a_command, (intptr_t)ap->a_data, 4155 ap->a_fflag, ap->a_cred, NULL, NULL)); 4156} 4157 4158static int 4159zfs_freebsd_read(ap) 4160 struct vop_read_args /* { 4161 struct vnode *a_vp; 4162 struct uio *a_uio; 4163 int a_ioflag; 4164 struct ucred *a_cred; 4165 } */ *ap; 4166{ 4167 4168 return (zfs_read(ap->a_vp, ap->a_uio, ap->a_ioflag, ap->a_cred, NULL)); 4169} 4170 4171static int 4172zfs_freebsd_write(ap) 4173 struct vop_write_args /* { 4174 struct vnode *a_vp; 4175 struct uio *a_uio; 4176 int a_ioflag; 4177 struct ucred *a_cred; 4178 } */ *ap; 4179{ 4180 4181 if (vn_rlimit_fsize(ap->a_vp, ap->a_uio, ap->a_uio->uio_td)) 4182 return (EFBIG); 4183 4184 return (zfs_write(ap->a_vp, ap->a_uio, ap->a_ioflag, ap->a_cred, NULL)); 4185} 4186 4187static int 4188zfs_freebsd_access(ap) 4189 struct vop_access_args /* { 4190 struct vnode *a_vp; 4191 accmode_t a_accmode; 4192 struct ucred *a_cred; 4193 struct thread *a_td; 4194 } */ *ap; 4195{ 4196 accmode_t accmode; 4197 int error = 0; 4198 4199 /* 4200 * ZFS itself only knowns about VREAD, VWRITE, VEXEC and VAPPEND, 4201 */ 4202 accmode = ap->a_accmode & (VREAD|VWRITE|VEXEC|VAPPEND); 4203 if (accmode != 0) 4204 error = zfs_access(ap->a_vp, accmode, 0, ap->a_cred, NULL); 4205 4206 /* 4207 * VADMIN has to be handled by vaccess(). 4208 */ 4209 if (error == 0) { 4210 accmode = ap->a_accmode & ~(VREAD|VWRITE|VEXEC|VAPPEND); 4211 if (accmode != 0) { 4212 vnode_t *vp = ap->a_vp; 4213 znode_t *zp = VTOZ(vp); 4214 znode_phys_t *zphys = zp->z_phys; 4215 4216 error = vaccess(vp->v_type, zphys->zp_mode, 4217 zphys->zp_uid, zphys->zp_gid, accmode, ap->a_cred, 4218 NULL); 4219 } 4220 } 4221 4222 return (error); 4223} 4224 4225static int 4226zfs_freebsd_lookup(ap) 4227 struct vop_lookup_args /* { 4228 struct vnode *a_dvp; 4229 struct vnode **a_vpp; 4230 struct componentname *a_cnp; 4231 } */ *ap; 4232{ 4233 struct componentname *cnp = ap->a_cnp; 4234 char nm[NAME_MAX + 1]; 4235 4236 ASSERT(cnp->cn_namelen < sizeof(nm)); 4237 strlcpy(nm, cnp->cn_nameptr, MIN(cnp->cn_namelen + 1, sizeof(nm))); 4238 4239 return (zfs_lookup(ap->a_dvp, nm, ap->a_vpp, cnp, cnp->cn_nameiop, 4240 cnp->cn_cred, cnp->cn_thread, 0)); 4241} 4242 4243static int 4244zfs_freebsd_create(ap) 4245 struct vop_create_args /* { 4246 struct vnode *a_dvp; 4247 struct vnode **a_vpp; 4248 struct componentname *a_cnp; 4249 struct vattr *a_vap; 4250 } */ *ap; 4251{ 4252 struct componentname *cnp = ap->a_cnp; 4253 vattr_t *vap = ap->a_vap; 4254 int mode; 4255 4256 ASSERT(cnp->cn_flags & SAVENAME); 4257 4258 vattr_init_mask(vap); 4259 mode = vap->va_mode & ALLPERMS; 4260 4261 return (zfs_create(ap->a_dvp, cnp->cn_nameptr, vap, !EXCL, mode, 4262 ap->a_vpp, cnp->cn_cred, cnp->cn_thread)); 4263} 4264 4265static int 4266zfs_freebsd_remove(ap) 4267 struct vop_remove_args /* { 4268 struct vnode *a_dvp; 4269 struct vnode *a_vp; 4270 struct componentname *a_cnp; 4271 } */ *ap; 4272{ 4273 4274 ASSERT(ap->a_cnp->cn_flags & SAVENAME); 4275 4276 return (zfs_remove(ap->a_dvp, ap->a_cnp->cn_nameptr, 4277 ap->a_cnp->cn_cred, NULL, 0)); 4278} 4279 4280static int 4281zfs_freebsd_mkdir(ap) 4282 struct vop_mkdir_args /* { 4283 struct vnode *a_dvp; 4284 struct vnode **a_vpp; 4285 struct componentname *a_cnp; 4286 struct vattr *a_vap; 4287 } */ *ap; 4288{ 4289 vattr_t *vap = ap->a_vap; 4290 4291 ASSERT(ap->a_cnp->cn_flags & SAVENAME); 4292 4293 vattr_init_mask(vap); 4294 4295 return (zfs_mkdir(ap->a_dvp, ap->a_cnp->cn_nameptr, vap, ap->a_vpp, 4296 ap->a_cnp->cn_cred, NULL, 0, NULL)); 4297} 4298 4299static int 4300zfs_freebsd_rmdir(ap) 4301 struct vop_rmdir_args /* { 4302 struct vnode *a_dvp; 4303 struct vnode *a_vp; 4304 struct componentname *a_cnp; 4305 } */ *ap; 4306{ 4307 struct componentname *cnp = ap->a_cnp; 4308 4309 ASSERT(cnp->cn_flags & SAVENAME); 4310 4311 return (zfs_rmdir(ap->a_dvp, cnp->cn_nameptr, NULL, cnp->cn_cred, NULL, 0)); 4312} 4313 4314static int 4315zfs_freebsd_readdir(ap) 4316 struct vop_readdir_args /* { 4317 struct vnode *a_vp; 4318 struct uio *a_uio; 4319 struct ucred *a_cred; 4320 int *a_eofflag; 4321 int *a_ncookies; 4322 u_long **a_cookies; 4323 } */ *ap; 4324{ 4325 4326 return (zfs_readdir(ap->a_vp, ap->a_uio, ap->a_cred, ap->a_eofflag, 4327 ap->a_ncookies, ap->a_cookies)); 4328} 4329 4330static int 4331zfs_freebsd_fsync(ap) 4332 struct vop_fsync_args /* { 4333 struct vnode *a_vp; 4334 int a_waitfor; 4335 struct thread *a_td; 4336 } */ *ap; 4337{ 4338 4339 vop_stdfsync(ap); 4340 return (zfs_fsync(ap->a_vp, 0, ap->a_td->td_ucred, NULL)); 4341} 4342 4343static int 4344zfs_freebsd_getattr(ap) 4345 struct vop_getattr_args /* { 4346 struct vnode *a_vp; 4347 struct vattr *a_vap; 4348 struct ucred *a_cred; 4349 struct thread *a_td; 4350 } */ *ap; 4351{ 4352 vattr_t *vap = ap->a_vap; 4353 xvattr_t xvap; 4354 u_long fflags = 0; 4355 int error; 4356 4357 xva_init(&xvap); 4358 xvap.xva_vattr = *vap; 4359 xvap.xva_vattr.va_mask |= AT_XVATTR; 4360 4361 /* Convert chflags into ZFS-type flags. */ 4362 /* XXX: what about SF_SETTABLE?. */ 4363 XVA_SET_REQ(&xvap, XAT_IMMUTABLE); 4364 XVA_SET_REQ(&xvap, XAT_APPENDONLY); 4365 XVA_SET_REQ(&xvap, XAT_NOUNLINK); 4366 XVA_SET_REQ(&xvap, XAT_NODUMP); 4367 error = zfs_getattr(ap->a_vp, (vattr_t *)&xvap, 0, ap->a_cred, NULL); 4368 if (error != 0) 4369 return (error); 4370 4371 /* Convert ZFS xattr into chflags. */ 4372#define FLAG_CHECK(fflag, xflag, xfield) do { \ 4373 if (XVA_ISSET_RTN(&xvap, (xflag)) && (xfield) != 0) \ 4374 fflags |= (fflag); \ 4375} while (0) 4376 FLAG_CHECK(SF_IMMUTABLE, XAT_IMMUTABLE, 4377 xvap.xva_xoptattrs.xoa_immutable); 4378 FLAG_CHECK(SF_APPEND, XAT_APPENDONLY, 4379 xvap.xva_xoptattrs.xoa_appendonly); 4380 FLAG_CHECK(SF_NOUNLINK, XAT_NOUNLINK, 4381 xvap.xva_xoptattrs.xoa_nounlink); 4382 FLAG_CHECK(UF_NODUMP, XAT_NODUMP, 4383 xvap.xva_xoptattrs.xoa_nodump); 4384#undef FLAG_CHECK 4385 *vap = xvap.xva_vattr; 4386 vap->va_flags = fflags; 4387 return (0); 4388} 4389 4390static int 4391zfs_freebsd_setattr(ap) 4392 struct vop_setattr_args /* { 4393 struct vnode *a_vp; 4394 struct vattr *a_vap; 4395 struct ucred *a_cred; 4396 struct thread *a_td; 4397 } */ *ap; 4398{ 4399 vnode_t *vp = ap->a_vp; 4400 vattr_t *vap = ap->a_vap; 4401 cred_t *cred = ap->a_cred; 4402 xvattr_t xvap; 4403 u_long fflags; 4404 uint64_t zflags; 4405 4406 vattr_init_mask(vap); 4407 vap->va_mask &= ~AT_NOSET; 4408 4409 xva_init(&xvap); 4410 xvap.xva_vattr = *vap; 4411 4412 zflags = VTOZ(vp)->z_phys->zp_flags; 4413 4414 if (vap->va_flags != VNOVAL) { 4415 zfsvfs_t *zfsvfs = VTOZ(vp)->z_zfsvfs; 4416 int error; 4417 4418 if (zfsvfs->z_use_fuids == B_FALSE) 4419 return (EOPNOTSUPP); 4420 4421 fflags = vap->va_flags; 4422 if ((fflags & ~(SF_IMMUTABLE|SF_APPEND|SF_NOUNLINK|UF_NODUMP)) != 0) 4423 return (EOPNOTSUPP); 4424 /* 4425 * Unprivileged processes are not permitted to unset system 4426 * flags, or modify flags if any system flags are set. 4427 * Privileged non-jail processes may not modify system flags 4428 * if securelevel > 0 and any existing system flags are set. 4429 * Privileged jail processes behave like privileged non-jail 4430 * processes if the security.jail.chflags_allowed sysctl is 4431 * is non-zero; otherwise, they behave like unprivileged 4432 * processes. 4433 */ 4434 if (secpolicy_fs_owner(vp->v_mount, cred) == 0 || 4435 priv_check_cred(cred, PRIV_VFS_SYSFLAGS, 0) == 0) { 4436 if (zflags & 4437 (ZFS_IMMUTABLE | ZFS_APPENDONLY | ZFS_NOUNLINK)) { 4438 error = securelevel_gt(cred, 0); 4439 if (error != 0) 4440 return (error); 4441 } 4442 } else { 4443 /* 4444 * Callers may only modify the file flags on objects they 4445 * have VADMIN rights for. 4446 */ 4447 if ((error = VOP_ACCESS(vp, VADMIN, cred, curthread)) != 0) 4448 return (error); 4449 if (zflags & 4450 (ZFS_IMMUTABLE | ZFS_APPENDONLY | ZFS_NOUNLINK)) { 4451 return (EPERM); 4452 } 4453 if (fflags & 4454 (SF_IMMUTABLE | SF_APPEND | SF_NOUNLINK)) { 4455 return (EPERM); 4456 } 4457 } 4458 4459#define FLAG_CHANGE(fflag, zflag, xflag, xfield) do { \ 4460 if (((fflags & (fflag)) && !(zflags & (zflag))) || \ 4461 ((zflags & (zflag)) && !(fflags & (fflag)))) { \ 4462 XVA_SET_REQ(&xvap, (xflag)); \ 4463 (xfield) = ((fflags & (fflag)) != 0); \ 4464 } \ 4465} while (0) 4466 /* Convert chflags into ZFS-type flags. */ 4467 /* XXX: what about SF_SETTABLE?. */ 4468 FLAG_CHANGE(SF_IMMUTABLE, ZFS_IMMUTABLE, XAT_IMMUTABLE, 4469 xvap.xva_xoptattrs.xoa_immutable); 4470 FLAG_CHANGE(SF_APPEND, ZFS_APPENDONLY, XAT_APPENDONLY, 4471 xvap.xva_xoptattrs.xoa_appendonly); 4472 FLAG_CHANGE(SF_NOUNLINK, ZFS_NOUNLINK, XAT_NOUNLINK, 4473 xvap.xva_xoptattrs.xoa_nounlink); 4474 FLAG_CHANGE(UF_NODUMP, ZFS_NODUMP, XAT_NODUMP, 4475 xvap.xva_xoptattrs.xoa_nodump); 4476#undef FLAG_CHANGE 4477 } 4478 return (zfs_setattr(vp, (vattr_t *)&xvap, 0, cred, NULL)); 4479} 4480 4481static int 4482zfs_freebsd_rename(ap) 4483 struct vop_rename_args /* { 4484 struct vnode *a_fdvp; 4485 struct vnode *a_fvp; 4486 struct componentname *a_fcnp; 4487 struct vnode *a_tdvp; 4488 struct vnode *a_tvp; 4489 struct componentname *a_tcnp; 4490 } */ *ap; 4491{ 4492 vnode_t *fdvp = ap->a_fdvp; 4493 vnode_t *fvp = ap->a_fvp; 4494 vnode_t *tdvp = ap->a_tdvp; 4495 vnode_t *tvp = ap->a_tvp; 4496 int error; 4497 4498 ASSERT(ap->a_fcnp->cn_flags & (SAVENAME|SAVESTART)); 4499 ASSERT(ap->a_tcnp->cn_flags & (SAVENAME|SAVESTART)); 4500 4501 error = zfs_rename(fdvp, ap->a_fcnp->cn_nameptr, tdvp, 4502 ap->a_tcnp->cn_nameptr, ap->a_fcnp->cn_cred, NULL, 0); 4503 4504 if (tdvp == tvp) 4505 VN_RELE(tdvp); 4506 else 4507 VN_URELE(tdvp); 4508 if (tvp) 4509 VN_URELE(tvp); 4510 VN_RELE(fdvp); 4511 VN_RELE(fvp); 4512 4513 return (error); 4514} 4515 4516static int 4517zfs_freebsd_symlink(ap) 4518 struct vop_symlink_args /* { 4519 struct vnode *a_dvp; 4520 struct vnode **a_vpp; 4521 struct componentname *a_cnp; 4522 struct vattr *a_vap; 4523 char *a_target; 4524 } */ *ap; 4525{ 4526 struct componentname *cnp = ap->a_cnp; 4527 vattr_t *vap = ap->a_vap; 4528 4529 ASSERT(cnp->cn_flags & SAVENAME); 4530 4531 vap->va_type = VLNK; /* FreeBSD: Syscall only sets va_mode. */ 4532 vattr_init_mask(vap); 4533 4534 return (zfs_symlink(ap->a_dvp, ap->a_vpp, cnp->cn_nameptr, vap, 4535 ap->a_target, cnp->cn_cred, cnp->cn_thread)); 4536} 4537 4538static int 4539zfs_freebsd_readlink(ap) 4540 struct vop_readlink_args /* { 4541 struct vnode *a_vp; 4542 struct uio *a_uio; 4543 struct ucred *a_cred; 4544 } */ *ap; 4545{ 4546 4547 return (zfs_readlink(ap->a_vp, ap->a_uio, ap->a_cred, NULL)); 4548} 4549 4550static int 4551zfs_freebsd_link(ap) 4552 struct vop_link_args /* { 4553 struct vnode *a_tdvp; 4554 struct vnode *a_vp; 4555 struct componentname *a_cnp; 4556 } */ *ap; 4557{ 4558 struct componentname *cnp = ap->a_cnp; 4559 4560 ASSERT(cnp->cn_flags & SAVENAME); 4561 4562 return (zfs_link(ap->a_tdvp, ap->a_vp, cnp->cn_nameptr, cnp->cn_cred, NULL, 0)); 4563} 4564 4565static int 4566zfs_freebsd_inactive(ap) 4567 struct vop_inactive_args /* { 4568 struct vnode *a_vp; 4569 struct thread *a_td; 4570 } */ *ap; 4571{ 4572 vnode_t *vp = ap->a_vp; 4573 4574 zfs_inactive(vp, ap->a_td->td_ucred, NULL); 4575 return (0); 4576} 4577 4578static void 4579zfs_reclaim_complete(void *arg, int pending) 4580{ 4581 znode_t *zp = arg; 4582 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 4583 4584 rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER); 4585 if (zp->z_dbuf != NULL) { 4586 ZFS_OBJ_HOLD_ENTER(zfsvfs, zp->z_id); 4587 zfs_znode_dmu_fini(zp); 4588 ZFS_OBJ_HOLD_EXIT(zfsvfs, zp->z_id); 4589 } 4590 zfs_znode_free(zp); 4591 rw_exit(&zfsvfs->z_teardown_inactive_lock); 4592 /* 4593 * If the file system is being unmounted, there is a process waiting 4594 * for us, wake it up. 4595 */ 4596 if (zfsvfs->z_unmounted) 4597 wakeup_one(zfsvfs); 4598} 4599 4600static int 4601zfs_freebsd_reclaim(ap) 4602 struct vop_reclaim_args /* { 4603 struct vnode *a_vp; 4604 struct thread *a_td; 4605 } */ *ap; 4606{ 4607 vnode_t *vp = ap->a_vp; 4608 znode_t *zp = VTOZ(vp); 4609 zfsvfs_t *zfsvfs = zp->z_zfsvfs; 4610 4611 rw_enter(&zfsvfs->z_teardown_inactive_lock, RW_READER); 4612 4613 ASSERT(zp != NULL); 4614 4615 /* 4616 * Destroy the vm object and flush associated pages. 4617 */ 4618 vnode_destroy_vobject(vp); 4619 4620 mutex_enter(&zp->z_lock); 4621 ASSERT(zp->z_phys != NULL); 4622 zp->z_vnode = NULL; 4623 mutex_exit(&zp->z_lock); 4624 4625 if (zp->z_unlinked) 4626 ; /* Do nothing. */ 4627 else if (zp->z_dbuf == NULL) 4628 zfs_znode_free(zp); 4629 else /* if (!zp->z_unlinked && zp->z_dbuf != NULL) */ { 4630 int locked; 4631 4632 locked = MUTEX_HELD(ZFS_OBJ_MUTEX(zfsvfs, zp->z_id)) ? 2 : 4633 ZFS_OBJ_HOLD_TRYENTER(zfsvfs, zp->z_id); 4634 if (locked == 0) { 4635 /* 4636 * Lock can't be obtained due to deadlock possibility, 4637 * so defer znode destruction. 4638 */ 4639 TASK_INIT(&zp->z_task, 0, zfs_reclaim_complete, zp); 4640 taskqueue_enqueue(taskqueue_thread, &zp->z_task); 4641 } else { 4642 zfs_znode_dmu_fini(zp); 4643 if (locked == 1) 4644 ZFS_OBJ_HOLD_EXIT(zfsvfs, zp->z_id); 4645 zfs_znode_free(zp); 4646 } 4647 } 4648 VI_LOCK(vp); 4649 vp->v_data = NULL; 4650 ASSERT(vp->v_holdcnt >= 1); 4651 VI_UNLOCK(vp); 4652 rw_exit(&zfsvfs->z_teardown_inactive_lock); 4653 return (0); 4654} 4655 4656static int 4657zfs_freebsd_fid(ap) 4658 struct vop_fid_args /* { 4659 struct vnode *a_vp; 4660 struct fid *a_fid; 4661 } */ *ap; 4662{ 4663 4664 return (zfs_fid(ap->a_vp, (void *)ap->a_fid, NULL)); 4665} 4666 4667static int 4668zfs_freebsd_pathconf(ap) 4669 struct vop_pathconf_args /* { 4670 struct vnode *a_vp; 4671 int a_name; 4672 register_t *a_retval; 4673 } */ *ap; 4674{ 4675 ulong_t val; 4676 int error; 4677 4678 error = zfs_pathconf(ap->a_vp, ap->a_name, &val, curthread->td_ucred, NULL); 4679 if (error == 0) 4680 *ap->a_retval = val; 4681 else if (error == EOPNOTSUPP) 4682 error = vop_stdpathconf(ap); 4683 return (error); 4684} 4685 4686static int 4687zfs_freebsd_fifo_pathconf(ap) 4688 struct vop_pathconf_args /* { 4689 struct vnode *a_vp; 4690 int a_name; 4691 register_t *a_retval; 4692 } */ *ap; 4693{ 4694 4695 switch (ap->a_name) { 4696 case _PC_ACL_EXTENDED: 4697 case _PC_ACL_NFS4: 4698 case _PC_ACL_PATH_MAX: 4699 case _PC_MAC_PRESENT: 4700 return (zfs_freebsd_pathconf(ap)); 4701 default: 4702 return (fifo_specops.vop_pathconf(ap)); 4703 } 4704} 4705 4706/* 4707 * FreeBSD's extended attributes namespace defines file name prefix for ZFS' 4708 * extended attribute name: 4709 * 4710 * NAMESPACE PREFIX 4711 * system freebsd:system: 4712 * user (none, can be used to access ZFS fsattr(5) attributes 4713 * created on Solaris) 4714 */ 4715static int 4716zfs_create_attrname(int attrnamespace, const char *name, char *attrname, 4717 size_t size) 4718{ 4719 const char *namespace, *prefix, *suffix; 4720 4721 /* We don't allow '/' character in attribute name. */ 4722 if (strchr(name, '/') != NULL) 4723 return (EINVAL); 4724 /* We don't allow attribute names that start with "freebsd:" string. */ 4725 if (strncmp(name, "freebsd:", 8) == 0) 4726 return (EINVAL); 4727 4728 bzero(attrname, size); 4729 4730 switch (attrnamespace) { 4731 case EXTATTR_NAMESPACE_USER: 4732#if 0 4733 prefix = "freebsd:"; 4734 namespace = EXTATTR_NAMESPACE_USER_STRING; 4735 suffix = ":"; 4736#else 4737 /* 4738 * This is the default namespace by which we can access all 4739 * attributes created on Solaris. 4740 */ 4741 prefix = namespace = suffix = ""; 4742#endif 4743 break; 4744 case EXTATTR_NAMESPACE_SYSTEM: 4745 prefix = "freebsd:"; 4746 namespace = EXTATTR_NAMESPACE_SYSTEM_STRING; 4747 suffix = ":"; 4748 break; 4749 case EXTATTR_NAMESPACE_EMPTY: 4750 default: 4751 return (EINVAL); 4752 } 4753 if (snprintf(attrname, size, "%s%s%s%s", prefix, namespace, suffix, 4754 name) >= size) { 4755 return (ENAMETOOLONG); 4756 } 4757 return (0); 4758} 4759 4760/* 4761 * Vnode operating to retrieve a named extended attribute. 4762 */ 4763static int 4764zfs_getextattr(struct vop_getextattr_args *ap) 4765/* 4766vop_getextattr { 4767 IN struct vnode *a_vp; 4768 IN int a_attrnamespace; 4769 IN const char *a_name; 4770 INOUT struct uio *a_uio; 4771 OUT size_t *a_size; 4772 IN struct ucred *a_cred; 4773 IN struct thread *a_td; 4774}; 4775*/ 4776{ 4777 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs; 4778 struct thread *td = ap->a_td; 4779 struct nameidata nd; 4780 char attrname[255]; 4781 struct vattr va; 4782 vnode_t *xvp = NULL, *vp; 4783 int error, flags; 4784 4785 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace, 4786 ap->a_cred, ap->a_td, VREAD); 4787 if (error != 0) 4788 return (error); 4789 4790 error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname, 4791 sizeof(attrname)); 4792 if (error != 0) 4793 return (error); 4794 4795 ZFS_ENTER(zfsvfs); 4796 4797 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td, 4798 LOOKUP_XATTR); 4799 if (error != 0) { 4800 ZFS_EXIT(zfsvfs); 4801 return (error); 4802 } 4803 4804 flags = FREAD; 4805 NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW | MPSAFE, UIO_SYSSPACE, attrname, 4806 xvp, td); 4807 error = vn_open_cred(&nd, &flags, 0, 0, ap->a_cred, NULL); 4808 vp = nd.ni_vp; 4809 NDFREE(&nd, NDF_ONLY_PNBUF); 4810 if (error != 0) { 4811 ZFS_EXIT(zfsvfs); 4812 if (error == ENOENT) 4813 error = ENOATTR; 4814 return (error); 4815 } 4816 4817 if (ap->a_size != NULL) { 4818 error = VOP_GETATTR(vp, &va, ap->a_cred); 4819 if (error == 0) 4820 *ap->a_size = (size_t)va.va_size; 4821 } else if (ap->a_uio != NULL) 4822 error = VOP_READ(vp, ap->a_uio, IO_UNIT | IO_SYNC, ap->a_cred); 4823 4824 VOP_UNLOCK(vp, 0); 4825 vn_close(vp, flags, ap->a_cred, td); 4826 ZFS_EXIT(zfsvfs); 4827 4828 return (error); 4829} 4830 4831/* 4832 * Vnode operation to remove a named attribute. 4833 */ 4834int 4835zfs_deleteextattr(struct vop_deleteextattr_args *ap) 4836/* 4837vop_deleteextattr { 4838 IN struct vnode *a_vp; 4839 IN int a_attrnamespace; 4840 IN const char *a_name; 4841 IN struct ucred *a_cred; 4842 IN struct thread *a_td; 4843}; 4844*/ 4845{ 4846 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs; 4847 struct thread *td = ap->a_td; 4848 struct nameidata nd; 4849 char attrname[255]; 4850 struct vattr va; 4851 vnode_t *xvp = NULL, *vp; 4852 int error, flags; 4853 4854 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace, 4855 ap->a_cred, ap->a_td, VWRITE); 4856 if (error != 0) 4857 return (error); 4858 4859 error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname, 4860 sizeof(attrname)); 4861 if (error != 0) 4862 return (error); 4863 4864 ZFS_ENTER(zfsvfs); 4865 4866 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td, 4867 LOOKUP_XATTR); 4868 if (error != 0) { 4869 ZFS_EXIT(zfsvfs); 4870 return (error); 4871 } 4872 4873 NDINIT_ATVP(&nd, DELETE, NOFOLLOW | LOCKPARENT | LOCKLEAF | MPSAFE, 4874 UIO_SYSSPACE, attrname, xvp, td); 4875 error = namei(&nd); 4876 vp = nd.ni_vp; 4877 NDFREE(&nd, NDF_ONLY_PNBUF); 4878 if (error != 0) { 4879 ZFS_EXIT(zfsvfs); 4880 if (error == ENOENT) 4881 error = ENOATTR; 4882 return (error); 4883 } 4884 error = VOP_REMOVE(nd.ni_dvp, vp, &nd.ni_cnd); 4885 4886 vput(nd.ni_dvp); 4887 if (vp == nd.ni_dvp) 4888 vrele(vp); 4889 else 4890 vput(vp); 4891 ZFS_EXIT(zfsvfs); 4892 4893 return (error); 4894} 4895 4896/* 4897 * Vnode operation to set a named attribute. 4898 */ 4899static int 4900zfs_setextattr(struct vop_setextattr_args *ap) 4901/* 4902vop_setextattr { 4903 IN struct vnode *a_vp; 4904 IN int a_attrnamespace; 4905 IN const char *a_name; 4906 INOUT struct uio *a_uio; 4907 IN struct ucred *a_cred; 4908 IN struct thread *a_td; 4909}; 4910*/ 4911{ 4912 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs; 4913 struct thread *td = ap->a_td; 4914 struct nameidata nd; 4915 char attrname[255]; 4916 struct vattr va; 4917 vnode_t *xvp = NULL, *vp; 4918 int error, flags; 4919 4920 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace, 4921 ap->a_cred, ap->a_td, VWRITE); 4922 if (error != 0) 4923 return (error); 4924 4925 error = zfs_create_attrname(ap->a_attrnamespace, ap->a_name, attrname, 4926 sizeof(attrname)); 4927 if (error != 0) 4928 return (error); 4929 4930 ZFS_ENTER(zfsvfs); 4931 4932 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td, 4933 LOOKUP_XATTR | CREATE_XATTR_DIR); 4934 if (error != 0) { 4935 ZFS_EXIT(zfsvfs); 4936 return (error); 4937 } 4938 4939 flags = FFLAGS(O_WRONLY | O_CREAT); 4940 NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW | MPSAFE, UIO_SYSSPACE, attrname, 4941 xvp, td); 4942 error = vn_open_cred(&nd, &flags, 0600, 0, ap->a_cred, NULL); 4943 vp = nd.ni_vp; 4944 NDFREE(&nd, NDF_ONLY_PNBUF); 4945 if (error != 0) { 4946 ZFS_EXIT(zfsvfs); 4947 return (error); 4948 } 4949 4950 VATTR_NULL(&va); 4951 va.va_size = 0; 4952 error = VOP_SETATTR(vp, &va, ap->a_cred); 4953 if (error == 0) 4954 VOP_WRITE(vp, ap->a_uio, IO_UNIT | IO_SYNC, ap->a_cred); 4955 4956 VOP_UNLOCK(vp, 0); 4957 vn_close(vp, flags, ap->a_cred, td); 4958 ZFS_EXIT(zfsvfs); 4959 4960 return (error); 4961} 4962 4963/* 4964 * Vnode operation to retrieve extended attributes on a vnode. 4965 */ 4966static int 4967zfs_listextattr(struct vop_listextattr_args *ap) 4968/* 4969vop_listextattr { 4970 IN struct vnode *a_vp; 4971 IN int a_attrnamespace; 4972 INOUT struct uio *a_uio; 4973 OUT size_t *a_size; 4974 IN struct ucred *a_cred; 4975 IN struct thread *a_td; 4976}; 4977*/ 4978{ 4979 zfsvfs_t *zfsvfs = VTOZ(ap->a_vp)->z_zfsvfs; 4980 struct thread *td = ap->a_td; 4981 struct nameidata nd; 4982 char attrprefix[16]; 4983 u_char dirbuf[sizeof(struct dirent)]; 4984 struct dirent *dp; 4985 struct iovec aiov; 4986 struct uio auio, *uio = ap->a_uio; 4987 size_t *sizep = ap->a_size; 4988 size_t plen; 4989 vnode_t *xvp = NULL, *vp; 4990 int done, error, eof, pos; 4991 4992 error = extattr_check_cred(ap->a_vp, ap->a_attrnamespace, 4993 ap->a_cred, ap->a_td, VREAD); 4994 if (error != 0) 4995 return (error); 4996 4997 error = zfs_create_attrname(ap->a_attrnamespace, "", attrprefix, 4998 sizeof(attrprefix)); 4999 if (error != 0) 5000 return (error); 5001 plen = strlen(attrprefix); 5002 5003 ZFS_ENTER(zfsvfs); 5004 5005 if (sizep != NULL) 5006 *sizep = 0; 5007 5008 error = zfs_lookup(ap->a_vp, NULL, &xvp, NULL, 0, ap->a_cred, td, 5009 LOOKUP_XATTR); 5010 if (error != 0) { 5011 ZFS_EXIT(zfsvfs); 5012 /* 5013 * ENOATTR means that the EA directory does not yet exist, 5014 * i.e. there are no extended attributes there. 5015 */ 5016 if (error == ENOATTR) 5017 error = 0; 5018 return (error); 5019 } 5020 5021 NDINIT_ATVP(&nd, LOOKUP, NOFOLLOW | LOCKLEAF | LOCKSHARED | MPSAFE, 5022 UIO_SYSSPACE, ".", xvp, td); 5023 error = namei(&nd); 5024 vp = nd.ni_vp; 5025 NDFREE(&nd, NDF_ONLY_PNBUF); 5026 if (error != 0) { 5027 ZFS_EXIT(zfsvfs); 5028 return (error); 5029 } 5030 5031 auio.uio_iov = &aiov; 5032 auio.uio_iovcnt = 1; 5033 auio.uio_segflg = UIO_SYSSPACE; 5034 auio.uio_td = td; 5035 auio.uio_rw = UIO_READ; 5036 auio.uio_offset = 0; 5037 5038 do { 5039 u_char nlen; 5040 5041 aiov.iov_base = (void *)dirbuf; 5042 aiov.iov_len = sizeof(dirbuf); 5043 auio.uio_resid = sizeof(dirbuf); 5044 error = VOP_READDIR(vp, &auio, ap->a_cred, &eof, NULL, NULL); 5045 done = sizeof(dirbuf) - auio.uio_resid; 5046 if (error != 0) 5047 break; 5048 for (pos = 0; pos < done;) { 5049 dp = (struct dirent *)(dirbuf + pos); 5050 pos += dp->d_reclen; 5051 /* 5052 * XXX: Temporarily we also accept DT_UNKNOWN, as this 5053 * is what we get when attribute was created on Solaris. 5054 */ 5055 if (dp->d_type != DT_REG && dp->d_type != DT_UNKNOWN) 5056 continue; 5057 if (plen == 0 && strncmp(dp->d_name, "freebsd:", 8) == 0) 5058 continue; 5059 else if (strncmp(dp->d_name, attrprefix, plen) != 0) 5060 continue; 5061 nlen = dp->d_namlen - plen; 5062 if (sizep != NULL) 5063 *sizep += 1 + nlen; 5064 else if (uio != NULL) { 5065 /* 5066 * Format of extattr name entry is one byte for 5067 * length and the rest for name. 5068 */ 5069 error = uiomove(&nlen, 1, uio->uio_rw, uio); 5070 if (error == 0) { 5071 error = uiomove(dp->d_name + plen, nlen, 5072 uio->uio_rw, uio); 5073 } 5074 if (error != 0) 5075 break; 5076 } 5077 } 5078 } while (!eof && error == 0); 5079 5080 vput(vp); 5081 ZFS_EXIT(zfsvfs); 5082 5083 return (error); 5084} 5085 5086int 5087zfs_freebsd_getacl(ap) 5088 struct vop_getacl_args /* { 5089 struct vnode *vp; 5090 acl_type_t type; 5091 struct acl *aclp; 5092 struct ucred *cred; 5093 struct thread *td; 5094 } */ *ap; 5095{ 5096 int error; 5097 vsecattr_t vsecattr; 5098 5099 if (ap->a_type != ACL_TYPE_NFS4) 5100 return (EINVAL); 5101 5102 vsecattr.vsa_mask = VSA_ACE | VSA_ACECNT; 5103 if (error = zfs_getsecattr(ap->a_vp, &vsecattr, 0, ap->a_cred, NULL)) 5104 return (error); 5105 5106 error = acl_from_aces(ap->a_aclp, vsecattr.vsa_aclentp, vsecattr.vsa_aclcnt); 5107 if (vsecattr.vsa_aclentp != NULL) 5108 kmem_free(vsecattr.vsa_aclentp, vsecattr.vsa_aclentsz); 5109 5110 return (error); 5111} 5112 5113int 5114zfs_freebsd_setacl(ap) 5115 struct vop_setacl_args /* { 5116 struct vnode *vp; 5117 acl_type_t type; 5118 struct acl *aclp; 5119 struct ucred *cred; 5120 struct thread *td; 5121 } */ *ap; 5122{ 5123 int error; 5124 vsecattr_t vsecattr; 5125 int aclbsize; /* size of acl list in bytes */ 5126 aclent_t *aaclp; 5127 5128 if (ap->a_type != ACL_TYPE_NFS4) 5129 return (EINVAL); 5130 5131 if (ap->a_aclp->acl_cnt < 1 || ap->a_aclp->acl_cnt > MAX_ACL_ENTRIES) 5132 return (EINVAL); 5133 5134 /* 5135 * With NFSv4 ACLs, chmod(2) may need to add additional entries, 5136 * splitting every entry into two and appending "canonical six" 5137 * entries at the end. Don't allow for setting an ACL that would 5138 * cause chmod(2) to run out of ACL entries. 5139 */ 5140 if (ap->a_aclp->acl_cnt * 2 + 6 > ACL_MAX_ENTRIES) 5141 return (ENOSPC); 5142 5143 error = acl_nfs4_check(ap->a_aclp, ap->a_vp->v_type == VDIR); 5144 if (error != 0) 5145 return (error); 5146 5147 vsecattr.vsa_mask = VSA_ACE; 5148 aclbsize = ap->a_aclp->acl_cnt * sizeof(ace_t); 5149 vsecattr.vsa_aclentp = kmem_alloc(aclbsize, KM_SLEEP); 5150 aaclp = vsecattr.vsa_aclentp; 5151 vsecattr.vsa_aclentsz = aclbsize; 5152 5153 aces_from_acl(vsecattr.vsa_aclentp, &vsecattr.vsa_aclcnt, ap->a_aclp); 5154 error = zfs_setsecattr(ap->a_vp, &vsecattr, 0, ap->a_cred, NULL); 5155 kmem_free(aaclp, aclbsize); 5156 5157 return (error); 5158} 5159 5160int 5161zfs_freebsd_aclcheck(ap) 5162 struct vop_aclcheck_args /* { 5163 struct vnode *vp; 5164 acl_type_t type; 5165 struct acl *aclp; 5166 struct ucred *cred; 5167 struct thread *td; 5168 } */ *ap; 5169{ 5170 5171 return (EOPNOTSUPP); 5172} 5173 5174struct vop_vector zfs_vnodeops; 5175struct vop_vector zfs_fifoops; 5176struct vop_vector zfs_shareops; 5177 5178struct vop_vector zfs_vnodeops = { 5179 .vop_default = &default_vnodeops, 5180 .vop_inactive = zfs_freebsd_inactive, 5181 .vop_reclaim = zfs_freebsd_reclaim, 5182 .vop_access = zfs_freebsd_access, 5183#ifdef FREEBSD_NAMECACHE 5184 .vop_lookup = vfs_cache_lookup, 5185 .vop_cachedlookup = zfs_freebsd_lookup, 5186#else 5187 .vop_lookup = zfs_freebsd_lookup, 5188#endif 5189 .vop_getattr = zfs_freebsd_getattr, 5190 .vop_setattr = zfs_freebsd_setattr, 5191 .vop_create = zfs_freebsd_create, 5192 .vop_mknod = zfs_freebsd_create, 5193 .vop_mkdir = zfs_freebsd_mkdir, 5194 .vop_readdir = zfs_freebsd_readdir, 5195 .vop_fsync = zfs_freebsd_fsync, 5196 .vop_open = zfs_freebsd_open, 5197 .vop_close = zfs_freebsd_close, 5198 .vop_rmdir = zfs_freebsd_rmdir, 5199 .vop_ioctl = zfs_freebsd_ioctl, 5200 .vop_link = zfs_freebsd_link, 5201 .vop_symlink = zfs_freebsd_symlink, 5202 .vop_readlink = zfs_freebsd_readlink, 5203 .vop_read = zfs_freebsd_read, 5204 .vop_write = zfs_freebsd_write, 5205 .vop_remove = zfs_freebsd_remove, 5206 .vop_rename = zfs_freebsd_rename, 5207 .vop_pathconf = zfs_freebsd_pathconf, 5208 .vop_bmap = VOP_EOPNOTSUPP, 5209 .vop_fid = zfs_freebsd_fid, 5210 .vop_getextattr = zfs_getextattr, 5211 .vop_deleteextattr = zfs_deleteextattr, 5212 .vop_setextattr = zfs_setextattr, 5213 .vop_listextattr = zfs_listextattr, 5214 .vop_getacl = zfs_freebsd_getacl, 5215 .vop_setacl = zfs_freebsd_setacl, 5216 .vop_aclcheck = zfs_freebsd_aclcheck, 5217}; 5218 5219struct vop_vector zfs_fifoops = { 5220 .vop_default = &fifo_specops, 5221 .vop_fsync = zfs_freebsd_fsync, 5222 .vop_access = zfs_freebsd_access, 5223 .vop_getattr = zfs_freebsd_getattr, 5224 .vop_inactive = zfs_freebsd_inactive, 5225 .vop_read = VOP_PANIC, 5226 .vop_reclaim = zfs_freebsd_reclaim, 5227 .vop_setattr = zfs_freebsd_setattr, 5228 .vop_write = VOP_PANIC, 5229 .vop_pathconf = zfs_freebsd_fifo_pathconf, 5230 .vop_fid = zfs_freebsd_fid, 5231 .vop_getacl = zfs_freebsd_getacl, 5232 .vop_setacl = zfs_freebsd_setacl, 5233 .vop_aclcheck = zfs_freebsd_aclcheck, 5234}; 5235 5236/* 5237 * special share hidden files vnode operations template 5238 */ 5239struct vop_vector zfs_shareops = { 5240 .vop_default = &default_vnodeops, 5241 .vop_access = zfs_freebsd_access, 5242 .vop_inactive = zfs_freebsd_inactive, 5243 .vop_reclaim = zfs_freebsd_reclaim, 5244 .vop_fid = zfs_freebsd_fid, 5245 .vop_pathconf = zfs_freebsd_pathconf, 5246}; 5247