1fa9e4066Sahrens /* 2fa9e4066Sahrens * CDDL HEADER START 3fa9e4066Sahrens * 4fa9e4066Sahrens * The contents of this file are subject to the terms of the 5736b9155Smarks * Common Development and Distribution License (the "License"). 6736b9155Smarks * You may not use this file except in compliance with the License. 7fa9e4066Sahrens * 8fa9e4066Sahrens * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9fa9e4066Sahrens * or http://www.opensolaris.org/os/licensing. 10fa9e4066Sahrens * See the License for the specific language governing permissions 11fa9e4066Sahrens * and limitations under the License. 12fa9e4066Sahrens * 13fa9e4066Sahrens * When distributing Covered Code, include this CDDL HEADER in each 14fa9e4066Sahrens * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15fa9e4066Sahrens * If applicable, add the following below this CDDL HEADER, with the 16fa9e4066Sahrens * fields enclosed by brackets "[]" replaced with your own identifying 17fa9e4066Sahrens * information: Portions Copyright [yyyy] [name of copyright owner] 18fa9e4066Sahrens * 19fa9e4066Sahrens * CDDL HEADER END 20fa9e4066Sahrens */ 21fa9e4066Sahrens /* 22d2443e76Smarks * Copyright 2006 Sun Microsystems, Inc. All rights reserved. 23fa9e4066Sahrens * Use is subject to license terms. 24fa9e4066Sahrens */ 25fa9e4066Sahrens 26fa9e4066Sahrens #pragma ident "%Z%%M% %I% %E% SMI" 27fa9e4066Sahrens 28fa9e4066Sahrens #include <sys/types.h> 29fa9e4066Sahrens #include <sys/param.h> 30fa9e4066Sahrens #include <sys/time.h> 31fa9e4066Sahrens #include <sys/systm.h> 32fa9e4066Sahrens #include <sys/sysmacros.h> 33fa9e4066Sahrens #include <sys/resource.h> 34fa9e4066Sahrens #include <sys/vfs.h> 35fa9e4066Sahrens #include <sys/vnode.h> 36fa9e4066Sahrens #include <sys/file.h> 37fa9e4066Sahrens #include <sys/stat.h> 38fa9e4066Sahrens #include <sys/kmem.h> 39fa9e4066Sahrens #include <sys/taskq.h> 40fa9e4066Sahrens #include <sys/uio.h> 41fa9e4066Sahrens #include <sys/vmsystm.h> 42fa9e4066Sahrens #include <sys/atomic.h> 43*44eda4d7Smaybee #include <sys/vm.h> 44fa9e4066Sahrens #include <vm/seg_vn.h> 45fa9e4066Sahrens #include <vm/pvn.h> 46fa9e4066Sahrens #include <vm/as.h> 47fa9e4066Sahrens #include <sys/mman.h> 48fa9e4066Sahrens #include <sys/pathname.h> 49fa9e4066Sahrens #include <sys/cmn_err.h> 50fa9e4066Sahrens #include <sys/errno.h> 51fa9e4066Sahrens #include <sys/unistd.h> 52fa9e4066Sahrens #include <sys/zfs_vfsops.h> 53fa9e4066Sahrens #include <sys/zfs_dir.h> 54fa9e4066Sahrens #include <sys/zfs_acl.h> 55fa9e4066Sahrens #include <sys/zfs_ioctl.h> 56fa9e4066Sahrens #include <sys/fs/zfs.h> 57fa9e4066Sahrens #include <sys/dmu.h> 58fa9e4066Sahrens #include <sys/spa.h> 59fa9e4066Sahrens #include <sys/txg.h> 60fa9e4066Sahrens #include <sys/dbuf.h> 61fa9e4066Sahrens #include <sys/zap.h> 62fa9e4066Sahrens #include <sys/dirent.h> 63fa9e4066Sahrens #include <sys/policy.h> 64fa9e4066Sahrens #include <sys/sunddi.h> 65fa9e4066Sahrens #include <sys/filio.h> 66fa9e4066Sahrens #include "fs/fs_subr.h" 67fa9e4066Sahrens #include <sys/zfs_ctldir.h> 68033f9833Sek #include <sys/dnlc.h> 69104e2ed7Sperrin #include <sys/zfs_rlock.h> 70fa9e4066Sahrens 71fa9e4066Sahrens /* 72fa9e4066Sahrens * Programming rules. 73fa9e4066Sahrens * 74fa9e4066Sahrens * Each vnode op performs some logical unit of work. To do this, the ZPL must 75fa9e4066Sahrens * properly lock its in-core state, create a DMU transaction, do the work, 76fa9e4066Sahrens * record this work in the intent log (ZIL), commit the DMU transaction, 77fa9e4066Sahrens * and wait the the intent log to commit if it's is a synchronous operation. 78fa9e4066Sahrens * Morover, the vnode ops must work in both normal and log replay context. 79fa9e4066Sahrens * The ordering of events is important to avoid deadlocks and references 80fa9e4066Sahrens * to freed memory. The example below illustrates the following Big Rules: 81fa9e4066Sahrens * 82fa9e4066Sahrens * (1) A check must be made in each zfs thread for a mounted file system. 83fa9e4066Sahrens * This is done avoiding races using ZFS_ENTER(zfsvfs). 84fa9e4066Sahrens * A ZFS_EXIT(zfsvfs) is needed before all returns. 85fa9e4066Sahrens * 86fa9e4066Sahrens * (2) VN_RELE() should always be the last thing except for zil_commit() 87b19a79ecSperrin * (if necessary) and ZFS_EXIT(). This is for 3 reasons: 88fa9e4066Sahrens * First, if it's the last reference, the vnode/znode 89fa9e4066Sahrens * can be freed, so the zp may point to freed memory. Second, the last 90fa9e4066Sahrens * reference will call zfs_zinactive(), which may induce a lot of work -- 91104e2ed7Sperrin * pushing cached pages (which acquires range locks) and syncing out 92fa9e4066Sahrens * cached atime changes. Third, zfs_zinactive() may require a new tx, 93fa9e4066Sahrens * which could deadlock the system if you were already holding one. 94fa9e4066Sahrens * 957885c754Sperrin * (3) All range locks must be grabbed before calling dmu_tx_assign(), 967885c754Sperrin * as they can span dmu_tx_assign() calls. 977885c754Sperrin * 987885c754Sperrin * (4) Always pass zfsvfs->z_assign as the second argument to dmu_tx_assign(). 99fa9e4066Sahrens * In normal operation, this will be TXG_NOWAIT. During ZIL replay, 100fa9e4066Sahrens * it will be a specific txg. Either way, dmu_tx_assign() never blocks. 101fa9e4066Sahrens * This is critical because we don't want to block while holding locks. 102fa9e4066Sahrens * Note, in particular, that if a lock is sometimes acquired before 103fa9e4066Sahrens * the tx assigns, and sometimes after (e.g. z_lock), then failing to 104fa9e4066Sahrens * use a non-blocking assign can deadlock the system. The scenario: 105fa9e4066Sahrens * 106fa9e4066Sahrens * Thread A has grabbed a lock before calling dmu_tx_assign(). 107fa9e4066Sahrens * Thread B is in an already-assigned tx, and blocks for this lock. 108fa9e4066Sahrens * Thread A calls dmu_tx_assign(TXG_WAIT) and blocks in txg_wait_open() 109fa9e4066Sahrens * forever, because the previous txg can't quiesce until B's tx commits. 110fa9e4066Sahrens * 111fa9e4066Sahrens * If dmu_tx_assign() returns ERESTART and zfsvfs->z_assign is TXG_NOWAIT, 1128a2f1b91Sahrens * then drop all locks, call dmu_tx_wait(), and try again. 113fa9e4066Sahrens * 1147885c754Sperrin * (5) If the operation succeeded, generate the intent log entry for it 115fa9e4066Sahrens * before dropping locks. This ensures that the ordering of events 116fa9e4066Sahrens * in the intent log matches the order in which they actually occurred. 117fa9e4066Sahrens * 1187885c754Sperrin * (6) At the end of each vnode op, the DMU tx must always commit, 119fa9e4066Sahrens * regardless of whether there were any errors. 120fa9e4066Sahrens * 121b19a79ecSperrin * (7) After dropping all locks, invoke zil_commit(zilog, seq, foid) 122fa9e4066Sahrens * to ensure that synchronous semantics are provided when necessary. 123fa9e4066Sahrens * 124fa9e4066Sahrens * In general, this is how things should be ordered in each vnode op: 125fa9e4066Sahrens * 126fa9e4066Sahrens * ZFS_ENTER(zfsvfs); // exit if unmounted 127fa9e4066Sahrens * top: 128fa9e4066Sahrens * zfs_dirent_lock(&dl, ...) // lock directory entry (may VN_HOLD()) 129fa9e4066Sahrens * rw_enter(...); // grab any other locks you need 130fa9e4066Sahrens * tx = dmu_tx_create(...); // get DMU tx 131fa9e4066Sahrens * dmu_tx_hold_*(); // hold each object you might modify 132fa9e4066Sahrens * error = dmu_tx_assign(tx, zfsvfs->z_assign); // try to assign 133fa9e4066Sahrens * if (error) { 134fa9e4066Sahrens * rw_exit(...); // drop locks 135fa9e4066Sahrens * zfs_dirent_unlock(dl); // unlock directory entry 136fa9e4066Sahrens * VN_RELE(...); // release held vnodes 137fa9e4066Sahrens * if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) { 1388a2f1b91Sahrens * dmu_tx_wait(tx); 1398a2f1b91Sahrens * dmu_tx_abort(tx); 140fa9e4066Sahrens * goto top; 141fa9e4066Sahrens * } 1428a2f1b91Sahrens * dmu_tx_abort(tx); // abort DMU tx 143fa9e4066Sahrens * ZFS_EXIT(zfsvfs); // finished in zfs 144fa9e4066Sahrens * return (error); // really out of space 145fa9e4066Sahrens * } 146fa9e4066Sahrens * error = do_real_work(); // do whatever this VOP does 147fa9e4066Sahrens * if (error == 0) 148b19a79ecSperrin * zfs_log_*(...); // on success, make ZIL entry 149fa9e4066Sahrens * dmu_tx_commit(tx); // commit DMU tx -- error or not 150fa9e4066Sahrens * rw_exit(...); // drop locks 151fa9e4066Sahrens * zfs_dirent_unlock(dl); // unlock directory entry 152fa9e4066Sahrens * VN_RELE(...); // release held vnodes 153b19a79ecSperrin * zil_commit(zilog, seq, foid); // synchronous when necessary 154fa9e4066Sahrens * ZFS_EXIT(zfsvfs); // finished in zfs 155fa9e4066Sahrens * return (error); // done, report error 156fa9e4066Sahrens */ 157fa9e4066Sahrens 158fa9e4066Sahrens /* ARGSUSED */ 159fa9e4066Sahrens static int 160fa9e4066Sahrens zfs_open(vnode_t **vpp, int flag, cred_t *cr) 161fa9e4066Sahrens { 162fa9e4066Sahrens return (0); 163fa9e4066Sahrens } 164fa9e4066Sahrens 165fa9e4066Sahrens /* ARGSUSED */ 166fa9e4066Sahrens static int 167fa9e4066Sahrens zfs_close(vnode_t *vp, int flag, int count, offset_t offset, cred_t *cr) 168fa9e4066Sahrens { 169fa9e4066Sahrens /* 170fa9e4066Sahrens * Clean up any locks held by this process on the vp. 171fa9e4066Sahrens */ 172fa9e4066Sahrens cleanlocks(vp, ddi_get_pid(), 0); 173fa9e4066Sahrens cleanshares(vp, ddi_get_pid()); 174fa9e4066Sahrens 175fa9e4066Sahrens return (0); 176fa9e4066Sahrens } 177fa9e4066Sahrens 178fa9e4066Sahrens /* 179fa9e4066Sahrens * Lseek support for finding holes (cmd == _FIO_SEEK_HOLE) and 180fa9e4066Sahrens * data (cmd == _FIO_SEEK_DATA). "off" is an in/out parameter. 181fa9e4066Sahrens */ 182fa9e4066Sahrens static int 183fa9e4066Sahrens zfs_holey(vnode_t *vp, int cmd, offset_t *off) 184fa9e4066Sahrens { 185fa9e4066Sahrens znode_t *zp = VTOZ(vp); 186fa9e4066Sahrens uint64_t noff = (uint64_t)*off; /* new offset */ 187fa9e4066Sahrens uint64_t file_sz; 188fa9e4066Sahrens int error; 189fa9e4066Sahrens boolean_t hole; 190fa9e4066Sahrens 191fa9e4066Sahrens file_sz = zp->z_phys->zp_size; 192fa9e4066Sahrens if (noff >= file_sz) { 193fa9e4066Sahrens return (ENXIO); 194fa9e4066Sahrens } 195fa9e4066Sahrens 196fa9e4066Sahrens if (cmd == _FIO_SEEK_HOLE) 197fa9e4066Sahrens hole = B_TRUE; 198fa9e4066Sahrens else 199fa9e4066Sahrens hole = B_FALSE; 200fa9e4066Sahrens 201fa9e4066Sahrens error = dmu_offset_next(zp->z_zfsvfs->z_os, zp->z_id, hole, &noff); 202fa9e4066Sahrens 203fa9e4066Sahrens /* end of file? */ 204fa9e4066Sahrens if ((error == ESRCH) || (noff > file_sz)) { 205fa9e4066Sahrens /* 206fa9e4066Sahrens * Handle the virtual hole at the end of file. 207fa9e4066Sahrens */ 208fa9e4066Sahrens if (hole) { 209fa9e4066Sahrens *off = file_sz; 210fa9e4066Sahrens return (0); 211fa9e4066Sahrens } 212fa9e4066Sahrens return (ENXIO); 213fa9e4066Sahrens } 214fa9e4066Sahrens 215fa9e4066Sahrens if (noff < *off) 216fa9e4066Sahrens return (error); 217fa9e4066Sahrens *off = noff; 218fa9e4066Sahrens return (error); 219fa9e4066Sahrens } 220fa9e4066Sahrens 221fa9e4066Sahrens /* ARGSUSED */ 222fa9e4066Sahrens static int 223fa9e4066Sahrens zfs_ioctl(vnode_t *vp, int com, intptr_t data, int flag, cred_t *cred, 224fa9e4066Sahrens int *rvalp) 225fa9e4066Sahrens { 226fa9e4066Sahrens offset_t off; 227fa9e4066Sahrens int error; 228fa9e4066Sahrens zfsvfs_t *zfsvfs; 229fa9e4066Sahrens 230fa9e4066Sahrens switch (com) { 231fa9e4066Sahrens case _FIOFFS: 232fa9e4066Sahrens return (zfs_sync(vp->v_vfsp, 0, cred)); 233fa9e4066Sahrens 234ea8dc4b6Seschrock /* 235ea8dc4b6Seschrock * The following two ioctls are used by bfu. Faking out, 236ea8dc4b6Seschrock * necessary to avoid bfu errors. 237ea8dc4b6Seschrock */ 238ea8dc4b6Seschrock case _FIOGDIO: 239ea8dc4b6Seschrock case _FIOSDIO: 240ea8dc4b6Seschrock return (0); 241ea8dc4b6Seschrock 242fa9e4066Sahrens case _FIO_SEEK_DATA: 243fa9e4066Sahrens case _FIO_SEEK_HOLE: 244fa9e4066Sahrens if (ddi_copyin((void *)data, &off, sizeof (off), flag)) 245fa9e4066Sahrens return (EFAULT); 246fa9e4066Sahrens 247fa9e4066Sahrens zfsvfs = VTOZ(vp)->z_zfsvfs; 248fa9e4066Sahrens ZFS_ENTER(zfsvfs); 249fa9e4066Sahrens 250fa9e4066Sahrens /* offset parameter is in/out */ 251fa9e4066Sahrens error = zfs_holey(vp, com, &off); 252fa9e4066Sahrens ZFS_EXIT(zfsvfs); 253fa9e4066Sahrens if (error) 254fa9e4066Sahrens return (error); 255fa9e4066Sahrens if (ddi_copyout(&off, (void *)data, sizeof (off), flag)) 256fa9e4066Sahrens return (EFAULT); 257fa9e4066Sahrens return (0); 258fa9e4066Sahrens } 259fa9e4066Sahrens return (ENOTTY); 260fa9e4066Sahrens } 261fa9e4066Sahrens 262fa9e4066Sahrens /* 263fa9e4066Sahrens * When a file is memory mapped, we must keep the IO data synchronized 264fa9e4066Sahrens * between the DMU cache and the memory mapped pages. What this means: 265fa9e4066Sahrens * 266fa9e4066Sahrens * On Write: If we find a memory mapped page, we write to *both* 267fa9e4066Sahrens * the page and the dmu buffer. 268fa9e4066Sahrens * 269fa9e4066Sahrens * NOTE: We will always "break up" the IO into PAGESIZE uiomoves when 270fa9e4066Sahrens * the file is memory mapped. 271fa9e4066Sahrens */ 272fa9e4066Sahrens static int 273fa9e4066Sahrens mappedwrite(vnode_t *vp, uint64_t woff, int nbytes, uio_t *uio, dmu_tx_t *tx) 274fa9e4066Sahrens { 275fa9e4066Sahrens znode_t *zp = VTOZ(vp); 276fa9e4066Sahrens zfsvfs_t *zfsvfs = zp->z_zfsvfs; 277fa9e4066Sahrens int64_t start, off; 278fa9e4066Sahrens int len = nbytes; 279fa9e4066Sahrens int error = 0; 280fa9e4066Sahrens 281fa9e4066Sahrens start = uio->uio_loffset; 282fa9e4066Sahrens off = start & PAGEOFFSET; 283fa9e4066Sahrens for (start &= PAGEMASK; len > 0; start += PAGESIZE) { 284fa9e4066Sahrens page_t *pp; 285fa9e4066Sahrens uint64_t bytes = MIN(PAGESIZE - off, len); 286fa9e4066Sahrens 287fa9e4066Sahrens /* 288fa9e4066Sahrens * We don't want a new page to "appear" in the middle of 289fa9e4066Sahrens * the file update (because it may not get the write 290fa9e4066Sahrens * update data), so we grab a lock to block 291fa9e4066Sahrens * zfs_getpage(). 292fa9e4066Sahrens */ 293fa9e4066Sahrens rw_enter(&zp->z_map_lock, RW_WRITER); 294fa9e4066Sahrens if (pp = page_lookup(vp, start, SE_SHARED)) { 295fa9e4066Sahrens caddr_t va; 296fa9e4066Sahrens 297fa9e4066Sahrens rw_exit(&zp->z_map_lock); 298fa9e4066Sahrens va = ppmapin(pp, PROT_READ | PROT_WRITE, (caddr_t)-1L); 299fa9e4066Sahrens error = uiomove(va+off, bytes, UIO_WRITE, uio); 300fa9e4066Sahrens if (error == 0) { 301fa9e4066Sahrens dmu_write(zfsvfs->z_os, zp->z_id, 302fa9e4066Sahrens woff, bytes, va+off, tx); 303fa9e4066Sahrens } 304fa9e4066Sahrens ppmapout(va); 305fa9e4066Sahrens page_unlock(pp); 306fa9e4066Sahrens } else { 307fa9e4066Sahrens error = dmu_write_uio(zfsvfs->z_os, zp->z_id, 308fa9e4066Sahrens woff, bytes, uio, tx); 309fa9e4066Sahrens rw_exit(&zp->z_map_lock); 310fa9e4066Sahrens } 311fa9e4066Sahrens len -= bytes; 312fa9e4066Sahrens woff += bytes; 313fa9e4066Sahrens off = 0; 314fa9e4066Sahrens if (error) 315fa9e4066Sahrens break; 316fa9e4066Sahrens } 317fa9e4066Sahrens return (error); 318fa9e4066Sahrens } 319fa9e4066Sahrens 320fa9e4066Sahrens /* 321fa9e4066Sahrens * When a file is memory mapped, we must keep the IO data synchronized 322fa9e4066Sahrens * between the DMU cache and the memory mapped pages. What this means: 323fa9e4066Sahrens * 324fa9e4066Sahrens * On Read: We "read" preferentially from memory mapped pages, 325fa9e4066Sahrens * else we default from the dmu buffer. 326fa9e4066Sahrens * 327fa9e4066Sahrens * NOTE: We will always "break up" the IO into PAGESIZE uiomoves when 328fa9e4066Sahrens * the file is memory mapped. 329fa9e4066Sahrens */ 330fa9e4066Sahrens static int 331fa9e4066Sahrens mappedread(vnode_t *vp, char *addr, int nbytes, uio_t *uio) 332fa9e4066Sahrens { 333fa9e4066Sahrens int64_t start, off, bytes; 334fa9e4066Sahrens int len = nbytes; 335fa9e4066Sahrens int error = 0; 336fa9e4066Sahrens 337fa9e4066Sahrens start = uio->uio_loffset; 338fa9e4066Sahrens off = start & PAGEOFFSET; 339fa9e4066Sahrens for (start &= PAGEMASK; len > 0; start += PAGESIZE) { 340fa9e4066Sahrens page_t *pp; 341fa9e4066Sahrens 342fa9e4066Sahrens bytes = MIN(PAGESIZE - off, len); 343fa9e4066Sahrens if (pp = page_lookup(vp, start, SE_SHARED)) { 344fa9e4066Sahrens caddr_t va; 345fa9e4066Sahrens 346*44eda4d7Smaybee va = ppmapin(pp, PROT_READ, (caddr_t)-1L); 347fa9e4066Sahrens error = uiomove(va + off, bytes, UIO_READ, uio); 348fa9e4066Sahrens ppmapout(va); 349fa9e4066Sahrens page_unlock(pp); 350fa9e4066Sahrens } else { 351fa9e4066Sahrens /* XXX use dmu_read here? */ 352fa9e4066Sahrens error = uiomove(addr, bytes, UIO_READ, uio); 353fa9e4066Sahrens } 354fa9e4066Sahrens len -= bytes; 355fa9e4066Sahrens addr += bytes; 356fa9e4066Sahrens off = 0; 357fa9e4066Sahrens if (error) 358fa9e4066Sahrens break; 359fa9e4066Sahrens } 360fa9e4066Sahrens return (error); 361fa9e4066Sahrens } 362fa9e4066Sahrens 363fa9e4066Sahrens uint_t zfs_read_chunk_size = 1024 * 1024; /* Tunable */ 364fa9e4066Sahrens 365fa9e4066Sahrens /* 366fa9e4066Sahrens * Read bytes from specified file into supplied buffer. 367fa9e4066Sahrens * 368fa9e4066Sahrens * IN: vp - vnode of file to be read from. 369fa9e4066Sahrens * uio - structure supplying read location, range info, 370fa9e4066Sahrens * and return buffer. 371fa9e4066Sahrens * ioflag - SYNC flags; used to provide FRSYNC semantics. 372fa9e4066Sahrens * cr - credentials of caller. 373fa9e4066Sahrens * 374fa9e4066Sahrens * OUT: uio - updated offset and range, buffer filled. 375fa9e4066Sahrens * 376fa9e4066Sahrens * RETURN: 0 if success 377fa9e4066Sahrens * error code if failure 378fa9e4066Sahrens * 379fa9e4066Sahrens * Side Effects: 380fa9e4066Sahrens * vp - atime updated if byte count > 0 381fa9e4066Sahrens */ 382fa9e4066Sahrens /* ARGSUSED */ 383fa9e4066Sahrens static int 384fa9e4066Sahrens zfs_read(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct) 385fa9e4066Sahrens { 386fa9e4066Sahrens znode_t *zp = VTOZ(vp); 387fa9e4066Sahrens zfsvfs_t *zfsvfs = zp->z_zfsvfs; 388fa9e4066Sahrens uint64_t delta; 389fa9e4066Sahrens ssize_t n, size, cnt, ndone; 390fa9e4066Sahrens int error, i, numbufs; 391fa9e4066Sahrens dmu_buf_t *dbp, **dbpp; 392104e2ed7Sperrin rl_t *rl; 393fa9e4066Sahrens 394fa9e4066Sahrens ZFS_ENTER(zfsvfs); 395fa9e4066Sahrens 396fa9e4066Sahrens /* 397fa9e4066Sahrens * Validate file offset 398fa9e4066Sahrens */ 399fa9e4066Sahrens if (uio->uio_loffset < (offset_t)0) { 400fa9e4066Sahrens ZFS_EXIT(zfsvfs); 401fa9e4066Sahrens return (EINVAL); 402fa9e4066Sahrens } 403fa9e4066Sahrens 404fa9e4066Sahrens /* 405fa9e4066Sahrens * Fasttrack empty reads 406fa9e4066Sahrens */ 407fa9e4066Sahrens if (uio->uio_resid == 0) { 408fa9e4066Sahrens ZFS_EXIT(zfsvfs); 409fa9e4066Sahrens return (0); 410fa9e4066Sahrens } 411fa9e4066Sahrens 412fa9e4066Sahrens /* 413104e2ed7Sperrin * Check for mandatory locks 414fa9e4066Sahrens */ 415fa9e4066Sahrens if (MANDMODE((mode_t)zp->z_phys->zp_mode)) { 416fa9e4066Sahrens if (error = chklock(vp, FREAD, 417fa9e4066Sahrens uio->uio_loffset, uio->uio_resid, uio->uio_fmode, ct)) { 418fa9e4066Sahrens ZFS_EXIT(zfsvfs); 419fa9e4066Sahrens return (error); 420fa9e4066Sahrens } 421fa9e4066Sahrens } 422fa9e4066Sahrens 423fa9e4066Sahrens /* 424fa9e4066Sahrens * If we're in FRSYNC mode, sync out this znode before reading it. 425fa9e4066Sahrens */ 426b19a79ecSperrin if (ioflag & FRSYNC) 427b19a79ecSperrin zil_commit(zfsvfs->z_log, zp->z_last_itx, zp->z_id); 428fa9e4066Sahrens 429fa9e4066Sahrens /* 430104e2ed7Sperrin * Lock the range against changes. 431fa9e4066Sahrens */ 432104e2ed7Sperrin rl = zfs_range_lock(zp, uio->uio_loffset, uio->uio_resid, RL_READER); 433104e2ed7Sperrin 434fa9e4066Sahrens /* 435fa9e4066Sahrens * If we are reading past end-of-file we can skip 436fa9e4066Sahrens * to the end; but we might still need to set atime. 437fa9e4066Sahrens */ 438fa9e4066Sahrens if (uio->uio_loffset >= zp->z_phys->zp_size) { 439fa9e4066Sahrens cnt = 0; 440fa9e4066Sahrens error = 0; 441fa9e4066Sahrens goto out; 442fa9e4066Sahrens } 443fa9e4066Sahrens 444fa9e4066Sahrens cnt = MIN(uio->uio_resid, zp->z_phys->zp_size - uio->uio_loffset); 445fa9e4066Sahrens 446fa9e4066Sahrens for (ndone = 0; ndone < cnt; ndone += zfs_read_chunk_size) { 447fa9e4066Sahrens ASSERT(uio->uio_loffset < zp->z_phys->zp_size); 448fa9e4066Sahrens n = MIN(zfs_read_chunk_size, 449fa9e4066Sahrens zp->z_phys->zp_size - uio->uio_loffset); 450fa9e4066Sahrens n = MIN(n, cnt); 45113506d1eSmaybee error = dmu_buf_hold_array_by_bonus(zp->z_dbuf, 452ea8dc4b6Seschrock uio->uio_loffset, n, TRUE, FTAG, &numbufs, &dbpp); 453ea8dc4b6Seschrock if (error) 454fa9e4066Sahrens goto out; 455fa9e4066Sahrens /* 456fa9e4066Sahrens * Compute the adjustment to align the dmu buffers 457fa9e4066Sahrens * with the uio buffer. 458fa9e4066Sahrens */ 459fa9e4066Sahrens delta = uio->uio_loffset - dbpp[0]->db_offset; 460fa9e4066Sahrens 461fa9e4066Sahrens for (i = 0; i < numbufs; i++) { 462fa9e4066Sahrens if (n < 0) 463fa9e4066Sahrens break; 464fa9e4066Sahrens dbp = dbpp[i]; 465fa9e4066Sahrens size = dbp->db_size - delta; 466fa9e4066Sahrens /* 467fa9e4066Sahrens * XXX -- this is correct, but may be suboptimal. 468fa9e4066Sahrens * If the pages are all clean, we don't need to 469fa9e4066Sahrens * go through mappedread(). Maybe the VMODSORT 470fa9e4066Sahrens * stuff can help us here. 471fa9e4066Sahrens */ 472fa9e4066Sahrens if (vn_has_cached_data(vp)) { 473fa9e4066Sahrens error = mappedread(vp, (caddr_t)dbp->db_data + 474fa9e4066Sahrens delta, (n < size ? n : size), uio); 475fa9e4066Sahrens } else { 476fa9e4066Sahrens error = uiomove((caddr_t)dbp->db_data + delta, 477fa9e4066Sahrens (n < size ? n : size), UIO_READ, uio); 478fa9e4066Sahrens } 479fa9e4066Sahrens if (error) { 480ea8dc4b6Seschrock dmu_buf_rele_array(dbpp, numbufs, FTAG); 481fa9e4066Sahrens goto out; 482fa9e4066Sahrens } 483fa9e4066Sahrens n -= dbp->db_size; 484fa9e4066Sahrens if (delta) { 485fa9e4066Sahrens n += delta; 486fa9e4066Sahrens delta = 0; 487fa9e4066Sahrens } 488fa9e4066Sahrens } 489ea8dc4b6Seschrock dmu_buf_rele_array(dbpp, numbufs, FTAG); 490fa9e4066Sahrens } 491fa9e4066Sahrens out: 492c5c6ffa0Smaybee zfs_range_unlock(rl); 493fa9e4066Sahrens 494fa9e4066Sahrens ZFS_ACCESSTIME_STAMP(zfsvfs, zp); 495fa9e4066Sahrens ZFS_EXIT(zfsvfs); 496fa9e4066Sahrens return (error); 497fa9e4066Sahrens } 498fa9e4066Sahrens 499fa9e4066Sahrens /* 500fa9e4066Sahrens * Fault in the pages of the first n bytes specified by the uio structure. 501fa9e4066Sahrens * 1 byte in each page is touched and the uio struct is unmodified. 502fa9e4066Sahrens * Any error will exit this routine as this is only a best 503fa9e4066Sahrens * attempt to get the pages resident. This is a copy of ufs_trans_touch(). 504fa9e4066Sahrens */ 505fa9e4066Sahrens static void 506fa9e4066Sahrens zfs_prefault_write(ssize_t n, struct uio *uio) 507fa9e4066Sahrens { 508fa9e4066Sahrens struct iovec *iov; 509fa9e4066Sahrens ulong_t cnt, incr; 510fa9e4066Sahrens caddr_t p; 511fa9e4066Sahrens uint8_t tmp; 512fa9e4066Sahrens 513fa9e4066Sahrens iov = uio->uio_iov; 514fa9e4066Sahrens 515fa9e4066Sahrens while (n) { 516fa9e4066Sahrens cnt = MIN(iov->iov_len, n); 517fa9e4066Sahrens if (cnt == 0) { 518fa9e4066Sahrens /* empty iov entry */ 519fa9e4066Sahrens iov++; 520fa9e4066Sahrens continue; 521fa9e4066Sahrens } 522fa9e4066Sahrens n -= cnt; 523fa9e4066Sahrens /* 524fa9e4066Sahrens * touch each page in this segment. 525fa9e4066Sahrens */ 526fa9e4066Sahrens p = iov->iov_base; 527fa9e4066Sahrens while (cnt) { 528fa9e4066Sahrens switch (uio->uio_segflg) { 529fa9e4066Sahrens case UIO_USERSPACE: 530fa9e4066Sahrens case UIO_USERISPACE: 531fa9e4066Sahrens if (fuword8(p, &tmp)) 532fa9e4066Sahrens return; 533fa9e4066Sahrens break; 534fa9e4066Sahrens case UIO_SYSSPACE: 535fa9e4066Sahrens if (kcopy(p, &tmp, 1)) 536fa9e4066Sahrens return; 537fa9e4066Sahrens break; 538fa9e4066Sahrens } 539fa9e4066Sahrens incr = MIN(cnt, PAGESIZE); 540fa9e4066Sahrens p += incr; 541fa9e4066Sahrens cnt -= incr; 542fa9e4066Sahrens } 543fa9e4066Sahrens /* 544fa9e4066Sahrens * touch the last byte in case it straddles a page. 545fa9e4066Sahrens */ 546fa9e4066Sahrens p--; 547fa9e4066Sahrens switch (uio->uio_segflg) { 548fa9e4066Sahrens case UIO_USERSPACE: 549fa9e4066Sahrens case UIO_USERISPACE: 550fa9e4066Sahrens if (fuword8(p, &tmp)) 551fa9e4066Sahrens return; 552fa9e4066Sahrens break; 553fa9e4066Sahrens case UIO_SYSSPACE: 554fa9e4066Sahrens if (kcopy(p, &tmp, 1)) 555fa9e4066Sahrens return; 556fa9e4066Sahrens break; 557fa9e4066Sahrens } 558fa9e4066Sahrens iov++; 559fa9e4066Sahrens } 560fa9e4066Sahrens } 561fa9e4066Sahrens 562fa9e4066Sahrens /* 563fa9e4066Sahrens * Write the bytes to a file. 564fa9e4066Sahrens * 565fa9e4066Sahrens * IN: vp - vnode of file to be written to. 566fa9e4066Sahrens * uio - structure supplying write location, range info, 567fa9e4066Sahrens * and data buffer. 568fa9e4066Sahrens * ioflag - FAPPEND flag set if in append mode. 569fa9e4066Sahrens * cr - credentials of caller. 570fa9e4066Sahrens * 571fa9e4066Sahrens * OUT: uio - updated offset and range. 572fa9e4066Sahrens * 573fa9e4066Sahrens * RETURN: 0 if success 574fa9e4066Sahrens * error code if failure 575fa9e4066Sahrens * 576fa9e4066Sahrens * Timestamps: 577fa9e4066Sahrens * vp - ctime|mtime updated if byte count > 0 578fa9e4066Sahrens */ 579fa9e4066Sahrens /* ARGSUSED */ 580fa9e4066Sahrens static int 581fa9e4066Sahrens zfs_write(vnode_t *vp, uio_t *uio, int ioflag, cred_t *cr, caller_context_t *ct) 582fa9e4066Sahrens { 583fa9e4066Sahrens znode_t *zp = VTOZ(vp); 584fa9e4066Sahrens rlim64_t limit = uio->uio_llimit; 585fa9e4066Sahrens ssize_t start_resid = uio->uio_resid; 586fa9e4066Sahrens ssize_t tx_bytes; 587fa9e4066Sahrens uint64_t end_size; 588fa9e4066Sahrens dmu_tx_t *tx; 589fa9e4066Sahrens zfsvfs_t *zfsvfs = zp->z_zfsvfs; 590fa9e4066Sahrens zilog_t *zilog = zfsvfs->z_log; 591fa9e4066Sahrens offset_t woff; 592fa9e4066Sahrens ssize_t n, nbytes; 593104e2ed7Sperrin rl_t *rl; 594fa9e4066Sahrens int max_blksz = zfsvfs->z_max_blksz; 595104e2ed7Sperrin int error; 596fa9e4066Sahrens 597fa9e4066Sahrens /* 598fa9e4066Sahrens * Fasttrack empty write 599fa9e4066Sahrens */ 600104e2ed7Sperrin n = start_resid; 601fa9e4066Sahrens if (n == 0) 602fa9e4066Sahrens return (0); 603fa9e4066Sahrens 604104e2ed7Sperrin if (limit == RLIM64_INFINITY || limit > MAXOFFSET_T) 605104e2ed7Sperrin limit = MAXOFFSET_T; 606104e2ed7Sperrin 607fa9e4066Sahrens ZFS_ENTER(zfsvfs); 608fa9e4066Sahrens 609fa9e4066Sahrens /* 610c5c6ffa0Smaybee * Pre-fault the pages to ensure slow (eg NFS) pages 611104e2ed7Sperrin * don't hold up txg. 612fa9e4066Sahrens */ 613c5c6ffa0Smaybee zfs_prefault_write(n, uio); 614fa9e4066Sahrens 615fa9e4066Sahrens /* 616fa9e4066Sahrens * If in append mode, set the io offset pointer to eof. 617fa9e4066Sahrens */ 618104e2ed7Sperrin if (ioflag & FAPPEND) { 619104e2ed7Sperrin /* 620104e2ed7Sperrin * Range lock for a file append: 621104e2ed7Sperrin * The value for the start of range will be determined by 622104e2ed7Sperrin * zfs_range_lock() (to guarantee append semantics). 623104e2ed7Sperrin * If this write will cause the block size to increase, 624104e2ed7Sperrin * zfs_range_lock() will lock the entire file, so we must 625104e2ed7Sperrin * later reduce the range after we grow the block size. 626104e2ed7Sperrin */ 627104e2ed7Sperrin rl = zfs_range_lock(zp, 0, n, RL_APPEND); 628104e2ed7Sperrin if (rl->r_len == UINT64_MAX) { 629104e2ed7Sperrin /* overlocked, zp_size can't change */ 630104e2ed7Sperrin woff = uio->uio_loffset = zp->z_phys->zp_size; 631104e2ed7Sperrin } else { 632104e2ed7Sperrin woff = uio->uio_loffset = rl->r_off; 633104e2ed7Sperrin } 634fa9e4066Sahrens } else { 635fa9e4066Sahrens woff = uio->uio_loffset; 636fa9e4066Sahrens /* 637fa9e4066Sahrens * Validate file offset 638fa9e4066Sahrens */ 639fa9e4066Sahrens if (woff < 0) { 640fa9e4066Sahrens ZFS_EXIT(zfsvfs); 641fa9e4066Sahrens return (EINVAL); 642fa9e4066Sahrens } 643fa9e4066Sahrens 644fa9e4066Sahrens /* 645104e2ed7Sperrin * If we need to grow the block size then zfs_range_lock() 646104e2ed7Sperrin * will lock a wider range than we request here. 647104e2ed7Sperrin * Later after growing the block size we reduce the range. 648fa9e4066Sahrens */ 649104e2ed7Sperrin rl = zfs_range_lock(zp, woff, n, RL_WRITER); 650fa9e4066Sahrens } 651fa9e4066Sahrens 652fa9e4066Sahrens if (woff >= limit) { 653fa9e4066Sahrens error = EFBIG; 654fa9e4066Sahrens goto no_tx_done; 655fa9e4066Sahrens } 656fa9e4066Sahrens 657fa9e4066Sahrens if ((woff + n) > limit || woff > (limit - n)) 658fa9e4066Sahrens n = limit - woff; 659fa9e4066Sahrens 660fa9e4066Sahrens /* 661104e2ed7Sperrin * Check for mandatory locks 662fa9e4066Sahrens */ 663fa9e4066Sahrens if (MANDMODE((mode_t)zp->z_phys->zp_mode) && 664fa9e4066Sahrens (error = chklock(vp, FWRITE, woff, n, uio->uio_fmode, ct)) != 0) 665fa9e4066Sahrens goto no_tx_done; 666fa9e4066Sahrens end_size = MAX(zp->z_phys->zp_size, woff + n); 667104e2ed7Sperrin top: 668fa9e4066Sahrens tx = dmu_tx_create(zfsvfs->z_os); 669fa9e4066Sahrens dmu_tx_hold_bonus(tx, zp->z_id); 670fa9e4066Sahrens dmu_tx_hold_write(tx, zp->z_id, woff, MIN(n, max_blksz)); 671fa9e4066Sahrens error = dmu_tx_assign(tx, zfsvfs->z_assign); 672fa9e4066Sahrens if (error) { 673fa9e4066Sahrens if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) { 6748a2f1b91Sahrens dmu_tx_wait(tx); 6758a2f1b91Sahrens dmu_tx_abort(tx); 676fa9e4066Sahrens goto top; 677fa9e4066Sahrens } 6788a2f1b91Sahrens dmu_tx_abort(tx); 679fa9e4066Sahrens goto no_tx_done; 680fa9e4066Sahrens } 681fa9e4066Sahrens 682104e2ed7Sperrin /* 683104e2ed7Sperrin * If zfs_range_lock() over-locked we grow the blocksize 684104e2ed7Sperrin * and then reduce the lock range. 685104e2ed7Sperrin */ 686104e2ed7Sperrin if (rl->r_len == UINT64_MAX) { 687fa9e4066Sahrens uint64_t new_blksz; 688104e2ed7Sperrin 689fa9e4066Sahrens if (zp->z_blksz > max_blksz) { 690fa9e4066Sahrens ASSERT(!ISP2(zp->z_blksz)); 691fa9e4066Sahrens new_blksz = MIN(end_size, SPA_MAXBLOCKSIZE); 692fa9e4066Sahrens } else { 693fa9e4066Sahrens new_blksz = MIN(end_size, max_blksz); 694fa9e4066Sahrens } 695104e2ed7Sperrin zfs_grow_blocksize(zp, new_blksz, tx); 696c5c6ffa0Smaybee zfs_range_reduce(rl, woff, n); 697fa9e4066Sahrens } 698fa9e4066Sahrens 699fa9e4066Sahrens /* 700fa9e4066Sahrens * The file data does not fit in the znode "cache", so we 701fa9e4066Sahrens * will be writing to the file block data buffers. 702fa9e4066Sahrens * Each buffer will be written in a separate transaction; 703fa9e4066Sahrens * this keeps the intent log records small and allows us 704fa9e4066Sahrens * to do more fine-grained space accounting. 705fa9e4066Sahrens */ 706fa9e4066Sahrens while (n > 0) { 707fa9e4066Sahrens /* 708fa9e4066Sahrens * XXX - should we really limit each write to z_max_blksz? 709fa9e4066Sahrens * Perhaps we should use SPA_MAXBLOCKSIZE chunks? 710fa9e4066Sahrens */ 711fa9e4066Sahrens nbytes = MIN(n, max_blksz - P2PHASE(woff, max_blksz)); 712fa9e4066Sahrens rw_enter(&zp->z_map_lock, RW_READER); 713fa9e4066Sahrens 714fa9e4066Sahrens tx_bytes = uio->uio_resid; 715fa9e4066Sahrens if (vn_has_cached_data(vp)) { 716fa9e4066Sahrens rw_exit(&zp->z_map_lock); 717fa9e4066Sahrens error = mappedwrite(vp, woff, nbytes, uio, tx); 718fa9e4066Sahrens } else { 719fa9e4066Sahrens error = dmu_write_uio(zfsvfs->z_os, zp->z_id, 720fa9e4066Sahrens woff, nbytes, uio, tx); 721fa9e4066Sahrens rw_exit(&zp->z_map_lock); 722fa9e4066Sahrens } 723fa9e4066Sahrens tx_bytes -= uio->uio_resid; 724fa9e4066Sahrens 725fa9e4066Sahrens if (error) { 726fa9e4066Sahrens /* XXX - do we need to "clean up" the dmu buffer? */ 727fa9e4066Sahrens break; 728fa9e4066Sahrens } 729fa9e4066Sahrens 730fa9e4066Sahrens ASSERT(tx_bytes == nbytes); 731fa9e4066Sahrens 732169cdae2Smarks /* 733169cdae2Smarks * Clear Set-UID/Set-GID bits on successful write if not 734169cdae2Smarks * privileged and at least one of the excute bits is set. 735169cdae2Smarks * 736169cdae2Smarks * It would be nice to to this after all writes have 737169cdae2Smarks * been done, but that would still expose the ISUID/ISGID 738169cdae2Smarks * to another app after the partial write is committed. 739169cdae2Smarks */ 740169cdae2Smarks 741169cdae2Smarks mutex_enter(&zp->z_acl_lock); 742169cdae2Smarks if ((zp->z_phys->zp_mode & (S_IXUSR | (S_IXUSR >> 3) | 743169cdae2Smarks (S_IXUSR >> 6))) != 0 && 744169cdae2Smarks (zp->z_phys->zp_mode & (S_ISUID | S_ISGID)) != 0 && 745169cdae2Smarks secpolicy_vnode_setid_retain(cr, 746169cdae2Smarks (zp->z_phys->zp_mode & S_ISUID) != 0 && 747169cdae2Smarks zp->z_phys->zp_uid == 0) != 0) { 748169cdae2Smarks zp->z_phys->zp_mode &= ~(S_ISUID | S_ISGID); 749169cdae2Smarks } 750169cdae2Smarks mutex_exit(&zp->z_acl_lock); 751169cdae2Smarks 752fa9e4066Sahrens n -= nbytes; 753fa9e4066Sahrens if (n <= 0) 754fa9e4066Sahrens break; 755fa9e4066Sahrens 756fa9e4066Sahrens /* 757fa9e4066Sahrens * We have more work ahead of us, so wrap up this transaction 758fa9e4066Sahrens * and start another. Exact same logic as tx_done below. 759fa9e4066Sahrens */ 760fa9e4066Sahrens while ((end_size = zp->z_phys->zp_size) < uio->uio_loffset) { 761fa9e4066Sahrens dmu_buf_will_dirty(zp->z_dbuf, tx); 762fa9e4066Sahrens (void) atomic_cas_64(&zp->z_phys->zp_size, end_size, 763fa9e4066Sahrens uio->uio_loffset); 764fa9e4066Sahrens } 765fa9e4066Sahrens zfs_time_stamper(zp, CONTENT_MODIFIED, tx); 766b19a79ecSperrin zfs_log_write(zilog, tx, TX_WRITE, zp, woff, tx_bytes, 767fa9e4066Sahrens ioflag, uio); 768fa9e4066Sahrens dmu_tx_commit(tx); 769fa9e4066Sahrens 770fa9e4066Sahrens /* 771fa9e4066Sahrens * Start another transaction. 772fa9e4066Sahrens */ 773fa9e4066Sahrens woff = uio->uio_loffset; 774fa9e4066Sahrens tx = dmu_tx_create(zfsvfs->z_os); 775fa9e4066Sahrens dmu_tx_hold_bonus(tx, zp->z_id); 776fa9e4066Sahrens dmu_tx_hold_write(tx, zp->z_id, woff, MIN(n, max_blksz)); 777fa9e4066Sahrens error = dmu_tx_assign(tx, zfsvfs->z_assign); 778fa9e4066Sahrens if (error) { 779fa9e4066Sahrens if (error == ERESTART && 780fa9e4066Sahrens zfsvfs->z_assign == TXG_NOWAIT) { 7818a2f1b91Sahrens dmu_tx_wait(tx); 7828a2f1b91Sahrens dmu_tx_abort(tx); 783fa9e4066Sahrens goto top; 784fa9e4066Sahrens } 7858a2f1b91Sahrens dmu_tx_abort(tx); 786fa9e4066Sahrens goto no_tx_done; 787fa9e4066Sahrens } 788fa9e4066Sahrens } 789fa9e4066Sahrens 790fa9e4066Sahrens tx_done: 791fa9e4066Sahrens 792fa9e4066Sahrens if (tx_bytes != 0) { 793fa9e4066Sahrens /* 794fa9e4066Sahrens * Update the file size if it has changed; account 795fa9e4066Sahrens * for possible concurrent updates. 796fa9e4066Sahrens */ 797fa9e4066Sahrens while ((end_size = zp->z_phys->zp_size) < uio->uio_loffset) { 798fa9e4066Sahrens dmu_buf_will_dirty(zp->z_dbuf, tx); 799fa9e4066Sahrens (void) atomic_cas_64(&zp->z_phys->zp_size, end_size, 800fa9e4066Sahrens uio->uio_loffset); 801fa9e4066Sahrens } 802fa9e4066Sahrens zfs_time_stamper(zp, CONTENT_MODIFIED, tx); 803b19a79ecSperrin zfs_log_write(zilog, tx, TX_WRITE, zp, woff, tx_bytes, 804fa9e4066Sahrens ioflag, uio); 805fa9e4066Sahrens } 806fa9e4066Sahrens dmu_tx_commit(tx); 807fa9e4066Sahrens 808fa9e4066Sahrens 809fa9e4066Sahrens no_tx_done: 810fa9e4066Sahrens 811c5c6ffa0Smaybee zfs_range_unlock(rl); 812fa9e4066Sahrens 813fa9e4066Sahrens /* 814fa9e4066Sahrens * If we're in replay mode, or we made no progress, return error. 815fa9e4066Sahrens * Otherwise, it's at least a partial write, so it's successful. 816fa9e4066Sahrens */ 817fa9e4066Sahrens if (zfsvfs->z_assign >= TXG_INITIAL || uio->uio_resid == start_resid) { 818fa9e4066Sahrens ZFS_EXIT(zfsvfs); 819fa9e4066Sahrens return (error); 820fa9e4066Sahrens } 821fa9e4066Sahrens 822b19a79ecSperrin if (ioflag & (FSYNC | FDSYNC)) 823b19a79ecSperrin zil_commit(zilog, zp->z_last_itx, zp->z_id); 824fa9e4066Sahrens 825fa9e4066Sahrens ZFS_EXIT(zfsvfs); 826fa9e4066Sahrens return (0); 827fa9e4066Sahrens } 828fa9e4066Sahrens 829c5c6ffa0Smaybee void 830c5c6ffa0Smaybee zfs_get_done(dmu_buf_t *db, void *vrl) 831c5c6ffa0Smaybee { 832c5c6ffa0Smaybee rl_t *rl = (rl_t *)vrl; 833c5c6ffa0Smaybee vnode_t *vp = ZTOV(rl->r_zp); 834c5c6ffa0Smaybee 835c5c6ffa0Smaybee dmu_buf_rele(db, rl); 836c5c6ffa0Smaybee zfs_range_unlock(rl); 837c5c6ffa0Smaybee VN_RELE(vp); 838c5c6ffa0Smaybee } 839c5c6ffa0Smaybee 840fa9e4066Sahrens /* 841fa9e4066Sahrens * Get data to generate a TX_WRITE intent log record. 842fa9e4066Sahrens */ 843fa9e4066Sahrens int 844c5c6ffa0Smaybee zfs_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio) 845fa9e4066Sahrens { 846fa9e4066Sahrens zfsvfs_t *zfsvfs = arg; 847fa9e4066Sahrens objset_t *os = zfsvfs->z_os; 848fa9e4066Sahrens znode_t *zp; 849fa9e4066Sahrens uint64_t off = lr->lr_offset; 850c5c6ffa0Smaybee dmu_buf_t *db; 851104e2ed7Sperrin rl_t *rl; 852fa9e4066Sahrens int dlen = lr->lr_length; /* length of user data */ 853fa9e4066Sahrens int error = 0; 854fa9e4066Sahrens 855fa9e4066Sahrens ASSERT(dlen != 0); 856fa9e4066Sahrens 857fa9e4066Sahrens /* 858104e2ed7Sperrin * Nothing to do if the file has been removed 859fa9e4066Sahrens */ 860fa9e4066Sahrens if (zfs_zget(zfsvfs, lr->lr_foid, &zp) != 0) 861fa9e4066Sahrens return (ENOENT); 862104e2ed7Sperrin if (zp->z_reap) { 863fa9e4066Sahrens VN_RELE(ZTOV(zp)); 864fa9e4066Sahrens return (ENOENT); 865fa9e4066Sahrens } 866fa9e4066Sahrens 867fa9e4066Sahrens /* 868fa9e4066Sahrens * Write records come in two flavors: immediate and indirect. 869fa9e4066Sahrens * For small writes it's cheaper to store the data with the 870fa9e4066Sahrens * log record (immediate); for large writes it's cheaper to 871fa9e4066Sahrens * sync the data and get a pointer to it (indirect) so that 872fa9e4066Sahrens * we don't have to write the data twice. 873fa9e4066Sahrens */ 874104e2ed7Sperrin if (buf != NULL) { /* immediate write */ 875104e2ed7Sperrin rl = zfs_range_lock(zp, off, dlen, RL_READER); 876104e2ed7Sperrin /* test for truncation needs to be done while range locked */ 877104e2ed7Sperrin if (off >= zp->z_phys->zp_size) { 878104e2ed7Sperrin error = ENOENT; 879104e2ed7Sperrin goto out; 880104e2ed7Sperrin } 88110d957caSmaybee VERIFY(0 == dmu_read(os, lr->lr_foid, off, dlen, buf)); 882104e2ed7Sperrin } else { /* indirect write */ 883104e2ed7Sperrin uint64_t boff; /* block starting offset */ 884104e2ed7Sperrin 88510d957caSmaybee ASSERT3U(dlen, <=, zp->z_blksz); 886fa9e4066Sahrens /* 887104e2ed7Sperrin * Have to lock the whole block to ensure when it's 888104e2ed7Sperrin * written out and it's checksum is being calculated 889104e2ed7Sperrin * that no one can change the data. We need to re-check 890104e2ed7Sperrin * blocksize after we get the lock in case it's changed! 891fa9e4066Sahrens */ 892104e2ed7Sperrin for (;;) { 8935263255cSperrin if (ISP2(zp->z_blksz)) { 8945263255cSperrin boff = P2ALIGN_TYPED(off, zp->z_blksz, 8955263255cSperrin uint64_t); 8965263255cSperrin } else { 8975263255cSperrin boff = 0; 8985263255cSperrin } 899104e2ed7Sperrin dlen = zp->z_blksz; 900104e2ed7Sperrin rl = zfs_range_lock(zp, boff, dlen, RL_READER); 901104e2ed7Sperrin if (zp->z_blksz == dlen) 902104e2ed7Sperrin break; 903c5c6ffa0Smaybee zfs_range_unlock(rl); 904104e2ed7Sperrin } 905104e2ed7Sperrin /* test for truncation needs to be done while range locked */ 906104e2ed7Sperrin if (off >= zp->z_phys->zp_size) { 907104e2ed7Sperrin error = ENOENT; 908104e2ed7Sperrin goto out; 909104e2ed7Sperrin } 910c5c6ffa0Smaybee VERIFY(0 == dmu_buf_hold(os, lr->lr_foid, boff, rl, &db)); 911c5c6ffa0Smaybee ASSERT(boff == db->db_offset); 912c5c6ffa0Smaybee lr->lr_blkoff = off - boff; 913c5c6ffa0Smaybee error = dmu_sync(zio, db, &lr->lr_blkptr, 914c5c6ffa0Smaybee lr->lr_common.lrc_txg, zio ? zfs_get_done : NULL, rl); 915c5c6ffa0Smaybee /* 916c5c6ffa0Smaybee * If we get EINPROGRESS, then we need to wait for a 917c5c6ffa0Smaybee * write IO initiated by dmu_sync() to complete before 918b19a79ecSperrin * we can release this dbuf. We will finish everything 919c5c6ffa0Smaybee * up in the zfs_get_done() callback. 920c5c6ffa0Smaybee */ 921c5c6ffa0Smaybee if (error == EINPROGRESS) 922c5c6ffa0Smaybee return (0); 923c5c6ffa0Smaybee dmu_buf_rele(db, rl); 924fa9e4066Sahrens } 925104e2ed7Sperrin out: 926c5c6ffa0Smaybee zfs_range_unlock(rl); 927fa9e4066Sahrens VN_RELE(ZTOV(zp)); 928fa9e4066Sahrens return (error); 929fa9e4066Sahrens } 930fa9e4066Sahrens 931fa9e4066Sahrens /*ARGSUSED*/ 932fa9e4066Sahrens static int 933fa9e4066Sahrens zfs_access(vnode_t *vp, int mode, int flags, cred_t *cr) 934fa9e4066Sahrens { 935fa9e4066Sahrens znode_t *zp = VTOZ(vp); 936fa9e4066Sahrens zfsvfs_t *zfsvfs = zp->z_zfsvfs; 937fa9e4066Sahrens int error; 938fa9e4066Sahrens 939fa9e4066Sahrens ZFS_ENTER(zfsvfs); 940fa9e4066Sahrens error = zfs_zaccess_rwx(zp, mode, cr); 941fa9e4066Sahrens ZFS_EXIT(zfsvfs); 942fa9e4066Sahrens return (error); 943fa9e4066Sahrens } 944fa9e4066Sahrens 945fa9e4066Sahrens /* 946fa9e4066Sahrens * Lookup an entry in a directory, or an extended attribute directory. 947fa9e4066Sahrens * If it exists, return a held vnode reference for it. 948fa9e4066Sahrens * 949fa9e4066Sahrens * IN: dvp - vnode of directory to search. 950fa9e4066Sahrens * nm - name of entry to lookup. 951fa9e4066Sahrens * pnp - full pathname to lookup [UNUSED]. 952fa9e4066Sahrens * flags - LOOKUP_XATTR set if looking for an attribute. 953fa9e4066Sahrens * rdir - root directory vnode [UNUSED]. 954fa9e4066Sahrens * cr - credentials of caller. 955fa9e4066Sahrens * 956fa9e4066Sahrens * OUT: vpp - vnode of located entry, NULL if not found. 957fa9e4066Sahrens * 958fa9e4066Sahrens * RETURN: 0 if success 959fa9e4066Sahrens * error code if failure 960fa9e4066Sahrens * 961fa9e4066Sahrens * Timestamps: 962fa9e4066Sahrens * NA 963fa9e4066Sahrens */ 964fa9e4066Sahrens /* ARGSUSED */ 965fa9e4066Sahrens static int 966fa9e4066Sahrens zfs_lookup(vnode_t *dvp, char *nm, vnode_t **vpp, struct pathname *pnp, 967fa9e4066Sahrens int flags, vnode_t *rdir, cred_t *cr) 968fa9e4066Sahrens { 969fa9e4066Sahrens 970fa9e4066Sahrens znode_t *zdp = VTOZ(dvp); 971fa9e4066Sahrens zfsvfs_t *zfsvfs = zdp->z_zfsvfs; 972fa9e4066Sahrens int error; 973fa9e4066Sahrens 974fa9e4066Sahrens ZFS_ENTER(zfsvfs); 975fa9e4066Sahrens 976fa9e4066Sahrens *vpp = NULL; 977fa9e4066Sahrens 978fa9e4066Sahrens if (flags & LOOKUP_XATTR) { 979fa9e4066Sahrens /* 980fa9e4066Sahrens * We don't allow recursive attributes.. 981fa9e4066Sahrens * Maybe someday we will. 982fa9e4066Sahrens */ 983fa9e4066Sahrens if (zdp->z_phys->zp_flags & ZFS_XATTR) { 984fa9e4066Sahrens ZFS_EXIT(zfsvfs); 985fa9e4066Sahrens return (EINVAL); 986fa9e4066Sahrens } 987fa9e4066Sahrens 988fa9e4066Sahrens if (error = zfs_get_xattrdir(VTOZ(dvp), vpp, cr)) { 989fa9e4066Sahrens ZFS_EXIT(zfsvfs); 990fa9e4066Sahrens return (error); 991fa9e4066Sahrens } 992fa9e4066Sahrens 993fa9e4066Sahrens /* 994fa9e4066Sahrens * Do we have permission to get into attribute directory? 995fa9e4066Sahrens */ 996fa9e4066Sahrens 997fa9e4066Sahrens if (error = zfs_zaccess(VTOZ(*vpp), ACE_EXECUTE, cr)) { 998fa9e4066Sahrens VN_RELE(*vpp); 999fa9e4066Sahrens } 1000fa9e4066Sahrens 1001fa9e4066Sahrens ZFS_EXIT(zfsvfs); 1002fa9e4066Sahrens return (error); 1003fa9e4066Sahrens } 1004fa9e4066Sahrens 10050f2dc02eSek if (dvp->v_type != VDIR) { 10060f2dc02eSek ZFS_EXIT(zfsvfs); 1007736b9155Smarks return (ENOTDIR); 10080f2dc02eSek } 1009736b9155Smarks 1010fa9e4066Sahrens /* 1011fa9e4066Sahrens * Check accessibility of directory. 1012fa9e4066Sahrens */ 1013fa9e4066Sahrens 1014fa9e4066Sahrens if (error = zfs_zaccess(zdp, ACE_EXECUTE, cr)) { 1015fa9e4066Sahrens ZFS_EXIT(zfsvfs); 1016fa9e4066Sahrens return (error); 1017fa9e4066Sahrens } 1018fa9e4066Sahrens 1019fa9e4066Sahrens if ((error = zfs_dirlook(zdp, nm, vpp)) == 0) { 1020fa9e4066Sahrens 1021fa9e4066Sahrens /* 1022fa9e4066Sahrens * Convert device special files 1023fa9e4066Sahrens */ 1024fa9e4066Sahrens if (IS_DEVVP(*vpp)) { 1025fa9e4066Sahrens vnode_t *svp; 1026fa9e4066Sahrens 1027fa9e4066Sahrens svp = specvp(*vpp, (*vpp)->v_rdev, (*vpp)->v_type, cr); 1028fa9e4066Sahrens VN_RELE(*vpp); 1029fa9e4066Sahrens if (svp == NULL) 1030fa9e4066Sahrens error = ENOSYS; 1031fa9e4066Sahrens else 1032fa9e4066Sahrens *vpp = svp; 1033fa9e4066Sahrens } 1034fa9e4066Sahrens } 1035fa9e4066Sahrens 1036fa9e4066Sahrens ZFS_EXIT(zfsvfs); 1037fa9e4066Sahrens return (error); 1038fa9e4066Sahrens } 1039fa9e4066Sahrens 1040fa9e4066Sahrens /* 1041fa9e4066Sahrens * Attempt to create a new entry in a directory. If the entry 1042fa9e4066Sahrens * already exists, truncate the file if permissible, else return 1043fa9e4066Sahrens * an error. Return the vp of the created or trunc'd file. 1044fa9e4066Sahrens * 1045fa9e4066Sahrens * IN: dvp - vnode of directory to put new file entry in. 1046fa9e4066Sahrens * name - name of new file entry. 1047fa9e4066Sahrens * vap - attributes of new file. 1048fa9e4066Sahrens * excl - flag indicating exclusive or non-exclusive mode. 1049fa9e4066Sahrens * mode - mode to open file with. 1050fa9e4066Sahrens * cr - credentials of caller. 1051fa9e4066Sahrens * flag - large file flag [UNUSED]. 1052fa9e4066Sahrens * 1053fa9e4066Sahrens * OUT: vpp - vnode of created or trunc'd entry. 1054fa9e4066Sahrens * 1055fa9e4066Sahrens * RETURN: 0 if success 1056fa9e4066Sahrens * error code if failure 1057fa9e4066Sahrens * 1058fa9e4066Sahrens * Timestamps: 1059fa9e4066Sahrens * dvp - ctime|mtime updated if new entry created 1060fa9e4066Sahrens * vp - ctime|mtime always, atime if new 1061fa9e4066Sahrens */ 1062fa9e4066Sahrens /* ARGSUSED */ 1063fa9e4066Sahrens static int 1064fa9e4066Sahrens zfs_create(vnode_t *dvp, char *name, vattr_t *vap, vcexcl_t excl, 1065fa9e4066Sahrens int mode, vnode_t **vpp, cred_t *cr, int flag) 1066fa9e4066Sahrens { 1067fa9e4066Sahrens znode_t *zp, *dzp = VTOZ(dvp); 1068fa9e4066Sahrens zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 1069fa9e4066Sahrens zilog_t *zilog = zfsvfs->z_log; 1070fa9e4066Sahrens objset_t *os = zfsvfs->z_os; 1071fa9e4066Sahrens zfs_dirlock_t *dl; 1072fa9e4066Sahrens dmu_tx_t *tx; 1073fa9e4066Sahrens int error; 1074fa9e4066Sahrens uint64_t zoid; 1075fa9e4066Sahrens 1076fa9e4066Sahrens ZFS_ENTER(zfsvfs); 1077fa9e4066Sahrens 1078fa9e4066Sahrens top: 1079fa9e4066Sahrens *vpp = NULL; 1080fa9e4066Sahrens 1081fa9e4066Sahrens if ((vap->va_mode & VSVTX) && secpolicy_vnode_stky_modify(cr)) 1082fa9e4066Sahrens vap->va_mode &= ~VSVTX; 1083fa9e4066Sahrens 1084fa9e4066Sahrens if (*name == '\0') { 1085fa9e4066Sahrens /* 1086fa9e4066Sahrens * Null component name refers to the directory itself. 1087fa9e4066Sahrens */ 1088fa9e4066Sahrens VN_HOLD(dvp); 1089fa9e4066Sahrens zp = dzp; 1090fa9e4066Sahrens dl = NULL; 1091fa9e4066Sahrens error = 0; 1092fa9e4066Sahrens } else { 1093fa9e4066Sahrens /* possible VN_HOLD(zp) */ 1094fa9e4066Sahrens if (error = zfs_dirent_lock(&dl, dzp, name, &zp, 0)) { 1095fa9e4066Sahrens if (strcmp(name, "..") == 0) 1096fa9e4066Sahrens error = EISDIR; 1097fa9e4066Sahrens ZFS_EXIT(zfsvfs); 1098fa9e4066Sahrens return (error); 1099fa9e4066Sahrens } 1100fa9e4066Sahrens } 1101fa9e4066Sahrens 1102fa9e4066Sahrens zoid = zp ? zp->z_id : -1ULL; 1103fa9e4066Sahrens 1104fa9e4066Sahrens if (zp == NULL) { 1105fa9e4066Sahrens /* 1106fa9e4066Sahrens * Create a new file object and update the directory 1107fa9e4066Sahrens * to reference it. 1108fa9e4066Sahrens */ 1109fa9e4066Sahrens if (error = zfs_zaccess(dzp, ACE_ADD_FILE, cr)) { 1110fa9e4066Sahrens goto out; 1111fa9e4066Sahrens } 1112fa9e4066Sahrens 1113fa9e4066Sahrens /* 1114fa9e4066Sahrens * We only support the creation of regular files in 1115fa9e4066Sahrens * extended attribute directories. 1116fa9e4066Sahrens */ 1117fa9e4066Sahrens if ((dzp->z_phys->zp_flags & ZFS_XATTR) && 1118fa9e4066Sahrens (vap->va_type != VREG)) { 1119fa9e4066Sahrens error = EINVAL; 1120fa9e4066Sahrens goto out; 1121fa9e4066Sahrens } 1122fa9e4066Sahrens 1123fa9e4066Sahrens tx = dmu_tx_create(os); 1124fa9e4066Sahrens dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT); 1125fa9e4066Sahrens dmu_tx_hold_bonus(tx, dzp->z_id); 1126ea8dc4b6Seschrock dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); 1127fa9e4066Sahrens if (dzp->z_phys->zp_flags & ZFS_INHERIT_ACE) 1128fa9e4066Sahrens dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 1129fa9e4066Sahrens 0, SPA_MAXBLOCKSIZE); 1130fa9e4066Sahrens error = dmu_tx_assign(tx, zfsvfs->z_assign); 1131fa9e4066Sahrens if (error) { 1132fa9e4066Sahrens zfs_dirent_unlock(dl); 1133fa9e4066Sahrens if (error == ERESTART && 1134fa9e4066Sahrens zfsvfs->z_assign == TXG_NOWAIT) { 11358a2f1b91Sahrens dmu_tx_wait(tx); 11368a2f1b91Sahrens dmu_tx_abort(tx); 1137fa9e4066Sahrens goto top; 1138fa9e4066Sahrens } 11398a2f1b91Sahrens dmu_tx_abort(tx); 1140fa9e4066Sahrens ZFS_EXIT(zfsvfs); 1141fa9e4066Sahrens return (error); 1142fa9e4066Sahrens } 1143fa9e4066Sahrens zfs_mknode(dzp, vap, &zoid, tx, cr, 0, &zp, 0); 1144fa9e4066Sahrens ASSERT(zp->z_id == zoid); 1145fa9e4066Sahrens (void) zfs_link_create(dl, zp, tx, ZNEW); 1146b19a79ecSperrin zfs_log_create(zilog, tx, TX_CREATE, dzp, zp, name); 1147fa9e4066Sahrens dmu_tx_commit(tx); 1148fa9e4066Sahrens } else { 1149fa9e4066Sahrens /* 1150fa9e4066Sahrens * A directory entry already exists for this name. 1151fa9e4066Sahrens */ 1152fa9e4066Sahrens /* 1153fa9e4066Sahrens * Can't truncate an existing file if in exclusive mode. 1154fa9e4066Sahrens */ 1155fa9e4066Sahrens if (excl == EXCL) { 1156fa9e4066Sahrens error = EEXIST; 1157fa9e4066Sahrens goto out; 1158fa9e4066Sahrens } 1159fa9e4066Sahrens /* 1160fa9e4066Sahrens * Can't open a directory for writing. 1161fa9e4066Sahrens */ 1162fa9e4066Sahrens if ((ZTOV(zp)->v_type == VDIR) && (mode & S_IWRITE)) { 1163fa9e4066Sahrens error = EISDIR; 1164fa9e4066Sahrens goto out; 1165fa9e4066Sahrens } 1166fa9e4066Sahrens /* 1167fa9e4066Sahrens * Verify requested access to file. 1168fa9e4066Sahrens */ 1169fa9e4066Sahrens if (mode && (error = zfs_zaccess_rwx(zp, mode, cr))) { 1170fa9e4066Sahrens goto out; 1171fa9e4066Sahrens } 1172fa9e4066Sahrens 1173fa9e4066Sahrens mutex_enter(&dzp->z_lock); 1174fa9e4066Sahrens dzp->z_seq++; 1175fa9e4066Sahrens mutex_exit(&dzp->z_lock); 1176fa9e4066Sahrens 11775730cc9aSmaybee /* 11785730cc9aSmaybee * Truncate regular files if requested. 11795730cc9aSmaybee */ 11805730cc9aSmaybee if ((ZTOV(zp)->v_type == VREG) && 11815730cc9aSmaybee (zp->z_phys->zp_size != 0) && 1182fa9e4066Sahrens (vap->va_mask & AT_SIZE) && (vap->va_size == 0)) { 11835730cc9aSmaybee error = zfs_freesp(zp, 0, 0, mode, TRUE); 11845730cc9aSmaybee if (error == ERESTART && 11855730cc9aSmaybee zfsvfs->z_assign == TXG_NOWAIT) { 11868a2f1b91Sahrens /* NB: we already did dmu_tx_wait() */ 11875730cc9aSmaybee zfs_dirent_unlock(dl); 11888991ab5bSperrin VN_RELE(ZTOV(zp)); 11895730cc9aSmaybee goto top; 1190fa9e4066Sahrens } 1191fa9e4066Sahrens } 1192fa9e4066Sahrens } 1193fa9e4066Sahrens out: 1194fa9e4066Sahrens 1195fa9e4066Sahrens if (dl) 1196fa9e4066Sahrens zfs_dirent_unlock(dl); 1197fa9e4066Sahrens 1198fa9e4066Sahrens if (error) { 1199fa9e4066Sahrens if (zp) 1200fa9e4066Sahrens VN_RELE(ZTOV(zp)); 1201fa9e4066Sahrens } else { 1202fa9e4066Sahrens *vpp = ZTOV(zp); 1203fa9e4066Sahrens /* 1204fa9e4066Sahrens * If vnode is for a device return a specfs vnode instead. 1205fa9e4066Sahrens */ 1206fa9e4066Sahrens if (IS_DEVVP(*vpp)) { 1207fa9e4066Sahrens struct vnode *svp; 1208fa9e4066Sahrens 1209fa9e4066Sahrens svp = specvp(*vpp, (*vpp)->v_rdev, (*vpp)->v_type, cr); 1210fa9e4066Sahrens VN_RELE(*vpp); 1211fa9e4066Sahrens if (svp == NULL) { 1212fa9e4066Sahrens error = ENOSYS; 1213fa9e4066Sahrens } 1214fa9e4066Sahrens *vpp = svp; 1215fa9e4066Sahrens } 1216fa9e4066Sahrens } 1217fa9e4066Sahrens 1218fa9e4066Sahrens ZFS_EXIT(zfsvfs); 1219fa9e4066Sahrens return (error); 1220fa9e4066Sahrens } 1221fa9e4066Sahrens 1222fa9e4066Sahrens /* 1223fa9e4066Sahrens * Remove an entry from a directory. 1224fa9e4066Sahrens * 1225fa9e4066Sahrens * IN: dvp - vnode of directory to remove entry from. 1226fa9e4066Sahrens * name - name of entry to remove. 1227fa9e4066Sahrens * cr - credentials of caller. 1228fa9e4066Sahrens * 1229fa9e4066Sahrens * RETURN: 0 if success 1230fa9e4066Sahrens * error code if failure 1231fa9e4066Sahrens * 1232fa9e4066Sahrens * Timestamps: 1233fa9e4066Sahrens * dvp - ctime|mtime 1234fa9e4066Sahrens * vp - ctime (if nlink > 0) 1235fa9e4066Sahrens */ 1236fa9e4066Sahrens static int 1237fa9e4066Sahrens zfs_remove(vnode_t *dvp, char *name, cred_t *cr) 1238fa9e4066Sahrens { 1239fa9e4066Sahrens znode_t *zp, *dzp = VTOZ(dvp); 1240fa9e4066Sahrens znode_t *xzp = NULL; 1241fa9e4066Sahrens vnode_t *vp; 1242fa9e4066Sahrens zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 1243fa9e4066Sahrens zilog_t *zilog = zfsvfs->z_log; 1244fa9e4066Sahrens uint64_t acl_obj, xattr_obj; 1245fa9e4066Sahrens zfs_dirlock_t *dl; 1246fa9e4066Sahrens dmu_tx_t *tx; 1247fa9e4066Sahrens int may_delete_now, delete_now = FALSE; 1248fa9e4066Sahrens int reaped; 1249fa9e4066Sahrens int error; 1250fa9e4066Sahrens 1251fa9e4066Sahrens ZFS_ENTER(zfsvfs); 1252fa9e4066Sahrens 1253fa9e4066Sahrens top: 1254fa9e4066Sahrens /* 1255fa9e4066Sahrens * Attempt to lock directory; fail if entry doesn't exist. 1256fa9e4066Sahrens */ 1257fa9e4066Sahrens if (error = zfs_dirent_lock(&dl, dzp, name, &zp, ZEXISTS)) { 1258fa9e4066Sahrens ZFS_EXIT(zfsvfs); 1259fa9e4066Sahrens return (error); 1260fa9e4066Sahrens } 1261fa9e4066Sahrens 1262fa9e4066Sahrens vp = ZTOV(zp); 1263fa9e4066Sahrens 1264fa9e4066Sahrens if (error = zfs_zaccess_delete(dzp, zp, cr)) { 1265fa9e4066Sahrens goto out; 1266fa9e4066Sahrens } 1267fa9e4066Sahrens 1268fa9e4066Sahrens /* 1269fa9e4066Sahrens * Need to use rmdir for removing directories. 1270fa9e4066Sahrens */ 1271fa9e4066Sahrens if (vp->v_type == VDIR) { 1272fa9e4066Sahrens error = EPERM; 1273fa9e4066Sahrens goto out; 1274fa9e4066Sahrens } 1275fa9e4066Sahrens 1276fa9e4066Sahrens vnevent_remove(vp); 1277fa9e4066Sahrens 1278033f9833Sek dnlc_remove(dvp, name); 1279033f9833Sek 1280fa9e4066Sahrens mutex_enter(&vp->v_lock); 1281fa9e4066Sahrens may_delete_now = vp->v_count == 1 && !vn_has_cached_data(vp); 1282fa9e4066Sahrens mutex_exit(&vp->v_lock); 1283fa9e4066Sahrens 1284fa9e4066Sahrens /* 1285fa9e4066Sahrens * We may delete the znode now, or we may put it on the delete queue; 1286fa9e4066Sahrens * it depends on whether we're the last link, and on whether there are 1287fa9e4066Sahrens * other holds on the vnode. So we dmu_tx_hold() the right things to 1288fa9e4066Sahrens * allow for either case. 1289fa9e4066Sahrens */ 1290fa9e4066Sahrens tx = dmu_tx_create(zfsvfs->z_os); 1291ea8dc4b6Seschrock dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); 1292fa9e4066Sahrens dmu_tx_hold_bonus(tx, zp->z_id); 1293fa9e4066Sahrens if (may_delete_now) 1294fa9e4066Sahrens dmu_tx_hold_free(tx, zp->z_id, 0, DMU_OBJECT_END); 1295fa9e4066Sahrens 1296fa9e4066Sahrens /* are there any extended attributes? */ 1297fa9e4066Sahrens if ((xattr_obj = zp->z_phys->zp_xattr) != 0) { 1298fa9e4066Sahrens /* 1299fa9e4066Sahrens * XXX - There is a possibility that the delete 1300fa9e4066Sahrens * of the parent file could succeed, but then we get 1301fa9e4066Sahrens * an ENOSPC when we try to delete the xattrs... 1302fa9e4066Sahrens * so we would need to re-try the deletes periodically 1303fa9e4066Sahrens */ 1304fa9e4066Sahrens /* XXX - do we need this if we are deleting? */ 1305fa9e4066Sahrens dmu_tx_hold_bonus(tx, xattr_obj); 1306fa9e4066Sahrens } 1307fa9e4066Sahrens 1308fa9e4066Sahrens /* are there any additional acls */ 1309fa9e4066Sahrens if ((acl_obj = zp->z_phys->zp_acl.z_acl_extern_obj) != 0 && 1310fa9e4066Sahrens may_delete_now) 1311fa9e4066Sahrens dmu_tx_hold_free(tx, acl_obj, 0, DMU_OBJECT_END); 1312fa9e4066Sahrens 1313fa9e4066Sahrens /* charge as an update -- would be nice not to charge at all */ 1314ea8dc4b6Seschrock dmu_tx_hold_zap(tx, zfsvfs->z_dqueue, FALSE, NULL); 1315fa9e4066Sahrens 1316fa9e4066Sahrens error = dmu_tx_assign(tx, zfsvfs->z_assign); 1317fa9e4066Sahrens if (error) { 1318fa9e4066Sahrens zfs_dirent_unlock(dl); 1319fa9e4066Sahrens VN_RELE(vp); 1320fa9e4066Sahrens if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) { 13218a2f1b91Sahrens dmu_tx_wait(tx); 13228a2f1b91Sahrens dmu_tx_abort(tx); 1323fa9e4066Sahrens goto top; 1324fa9e4066Sahrens } 13258a2f1b91Sahrens dmu_tx_abort(tx); 1326fa9e4066Sahrens ZFS_EXIT(zfsvfs); 1327fa9e4066Sahrens return (error); 1328fa9e4066Sahrens } 1329fa9e4066Sahrens 1330fa9e4066Sahrens /* 1331fa9e4066Sahrens * Remove the directory entry. 1332fa9e4066Sahrens */ 1333fa9e4066Sahrens error = zfs_link_destroy(dl, zp, tx, 0, &reaped); 1334fa9e4066Sahrens 1335fa9e4066Sahrens if (error) { 1336fa9e4066Sahrens dmu_tx_commit(tx); 1337fa9e4066Sahrens goto out; 1338fa9e4066Sahrens } 1339fa9e4066Sahrens 1340fa9e4066Sahrens if (reaped) { 1341fa9e4066Sahrens mutex_enter(&vp->v_lock); 1342fa9e4066Sahrens delete_now = may_delete_now && 1343fa9e4066Sahrens vp->v_count == 1 && !vn_has_cached_data(vp) && 1344fa9e4066Sahrens zp->z_phys->zp_xattr == xattr_obj && 1345fa9e4066Sahrens zp->z_phys->zp_acl.z_acl_extern_obj == acl_obj; 1346fa9e4066Sahrens mutex_exit(&vp->v_lock); 1347fa9e4066Sahrens } 1348fa9e4066Sahrens 1349fa9e4066Sahrens if (delete_now) { 1350fa9e4066Sahrens if (zp->z_phys->zp_xattr) { 1351fa9e4066Sahrens error = zfs_zget(zfsvfs, zp->z_phys->zp_xattr, &xzp); 1352fa9e4066Sahrens ASSERT3U(error, ==, 0); 1353fa9e4066Sahrens ASSERT3U(xzp->z_phys->zp_links, ==, 2); 1354fa9e4066Sahrens dmu_buf_will_dirty(xzp->z_dbuf, tx); 1355fa9e4066Sahrens mutex_enter(&xzp->z_lock); 1356fa9e4066Sahrens xzp->z_reap = 1; 1357fa9e4066Sahrens xzp->z_phys->zp_links = 0; 1358fa9e4066Sahrens mutex_exit(&xzp->z_lock); 1359fa9e4066Sahrens zfs_dq_add(xzp, tx); 1360fa9e4066Sahrens zp->z_phys->zp_xattr = 0; /* probably unnecessary */ 1361fa9e4066Sahrens } 1362fa9e4066Sahrens mutex_enter(&zp->z_lock); 1363fa9e4066Sahrens mutex_enter(&vp->v_lock); 1364fa9e4066Sahrens vp->v_count--; 1365fa9e4066Sahrens ASSERT3U(vp->v_count, ==, 0); 1366fa9e4066Sahrens mutex_exit(&vp->v_lock); 1367fa9e4066Sahrens mutex_exit(&zp->z_lock); 1368fa9e4066Sahrens zfs_znode_delete(zp, tx); 1369fa9e4066Sahrens VFS_RELE(zfsvfs->z_vfs); 1370fa9e4066Sahrens } else if (reaped) { 1371fa9e4066Sahrens zfs_dq_add(zp, tx); 1372fa9e4066Sahrens } 1373fa9e4066Sahrens 1374b19a79ecSperrin zfs_log_remove(zilog, tx, TX_REMOVE, dzp, name); 1375fa9e4066Sahrens 1376fa9e4066Sahrens dmu_tx_commit(tx); 1377fa9e4066Sahrens out: 1378fa9e4066Sahrens zfs_dirent_unlock(dl); 1379fa9e4066Sahrens 1380fa9e4066Sahrens if (!delete_now) { 1381fa9e4066Sahrens VN_RELE(vp); 1382fa9e4066Sahrens } else if (xzp) { 1383fa9e4066Sahrens /* this rele delayed to prevent nesting transactions */ 1384fa9e4066Sahrens VN_RELE(ZTOV(xzp)); 1385fa9e4066Sahrens } 1386fa9e4066Sahrens 1387fa9e4066Sahrens ZFS_EXIT(zfsvfs); 1388fa9e4066Sahrens return (error); 1389fa9e4066Sahrens } 1390fa9e4066Sahrens 1391fa9e4066Sahrens /* 1392fa9e4066Sahrens * Create a new directory and insert it into dvp using the name 1393fa9e4066Sahrens * provided. Return a pointer to the inserted directory. 1394fa9e4066Sahrens * 1395fa9e4066Sahrens * IN: dvp - vnode of directory to add subdir to. 1396fa9e4066Sahrens * dirname - name of new directory. 1397fa9e4066Sahrens * vap - attributes of new directory. 1398fa9e4066Sahrens * cr - credentials of caller. 1399fa9e4066Sahrens * 1400fa9e4066Sahrens * OUT: vpp - vnode of created directory. 1401fa9e4066Sahrens * 1402fa9e4066Sahrens * RETURN: 0 if success 1403fa9e4066Sahrens * error code if failure 1404fa9e4066Sahrens * 1405fa9e4066Sahrens * Timestamps: 1406fa9e4066Sahrens * dvp - ctime|mtime updated 1407fa9e4066Sahrens * vp - ctime|mtime|atime updated 1408fa9e4066Sahrens */ 1409fa9e4066Sahrens static int 1410fa9e4066Sahrens zfs_mkdir(vnode_t *dvp, char *dirname, vattr_t *vap, vnode_t **vpp, cred_t *cr) 1411fa9e4066Sahrens { 1412fa9e4066Sahrens znode_t *zp, *dzp = VTOZ(dvp); 1413fa9e4066Sahrens zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 1414fa9e4066Sahrens zilog_t *zilog = zfsvfs->z_log; 1415fa9e4066Sahrens zfs_dirlock_t *dl; 1416fa9e4066Sahrens uint64_t zoid = 0; 1417fa9e4066Sahrens dmu_tx_t *tx; 1418fa9e4066Sahrens int error; 1419fa9e4066Sahrens 1420fa9e4066Sahrens ASSERT(vap->va_type == VDIR); 1421fa9e4066Sahrens 1422fa9e4066Sahrens ZFS_ENTER(zfsvfs); 1423fa9e4066Sahrens 1424fa9e4066Sahrens if (dzp->z_phys->zp_flags & ZFS_XATTR) { 1425fa9e4066Sahrens ZFS_EXIT(zfsvfs); 1426fa9e4066Sahrens return (EINVAL); 1427fa9e4066Sahrens } 1428fa9e4066Sahrens top: 1429fa9e4066Sahrens *vpp = NULL; 1430fa9e4066Sahrens 1431fa9e4066Sahrens /* 1432fa9e4066Sahrens * First make sure the new directory doesn't exist. 1433fa9e4066Sahrens */ 1434fa9e4066Sahrens if (error = zfs_dirent_lock(&dl, dzp, dirname, &zp, ZNEW)) { 1435fa9e4066Sahrens ZFS_EXIT(zfsvfs); 1436fa9e4066Sahrens return (error); 1437fa9e4066Sahrens } 1438fa9e4066Sahrens 1439d2443e76Smarks if (error = zfs_zaccess(dzp, ACE_ADD_SUBDIRECTORY, cr)) { 1440d2443e76Smarks zfs_dirent_unlock(dl); 1441d2443e76Smarks ZFS_EXIT(zfsvfs); 1442d2443e76Smarks return (error); 1443d2443e76Smarks } 1444d2443e76Smarks 1445fa9e4066Sahrens /* 1446fa9e4066Sahrens * Add a new entry to the directory. 1447fa9e4066Sahrens */ 1448fa9e4066Sahrens tx = dmu_tx_create(zfsvfs->z_os); 1449ea8dc4b6Seschrock dmu_tx_hold_zap(tx, dzp->z_id, TRUE, dirname); 1450ea8dc4b6Seschrock dmu_tx_hold_zap(tx, DMU_NEW_OBJECT, FALSE, NULL); 1451fa9e4066Sahrens if (dzp->z_phys->zp_flags & ZFS_INHERIT_ACE) 1452fa9e4066Sahrens dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 1453fa9e4066Sahrens 0, SPA_MAXBLOCKSIZE); 1454fa9e4066Sahrens error = dmu_tx_assign(tx, zfsvfs->z_assign); 1455fa9e4066Sahrens if (error) { 1456fa9e4066Sahrens zfs_dirent_unlock(dl); 1457fa9e4066Sahrens if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) { 14588a2f1b91Sahrens dmu_tx_wait(tx); 14598a2f1b91Sahrens dmu_tx_abort(tx); 1460fa9e4066Sahrens goto top; 1461fa9e4066Sahrens } 14628a2f1b91Sahrens dmu_tx_abort(tx); 1463fa9e4066Sahrens ZFS_EXIT(zfsvfs); 1464fa9e4066Sahrens return (error); 1465fa9e4066Sahrens } 1466fa9e4066Sahrens 1467fa9e4066Sahrens /* 1468fa9e4066Sahrens * Create new node. 1469fa9e4066Sahrens */ 1470fa9e4066Sahrens zfs_mknode(dzp, vap, &zoid, tx, cr, 0, &zp, 0); 1471fa9e4066Sahrens 1472fa9e4066Sahrens /* 1473fa9e4066Sahrens * Now put new name in parent dir. 1474fa9e4066Sahrens */ 1475fa9e4066Sahrens (void) zfs_link_create(dl, zp, tx, ZNEW); 1476fa9e4066Sahrens 1477fa9e4066Sahrens *vpp = ZTOV(zp); 1478fa9e4066Sahrens 1479b19a79ecSperrin zfs_log_create(zilog, tx, TX_MKDIR, dzp, zp, dirname); 1480fa9e4066Sahrens dmu_tx_commit(tx); 1481fa9e4066Sahrens 1482fa9e4066Sahrens zfs_dirent_unlock(dl); 1483fa9e4066Sahrens 1484fa9e4066Sahrens ZFS_EXIT(zfsvfs); 1485fa9e4066Sahrens return (0); 1486fa9e4066Sahrens } 1487fa9e4066Sahrens 1488fa9e4066Sahrens /* 1489fa9e4066Sahrens * Remove a directory subdir entry. If the current working 1490fa9e4066Sahrens * directory is the same as the subdir to be removed, the 1491fa9e4066Sahrens * remove will fail. 1492fa9e4066Sahrens * 1493fa9e4066Sahrens * IN: dvp - vnode of directory to remove from. 1494fa9e4066Sahrens * name - name of directory to be removed. 1495fa9e4066Sahrens * cwd - vnode of current working directory. 1496fa9e4066Sahrens * cr - credentials of caller. 1497fa9e4066Sahrens * 1498fa9e4066Sahrens * RETURN: 0 if success 1499fa9e4066Sahrens * error code if failure 1500fa9e4066Sahrens * 1501fa9e4066Sahrens * Timestamps: 1502fa9e4066Sahrens * dvp - ctime|mtime updated 1503fa9e4066Sahrens */ 1504fa9e4066Sahrens static int 1505fa9e4066Sahrens zfs_rmdir(vnode_t *dvp, char *name, vnode_t *cwd, cred_t *cr) 1506fa9e4066Sahrens { 1507fa9e4066Sahrens znode_t *dzp = VTOZ(dvp); 1508fa9e4066Sahrens znode_t *zp; 1509fa9e4066Sahrens vnode_t *vp; 1510fa9e4066Sahrens zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 1511fa9e4066Sahrens zilog_t *zilog = zfsvfs->z_log; 1512fa9e4066Sahrens zfs_dirlock_t *dl; 1513fa9e4066Sahrens dmu_tx_t *tx; 1514fa9e4066Sahrens int error; 1515fa9e4066Sahrens 1516fa9e4066Sahrens ZFS_ENTER(zfsvfs); 1517fa9e4066Sahrens 1518fa9e4066Sahrens top: 1519fa9e4066Sahrens zp = NULL; 1520fa9e4066Sahrens 1521fa9e4066Sahrens /* 1522fa9e4066Sahrens * Attempt to lock directory; fail if entry doesn't exist. 1523fa9e4066Sahrens */ 1524fa9e4066Sahrens if (error = zfs_dirent_lock(&dl, dzp, name, &zp, ZEXISTS)) { 1525fa9e4066Sahrens ZFS_EXIT(zfsvfs); 1526fa9e4066Sahrens return (error); 1527fa9e4066Sahrens } 1528fa9e4066Sahrens 1529fa9e4066Sahrens vp = ZTOV(zp); 1530fa9e4066Sahrens 1531fa9e4066Sahrens if (error = zfs_zaccess_delete(dzp, zp, cr)) { 1532fa9e4066Sahrens goto out; 1533fa9e4066Sahrens } 1534fa9e4066Sahrens 1535fa9e4066Sahrens if (vp->v_type != VDIR) { 1536fa9e4066Sahrens error = ENOTDIR; 1537fa9e4066Sahrens goto out; 1538fa9e4066Sahrens } 1539fa9e4066Sahrens 1540fa9e4066Sahrens if (vp == cwd) { 1541fa9e4066Sahrens error = EINVAL; 1542fa9e4066Sahrens goto out; 1543fa9e4066Sahrens } 1544fa9e4066Sahrens 1545fa9e4066Sahrens vnevent_rmdir(vp); 1546fa9e4066Sahrens 1547fa9e4066Sahrens /* 1548fa9e4066Sahrens * Grab a lock on the parent pointer make sure we play well 1549fa9e4066Sahrens * with the treewalk and directory rename code. 1550fa9e4066Sahrens */ 1551fa9e4066Sahrens rw_enter(&zp->z_parent_lock, RW_WRITER); 1552fa9e4066Sahrens 1553fa9e4066Sahrens tx = dmu_tx_create(zfsvfs->z_os); 1554ea8dc4b6Seschrock dmu_tx_hold_zap(tx, dzp->z_id, FALSE, name); 1555fa9e4066Sahrens dmu_tx_hold_bonus(tx, zp->z_id); 1556ea8dc4b6Seschrock dmu_tx_hold_zap(tx, zfsvfs->z_dqueue, FALSE, NULL); 1557fa9e4066Sahrens error = dmu_tx_assign(tx, zfsvfs->z_assign); 1558fa9e4066Sahrens if (error) { 1559fa9e4066Sahrens rw_exit(&zp->z_parent_lock); 1560fa9e4066Sahrens zfs_dirent_unlock(dl); 1561fa9e4066Sahrens VN_RELE(vp); 1562fa9e4066Sahrens if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) { 15638a2f1b91Sahrens dmu_tx_wait(tx); 15648a2f1b91Sahrens dmu_tx_abort(tx); 1565fa9e4066Sahrens goto top; 1566fa9e4066Sahrens } 15678a2f1b91Sahrens dmu_tx_abort(tx); 1568fa9e4066Sahrens ZFS_EXIT(zfsvfs); 1569fa9e4066Sahrens return (error); 1570fa9e4066Sahrens } 1571fa9e4066Sahrens 1572fa9e4066Sahrens error = zfs_link_destroy(dl, zp, tx, 0, NULL); 1573fa9e4066Sahrens 1574fa9e4066Sahrens if (error == 0) 1575b19a79ecSperrin zfs_log_remove(zilog, tx, TX_RMDIR, dzp, name); 1576fa9e4066Sahrens 1577fa9e4066Sahrens dmu_tx_commit(tx); 1578fa9e4066Sahrens 1579fa9e4066Sahrens rw_exit(&zp->z_parent_lock); 1580fa9e4066Sahrens out: 1581fa9e4066Sahrens zfs_dirent_unlock(dl); 1582fa9e4066Sahrens 1583fa9e4066Sahrens VN_RELE(vp); 1584fa9e4066Sahrens 1585fa9e4066Sahrens ZFS_EXIT(zfsvfs); 1586fa9e4066Sahrens return (error); 1587fa9e4066Sahrens } 1588fa9e4066Sahrens 1589fa9e4066Sahrens /* 1590fa9e4066Sahrens * Read as many directory entries as will fit into the provided 1591fa9e4066Sahrens * buffer from the given directory cursor position (specified in 1592fa9e4066Sahrens * the uio structure. 1593fa9e4066Sahrens * 1594fa9e4066Sahrens * IN: vp - vnode of directory to read. 1595fa9e4066Sahrens * uio - structure supplying read location, range info, 1596fa9e4066Sahrens * and return buffer. 1597fa9e4066Sahrens * cr - credentials of caller. 1598fa9e4066Sahrens * 1599fa9e4066Sahrens * OUT: uio - updated offset and range, buffer filled. 1600fa9e4066Sahrens * eofp - set to true if end-of-file detected. 1601fa9e4066Sahrens * 1602fa9e4066Sahrens * RETURN: 0 if success 1603fa9e4066Sahrens * error code if failure 1604fa9e4066Sahrens * 1605fa9e4066Sahrens * Timestamps: 1606fa9e4066Sahrens * vp - atime updated 1607fa9e4066Sahrens * 1608fa9e4066Sahrens * Note that the low 4 bits of the cookie returned by zap is always zero. 1609fa9e4066Sahrens * This allows us to use the low range for "special" directory entries: 1610fa9e4066Sahrens * We use 0 for '.', and 1 for '..'. If this is the root of the filesystem, 1611fa9e4066Sahrens * we use the offset 2 for the '.zfs' directory. 1612fa9e4066Sahrens */ 1613fa9e4066Sahrens /* ARGSUSED */ 1614fa9e4066Sahrens static int 1615fa9e4066Sahrens zfs_readdir(vnode_t *vp, uio_t *uio, cred_t *cr, int *eofp) 1616fa9e4066Sahrens { 1617fa9e4066Sahrens znode_t *zp = VTOZ(vp); 1618fa9e4066Sahrens iovec_t *iovp; 1619fa9e4066Sahrens dirent64_t *odp; 1620fa9e4066Sahrens zfsvfs_t *zfsvfs = zp->z_zfsvfs; 16217f6e3e7dSperrin objset_t *os; 1622fa9e4066Sahrens caddr_t outbuf; 1623fa9e4066Sahrens size_t bufsize; 1624fa9e4066Sahrens zap_cursor_t zc; 1625fa9e4066Sahrens zap_attribute_t zap; 1626fa9e4066Sahrens uint_t bytes_wanted; 1627fa9e4066Sahrens ushort_t this_reclen; 1628fa9e4066Sahrens uint64_t offset; /* must be unsigned; checks for < 1 */ 1629fa9e4066Sahrens off64_t *next; 1630fa9e4066Sahrens int local_eof; 16317f6e3e7dSperrin int outcount; 16327f6e3e7dSperrin int error; 16337f6e3e7dSperrin uint8_t prefetch; 1634fa9e4066Sahrens 1635fa9e4066Sahrens ZFS_ENTER(zfsvfs); 1636fa9e4066Sahrens 1637fa9e4066Sahrens /* 1638fa9e4066Sahrens * If we are not given an eof variable, 1639fa9e4066Sahrens * use a local one. 1640fa9e4066Sahrens */ 1641fa9e4066Sahrens if (eofp == NULL) 1642fa9e4066Sahrens eofp = &local_eof; 1643fa9e4066Sahrens 1644fa9e4066Sahrens /* 1645fa9e4066Sahrens * Check for valid iov_len. 1646fa9e4066Sahrens */ 1647fa9e4066Sahrens if (uio->uio_iov->iov_len <= 0) { 1648fa9e4066Sahrens ZFS_EXIT(zfsvfs); 1649fa9e4066Sahrens return (EINVAL); 1650fa9e4066Sahrens } 1651fa9e4066Sahrens 1652fa9e4066Sahrens /* 1653fa9e4066Sahrens * Quit if directory has been removed (posix) 1654fa9e4066Sahrens */ 1655fa9e4066Sahrens if ((*eofp = zp->z_reap) != 0) { 1656fa9e4066Sahrens ZFS_EXIT(zfsvfs); 1657fa9e4066Sahrens return (0); 1658fa9e4066Sahrens } 1659fa9e4066Sahrens 16607f6e3e7dSperrin error = 0; 16617f6e3e7dSperrin os = zfsvfs->z_os; 16627f6e3e7dSperrin offset = uio->uio_loffset; 16637f6e3e7dSperrin prefetch = zp->z_zn_prefetch; 16647f6e3e7dSperrin 1665fa9e4066Sahrens /* 1666fa9e4066Sahrens * Initialize the iterator cursor. 1667fa9e4066Sahrens */ 1668fa9e4066Sahrens if (offset <= 3) { 1669fa9e4066Sahrens /* 1670fa9e4066Sahrens * Start iteration from the beginning of the directory. 1671fa9e4066Sahrens */ 16727f6e3e7dSperrin zap_cursor_init(&zc, os, zp->z_id); 1673fa9e4066Sahrens } else { 1674fa9e4066Sahrens /* 1675fa9e4066Sahrens * The offset is a serialized cursor. 1676fa9e4066Sahrens */ 16777f6e3e7dSperrin zap_cursor_init_serialized(&zc, os, zp->z_id, offset); 1678fa9e4066Sahrens } 1679fa9e4066Sahrens 1680fa9e4066Sahrens /* 1681fa9e4066Sahrens * Get space to change directory entries into fs independent format. 1682fa9e4066Sahrens */ 1683fa9e4066Sahrens iovp = uio->uio_iov; 1684fa9e4066Sahrens bytes_wanted = iovp->iov_len; 1685fa9e4066Sahrens if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1) { 1686fa9e4066Sahrens bufsize = bytes_wanted; 1687fa9e4066Sahrens outbuf = kmem_alloc(bufsize, KM_SLEEP); 1688fa9e4066Sahrens odp = (struct dirent64 *)outbuf; 1689fa9e4066Sahrens } else { 1690fa9e4066Sahrens bufsize = bytes_wanted; 1691fa9e4066Sahrens odp = (struct dirent64 *)iovp->iov_base; 1692fa9e4066Sahrens } 1693fa9e4066Sahrens 1694fa9e4066Sahrens /* 1695fa9e4066Sahrens * Transform to file-system independent format 1696fa9e4066Sahrens */ 1697fa9e4066Sahrens outcount = 0; 1698fa9e4066Sahrens while (outcount < bytes_wanted) { 1699fa9e4066Sahrens /* 1700fa9e4066Sahrens * Special case `.', `..', and `.zfs'. 1701fa9e4066Sahrens */ 1702fa9e4066Sahrens if (offset == 0) { 1703fa9e4066Sahrens (void) strcpy(zap.za_name, "."); 1704fa9e4066Sahrens zap.za_first_integer = zp->z_id; 1705fa9e4066Sahrens this_reclen = DIRENT64_RECLEN(1); 1706fa9e4066Sahrens } else if (offset == 1) { 1707fa9e4066Sahrens (void) strcpy(zap.za_name, ".."); 1708fa9e4066Sahrens zap.za_first_integer = zp->z_phys->zp_parent; 1709fa9e4066Sahrens this_reclen = DIRENT64_RECLEN(2); 1710fa9e4066Sahrens } else if (offset == 2 && zfs_show_ctldir(zp)) { 1711fa9e4066Sahrens (void) strcpy(zap.za_name, ZFS_CTLDIR_NAME); 1712fa9e4066Sahrens zap.za_first_integer = ZFSCTL_INO_ROOT; 1713fa9e4066Sahrens this_reclen = 1714fa9e4066Sahrens DIRENT64_RECLEN(sizeof (ZFS_CTLDIR_NAME) - 1); 1715fa9e4066Sahrens } else { 1716fa9e4066Sahrens /* 1717fa9e4066Sahrens * Grab next entry. 1718fa9e4066Sahrens */ 1719fa9e4066Sahrens if (error = zap_cursor_retrieve(&zc, &zap)) { 1720fa9e4066Sahrens if ((*eofp = (error == ENOENT)) != 0) 1721fa9e4066Sahrens break; 1722fa9e4066Sahrens else 1723fa9e4066Sahrens goto update; 1724fa9e4066Sahrens } 1725fa9e4066Sahrens 1726fa9e4066Sahrens if (zap.za_integer_length != 8 || 1727fa9e4066Sahrens zap.za_num_integers != 1) { 1728fa9e4066Sahrens cmn_err(CE_WARN, "zap_readdir: bad directory " 1729fa9e4066Sahrens "entry, obj = %lld, offset = %lld\n", 1730fa9e4066Sahrens (u_longlong_t)zp->z_id, 1731fa9e4066Sahrens (u_longlong_t)offset); 1732fa9e4066Sahrens error = ENXIO; 1733fa9e4066Sahrens goto update; 1734fa9e4066Sahrens } 1735fa9e4066Sahrens this_reclen = DIRENT64_RECLEN(strlen(zap.za_name)); 1736fa9e4066Sahrens } 1737fa9e4066Sahrens 1738fa9e4066Sahrens /* 1739fa9e4066Sahrens * Will this entry fit in the buffer? 1740fa9e4066Sahrens */ 1741fa9e4066Sahrens if (outcount + this_reclen > bufsize) { 1742fa9e4066Sahrens /* 1743fa9e4066Sahrens * Did we manage to fit anything in the buffer? 1744fa9e4066Sahrens */ 1745fa9e4066Sahrens if (!outcount) { 1746fa9e4066Sahrens error = EINVAL; 1747fa9e4066Sahrens goto update; 1748fa9e4066Sahrens } 1749fa9e4066Sahrens break; 1750fa9e4066Sahrens } 1751fa9e4066Sahrens /* 1752fa9e4066Sahrens * Add this entry: 1753fa9e4066Sahrens */ 1754fa9e4066Sahrens odp->d_ino = (ino64_t)zap.za_first_integer; 1755fa9e4066Sahrens odp->d_reclen = (ushort_t)this_reclen; 1756fa9e4066Sahrens /* NOTE: d_off is the offset for the *next* entry */ 1757fa9e4066Sahrens next = &(odp->d_off); 1758fa9e4066Sahrens (void) strncpy(odp->d_name, zap.za_name, 1759fa9e4066Sahrens DIRENT64_NAMELEN(this_reclen)); 1760fa9e4066Sahrens outcount += this_reclen; 1761fa9e4066Sahrens odp = (dirent64_t *)((intptr_t)odp + this_reclen); 1762fa9e4066Sahrens 1763fa9e4066Sahrens ASSERT(outcount <= bufsize); 1764fa9e4066Sahrens 1765fa9e4066Sahrens /* Prefetch znode */ 17667f6e3e7dSperrin if (prefetch) 17677f6e3e7dSperrin dmu_prefetch(os, zap.za_first_integer, 0, 0); 1768fa9e4066Sahrens 1769fa9e4066Sahrens /* 1770fa9e4066Sahrens * Move to the next entry, fill in the previous offset. 1771fa9e4066Sahrens */ 1772fa9e4066Sahrens if (offset > 2 || (offset == 2 && !zfs_show_ctldir(zp))) { 1773fa9e4066Sahrens zap_cursor_advance(&zc); 1774fa9e4066Sahrens offset = zap_cursor_serialize(&zc); 1775fa9e4066Sahrens } else { 1776fa9e4066Sahrens offset += 1; 1777fa9e4066Sahrens } 1778fa9e4066Sahrens *next = offset; 1779fa9e4066Sahrens } 17807f6e3e7dSperrin zp->z_zn_prefetch = B_FALSE; /* a lookup will re-enable pre-fetching */ 1781fa9e4066Sahrens 1782fa9e4066Sahrens if (uio->uio_segflg == UIO_SYSSPACE && uio->uio_iovcnt == 1) { 1783fa9e4066Sahrens iovp->iov_base += outcount; 1784fa9e4066Sahrens iovp->iov_len -= outcount; 1785fa9e4066Sahrens uio->uio_resid -= outcount; 1786fa9e4066Sahrens } else if (error = uiomove(outbuf, (long)outcount, UIO_READ, uio)) { 1787fa9e4066Sahrens /* 1788fa9e4066Sahrens * Reset the pointer. 1789fa9e4066Sahrens */ 1790fa9e4066Sahrens offset = uio->uio_loffset; 1791fa9e4066Sahrens } 1792fa9e4066Sahrens 1793fa9e4066Sahrens update: 179487e5029aSahrens zap_cursor_fini(&zc); 1795fa9e4066Sahrens if (uio->uio_segflg != UIO_SYSSPACE || uio->uio_iovcnt != 1) 1796fa9e4066Sahrens kmem_free(outbuf, bufsize); 1797fa9e4066Sahrens 1798fa9e4066Sahrens if (error == ENOENT) 1799fa9e4066Sahrens error = 0; 1800fa9e4066Sahrens 1801fa9e4066Sahrens ZFS_ACCESSTIME_STAMP(zfsvfs, zp); 1802fa9e4066Sahrens 1803fa9e4066Sahrens uio->uio_loffset = offset; 1804fa9e4066Sahrens ZFS_EXIT(zfsvfs); 1805fa9e4066Sahrens return (error); 1806fa9e4066Sahrens } 1807fa9e4066Sahrens 1808fa9e4066Sahrens static int 1809fa9e4066Sahrens zfs_fsync(vnode_t *vp, int syncflag, cred_t *cr) 1810fa9e4066Sahrens { 1811fa9e4066Sahrens znode_t *zp = VTOZ(vp); 1812fa9e4066Sahrens zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1813fa9e4066Sahrens 1814b468a217Seschrock /* 1815b468a217Seschrock * Regardless of whether this is required for standards conformance, 1816b468a217Seschrock * this is the logical behavior when fsync() is called on a file with 1817b468a217Seschrock * dirty pages. We use B_ASYNC since the ZIL transactions are already 1818b468a217Seschrock * going to be pushed out as part of the zil_commit(). 1819b468a217Seschrock */ 1820b468a217Seschrock if (vn_has_cached_data(vp) && !(syncflag & FNODSYNC) && 1821b468a217Seschrock (vp->v_type == VREG) && !(IS_SWAPVP(vp))) 1822b468a217Seschrock (void) VOP_PUTPAGE(vp, (offset_t)0, (size_t)0, B_ASYNC, cr); 1823b468a217Seschrock 1824fa9e4066Sahrens ZFS_ENTER(zfsvfs); 1825b19a79ecSperrin zil_commit(zfsvfs->z_log, zp->z_last_itx, zp->z_id); 1826fa9e4066Sahrens ZFS_EXIT(zfsvfs); 1827fa9e4066Sahrens return (0); 1828fa9e4066Sahrens } 1829fa9e4066Sahrens 1830fa9e4066Sahrens /* 1831fa9e4066Sahrens * Get the requested file attributes and place them in the provided 1832fa9e4066Sahrens * vattr structure. 1833fa9e4066Sahrens * 1834fa9e4066Sahrens * IN: vp - vnode of file. 1835fa9e4066Sahrens * vap - va_mask identifies requested attributes. 1836fa9e4066Sahrens * flags - [UNUSED] 1837fa9e4066Sahrens * cr - credentials of caller. 1838fa9e4066Sahrens * 1839fa9e4066Sahrens * OUT: vap - attribute values. 1840fa9e4066Sahrens * 1841fa9e4066Sahrens * RETURN: 0 (always succeeds) 1842fa9e4066Sahrens */ 1843fa9e4066Sahrens /* ARGSUSED */ 1844fa9e4066Sahrens static int 1845fa9e4066Sahrens zfs_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr) 1846fa9e4066Sahrens { 1847fa9e4066Sahrens znode_t *zp = VTOZ(vp); 1848fa9e4066Sahrens zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1849fa9e4066Sahrens znode_phys_t *pzp = zp->z_phys; 1850fa9e4066Sahrens int error; 1851fa9e4066Sahrens 1852fa9e4066Sahrens ZFS_ENTER(zfsvfs); 1853fa9e4066Sahrens 1854fa9e4066Sahrens /* 1855fa9e4066Sahrens * Return all attributes. It's cheaper to provide the answer 1856fa9e4066Sahrens * than to determine whether we were asked the question. 1857fa9e4066Sahrens */ 1858fa9e4066Sahrens mutex_enter(&zp->z_lock); 1859fa9e4066Sahrens 1860fa9e4066Sahrens vap->va_type = vp->v_type; 1861fa9e4066Sahrens vap->va_mode = pzp->zp_mode & MODEMASK; 1862fa9e4066Sahrens vap->va_uid = zp->z_phys->zp_uid; 1863fa9e4066Sahrens vap->va_gid = zp->z_phys->zp_gid; 1864fa9e4066Sahrens vap->va_fsid = zp->z_zfsvfs->z_vfs->vfs_dev; 1865fa9e4066Sahrens vap->va_nodeid = zp->z_id; 1866fa9e4066Sahrens vap->va_nlink = MIN(pzp->zp_links, UINT32_MAX); /* nlink_t limit! */ 1867fa9e4066Sahrens vap->va_size = pzp->zp_size; 186872fc53bcSmarks vap->va_rdev = vp->v_rdev; 1869fa9e4066Sahrens vap->va_seq = zp->z_seq; 1870fa9e4066Sahrens 1871fa9e4066Sahrens ZFS_TIME_DECODE(&vap->va_atime, pzp->zp_atime); 1872fa9e4066Sahrens ZFS_TIME_DECODE(&vap->va_mtime, pzp->zp_mtime); 1873fa9e4066Sahrens ZFS_TIME_DECODE(&vap->va_ctime, pzp->zp_ctime); 1874fa9e4066Sahrens 1875fa9e4066Sahrens /* 1876de122929Smarks * If ACL is trivial don't bother looking for ACE_READ_ATTRIBUTES. 1877de122929Smarks * Also, if we are the owner don't bother, since owner should 1878de122929Smarks * always be allowed to read basic attributes of file. 1879fa9e4066Sahrens */ 1880de122929Smarks if (!(zp->z_phys->zp_flags & ZFS_ACL_TRIVIAL) && 1881de122929Smarks (zp->z_phys->zp_uid != crgetuid(cr))) { 1882de122929Smarks if (error = zfs_zaccess(zp, ACE_READ_ATTRIBUTES, cr)) { 1883fa9e4066Sahrens mutex_exit(&zp->z_lock); 1884fa9e4066Sahrens ZFS_EXIT(zfsvfs); 1885fa9e4066Sahrens return (error); 1886fa9e4066Sahrens } 1887fa9e4066Sahrens } 1888fa9e4066Sahrens 1889fa9e4066Sahrens mutex_exit(&zp->z_lock); 1890fa9e4066Sahrens 1891fa9e4066Sahrens dmu_object_size_from_db(zp->z_dbuf, &vap->va_blksize, &vap->va_nblocks); 1892fa9e4066Sahrens 1893fa9e4066Sahrens if (zp->z_blksz == 0) { 1894fa9e4066Sahrens /* 1895fa9e4066Sahrens * Block size hasn't been set; suggest maximal I/O transfers. 1896fa9e4066Sahrens */ 1897fa9e4066Sahrens vap->va_blksize = zfsvfs->z_max_blksz; 1898fa9e4066Sahrens } 1899fa9e4066Sahrens 1900fa9e4066Sahrens ZFS_EXIT(zfsvfs); 1901fa9e4066Sahrens return (0); 1902fa9e4066Sahrens } 1903fa9e4066Sahrens 1904fa9e4066Sahrens /* 1905fa9e4066Sahrens * Set the file attributes to the values contained in the 1906fa9e4066Sahrens * vattr structure. 1907fa9e4066Sahrens * 1908fa9e4066Sahrens * IN: vp - vnode of file to be modified. 1909fa9e4066Sahrens * vap - new attribute values. 1910fa9e4066Sahrens * flags - ATTR_UTIME set if non-default time values provided. 1911fa9e4066Sahrens * cr - credentials of caller. 1912fa9e4066Sahrens * 1913fa9e4066Sahrens * RETURN: 0 if success 1914fa9e4066Sahrens * error code if failure 1915fa9e4066Sahrens * 1916fa9e4066Sahrens * Timestamps: 1917fa9e4066Sahrens * vp - ctime updated, mtime updated if size changed. 1918fa9e4066Sahrens */ 1919fa9e4066Sahrens /* ARGSUSED */ 1920fa9e4066Sahrens static int 1921fa9e4066Sahrens zfs_setattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr, 1922fa9e4066Sahrens caller_context_t *ct) 1923fa9e4066Sahrens { 1924fa9e4066Sahrens struct znode *zp = VTOZ(vp); 1925fa9e4066Sahrens znode_phys_t *pzp = zp->z_phys; 1926fa9e4066Sahrens zfsvfs_t *zfsvfs = zp->z_zfsvfs; 1927fa9e4066Sahrens zilog_t *zilog = zfsvfs->z_log; 1928fa9e4066Sahrens dmu_tx_t *tx; 1929fa9e4066Sahrens vattr_t oldva; 19305730cc9aSmaybee uint_t mask = vap->va_mask; 19315730cc9aSmaybee uint_t saved_mask; 193213f9f30eSmarks int trim_mask = FALSE; 1933fa9e4066Sahrens uint64_t new_mode; 1934d2443e76Smarks znode_t *attrzp; 1935fa9e4066Sahrens int need_policy = FALSE; 1936fa9e4066Sahrens int err; 1937fa9e4066Sahrens 1938fa9e4066Sahrens if (mask == 0) 1939fa9e4066Sahrens return (0); 1940fa9e4066Sahrens 1941fa9e4066Sahrens if (mask & AT_NOSET) 1942fa9e4066Sahrens return (EINVAL); 1943fa9e4066Sahrens 1944fa9e4066Sahrens if (mask & AT_SIZE && vp->v_type == VDIR) 1945fa9e4066Sahrens return (EISDIR); 1946fa9e4066Sahrens 19478ef9dde0Smarks if (mask & AT_SIZE && vp->v_type != VREG && vp->v_type != VFIFO) 194884c5a155Smarks return (EINVAL); 194984c5a155Smarks 1950fa9e4066Sahrens ZFS_ENTER(zfsvfs); 1951fa9e4066Sahrens 1952fa9e4066Sahrens top: 1953d2443e76Smarks attrzp = NULL; 1954fa9e4066Sahrens 1955fa9e4066Sahrens if (zfsvfs->z_vfs->vfs_flag & VFS_RDONLY) { 1956fa9e4066Sahrens ZFS_EXIT(zfsvfs); 1957fa9e4066Sahrens return (EROFS); 1958fa9e4066Sahrens } 1959fa9e4066Sahrens 1960fa9e4066Sahrens /* 1961fa9e4066Sahrens * First validate permissions 1962fa9e4066Sahrens */ 1963fa9e4066Sahrens 1964fa9e4066Sahrens if (mask & AT_SIZE) { 1965fa9e4066Sahrens err = zfs_zaccess(zp, ACE_WRITE_DATA, cr); 1966fa9e4066Sahrens if (err) { 1967fa9e4066Sahrens ZFS_EXIT(zfsvfs); 1968fa9e4066Sahrens return (err); 1969fa9e4066Sahrens } 19705730cc9aSmaybee /* 19715730cc9aSmaybee * XXX - Note, we are not providing any open 19725730cc9aSmaybee * mode flags here (like FNDELAY), so we may 19735730cc9aSmaybee * block if there are locks present... this 19745730cc9aSmaybee * should be addressed in openat(). 19755730cc9aSmaybee */ 19765730cc9aSmaybee do { 19775730cc9aSmaybee err = zfs_freesp(zp, vap->va_size, 0, 0, FALSE); 19788a2f1b91Sahrens /* NB: we already did dmu_tx_wait() if necessary */ 19795730cc9aSmaybee } while (err == ERESTART && zfsvfs->z_assign == TXG_NOWAIT); 19805730cc9aSmaybee if (err) { 19815730cc9aSmaybee ZFS_EXIT(zfsvfs); 19825730cc9aSmaybee return (err); 19835730cc9aSmaybee } 1984fa9e4066Sahrens } 1985fa9e4066Sahrens 1986fa9e4066Sahrens if (mask & (AT_ATIME|AT_MTIME)) 1987fa9e4066Sahrens need_policy = zfs_zaccess_v4_perm(zp, ACE_WRITE_ATTRIBUTES, cr); 1988fa9e4066Sahrens 1989fa9e4066Sahrens if (mask & (AT_UID|AT_GID)) { 1990fa9e4066Sahrens int idmask = (mask & (AT_UID|AT_GID)); 1991fa9e4066Sahrens int take_owner; 1992fa9e4066Sahrens int take_group; 1993fa9e4066Sahrens 1994a933bc41Smarks /* 1995a933bc41Smarks * NOTE: even if a new mode is being set, 1996a933bc41Smarks * we may clear S_ISUID/S_ISGID bits. 1997a933bc41Smarks */ 1998a933bc41Smarks 1999a933bc41Smarks if (!(mask & AT_MODE)) 2000a933bc41Smarks vap->va_mode = pzp->zp_mode; 2001a933bc41Smarks 2002fa9e4066Sahrens /* 2003fa9e4066Sahrens * Take ownership or chgrp to group we are a member of 2004fa9e4066Sahrens */ 2005fa9e4066Sahrens 2006fa9e4066Sahrens take_owner = (mask & AT_UID) && (vap->va_uid == crgetuid(cr)); 2007fa9e4066Sahrens take_group = (mask & AT_GID) && groupmember(vap->va_gid, cr); 2008fa9e4066Sahrens 2009fa9e4066Sahrens /* 2010fa9e4066Sahrens * If both AT_UID and AT_GID are set then take_owner and 2011fa9e4066Sahrens * take_group must both be set in order to allow taking 2012fa9e4066Sahrens * ownership. 2013fa9e4066Sahrens * 2014fa9e4066Sahrens * Otherwise, send the check through secpolicy_vnode_setattr() 2015fa9e4066Sahrens * 2016fa9e4066Sahrens */ 2017fa9e4066Sahrens 2018fa9e4066Sahrens if (((idmask == (AT_UID|AT_GID)) && take_owner && take_group) || 2019fa9e4066Sahrens ((idmask == AT_UID) && take_owner) || 2020fa9e4066Sahrens ((idmask == AT_GID) && take_group)) { 2021fa9e4066Sahrens if (zfs_zaccess_v4_perm(zp, ACE_WRITE_OWNER, cr) == 0) { 2022fa9e4066Sahrens /* 2023fa9e4066Sahrens * Remove setuid/setgid for non-privileged users 2024fa9e4066Sahrens */ 202513f9f30eSmarks secpolicy_setid_clear(vap, cr); 202613f9f30eSmarks trim_mask = TRUE; 202713f9f30eSmarks saved_mask = vap->va_mask; 2028fa9e4066Sahrens } else { 2029fa9e4066Sahrens need_policy = TRUE; 2030fa9e4066Sahrens } 2031fa9e4066Sahrens } else { 2032fa9e4066Sahrens need_policy = TRUE; 2033fa9e4066Sahrens } 2034fa9e4066Sahrens } 2035fa9e4066Sahrens 2036fa9e4066Sahrens if (mask & AT_MODE) 2037fa9e4066Sahrens need_policy = TRUE; 2038fa9e4066Sahrens 2039fa9e4066Sahrens if (need_policy) { 2040fa9e4066Sahrens mutex_enter(&zp->z_lock); 2041fa9e4066Sahrens oldva.va_mode = pzp->zp_mode; 2042fa9e4066Sahrens oldva.va_uid = zp->z_phys->zp_uid; 2043fa9e4066Sahrens oldva.va_gid = zp->z_phys->zp_gid; 2044fa9e4066Sahrens mutex_exit(&zp->z_lock); 204513f9f30eSmarks 204613f9f30eSmarks /* 204713f9f30eSmarks * If trim_mask is set then take ownership 204813f9f30eSmarks * has been granted. In that case remove 204913f9f30eSmarks * UID|GID from mask so that 205013f9f30eSmarks * secpolicy_vnode_setattr() doesn't revoke it. 205113f9f30eSmarks */ 205213f9f30eSmarks if (trim_mask) 205313f9f30eSmarks vap->va_mask &= ~(AT_UID|AT_GID); 205413f9f30eSmarks 2055fa9e4066Sahrens err = secpolicy_vnode_setattr(cr, vp, vap, &oldva, flags, 2056fa9e4066Sahrens (int (*)(void *, int, cred_t *))zfs_zaccess_rwx, zp); 2057fa9e4066Sahrens if (err) { 2058fa9e4066Sahrens ZFS_EXIT(zfsvfs); 2059fa9e4066Sahrens return (err); 2060fa9e4066Sahrens } 206113f9f30eSmarks 206213f9f30eSmarks if (trim_mask) 206313f9f30eSmarks vap->va_mask |= (saved_mask & (AT_UID|AT_GID)); 2064fa9e4066Sahrens } 2065fa9e4066Sahrens 2066fa9e4066Sahrens /* 2067fa9e4066Sahrens * secpolicy_vnode_setattr, or take ownership may have 2068fa9e4066Sahrens * changed va_mask 2069fa9e4066Sahrens */ 2070fa9e4066Sahrens mask = vap->va_mask; 2071fa9e4066Sahrens 2072fa9e4066Sahrens tx = dmu_tx_create(zfsvfs->z_os); 2073fa9e4066Sahrens dmu_tx_hold_bonus(tx, zp->z_id); 2074fa9e4066Sahrens 2075fa9e4066Sahrens if (mask & AT_MODE) { 2076169cdae2Smarks uint64_t pmode = pzp->zp_mode; 2077fa9e4066Sahrens 2078169cdae2Smarks new_mode = (pmode & S_IFMT) | (vap->va_mode & ~S_IFMT); 2079fa9e4066Sahrens 2080fa9e4066Sahrens if (zp->z_phys->zp_acl.z_acl_extern_obj) 2081fa9e4066Sahrens dmu_tx_hold_write(tx, 2082fa9e4066Sahrens pzp->zp_acl.z_acl_extern_obj, 0, SPA_MAXBLOCKSIZE); 2083fa9e4066Sahrens else 2084fa9e4066Sahrens dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 2085fa9e4066Sahrens 0, ZFS_ACL_SIZE(MAX_ACL_SIZE)); 2086fa9e4066Sahrens } 2087fa9e4066Sahrens 2088d2443e76Smarks if ((mask & (AT_UID | AT_GID)) && zp->z_phys->zp_xattr != 0) { 2089d2443e76Smarks err = zfs_zget(zp->z_zfsvfs, zp->z_phys->zp_xattr, &attrzp); 2090d2443e76Smarks if (err) { 2091d2443e76Smarks dmu_tx_abort(tx); 2092d2443e76Smarks ZFS_EXIT(zfsvfs); 2093d2443e76Smarks return (err); 2094d2443e76Smarks } 2095d2443e76Smarks dmu_tx_hold_bonus(tx, attrzp->z_id); 2096d2443e76Smarks } 2097d2443e76Smarks 2098fa9e4066Sahrens err = dmu_tx_assign(tx, zfsvfs->z_assign); 2099fa9e4066Sahrens if (err) { 2100d2443e76Smarks if (attrzp) 2101d2443e76Smarks VN_RELE(ZTOV(attrzp)); 2102fa9e4066Sahrens if (err == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) { 21038a2f1b91Sahrens dmu_tx_wait(tx); 21048a2f1b91Sahrens dmu_tx_abort(tx); 2105fa9e4066Sahrens goto top; 2106fa9e4066Sahrens } 21078a2f1b91Sahrens dmu_tx_abort(tx); 2108fa9e4066Sahrens ZFS_EXIT(zfsvfs); 2109fa9e4066Sahrens return (err); 2110fa9e4066Sahrens } 2111fa9e4066Sahrens 2112fa9e4066Sahrens dmu_buf_will_dirty(zp->z_dbuf, tx); 2113fa9e4066Sahrens 2114fa9e4066Sahrens /* 2115fa9e4066Sahrens * Set each attribute requested. 2116fa9e4066Sahrens * We group settings according to the locks they need to acquire. 2117fa9e4066Sahrens * 2118fa9e4066Sahrens * Note: you cannot set ctime directly, although it will be 2119fa9e4066Sahrens * updated as a side-effect of calling this function. 2120fa9e4066Sahrens */ 2121fa9e4066Sahrens 2122fa9e4066Sahrens mutex_enter(&zp->z_lock); 2123fa9e4066Sahrens 2124fa9e4066Sahrens if (mask & AT_MODE) { 2125fa9e4066Sahrens err = zfs_acl_chmod_setattr(zp, new_mode, tx); 2126fa9e4066Sahrens ASSERT3U(err, ==, 0); 2127fa9e4066Sahrens } 2128fa9e4066Sahrens 2129d2443e76Smarks if (attrzp) 2130d2443e76Smarks mutex_enter(&attrzp->z_lock); 2131d2443e76Smarks 2132d2443e76Smarks if (mask & AT_UID) { 2133fa9e4066Sahrens zp->z_phys->zp_uid = (uint64_t)vap->va_uid; 2134d2443e76Smarks if (attrzp) { 2135d2443e76Smarks attrzp->z_phys->zp_uid = (uint64_t)vap->va_uid; 2136d2443e76Smarks } 2137d2443e76Smarks } 2138fa9e4066Sahrens 2139d2443e76Smarks if (mask & AT_GID) { 2140fa9e4066Sahrens zp->z_phys->zp_gid = (uint64_t)vap->va_gid; 2141d2443e76Smarks if (attrzp) 2142d2443e76Smarks attrzp->z_phys->zp_gid = (uint64_t)vap->va_gid; 2143d2443e76Smarks } 2144d2443e76Smarks 2145d2443e76Smarks if (attrzp) 2146d2443e76Smarks mutex_exit(&attrzp->z_lock); 2147fa9e4066Sahrens 2148fa9e4066Sahrens if (mask & AT_ATIME) 2149fa9e4066Sahrens ZFS_TIME_ENCODE(&vap->va_atime, pzp->zp_atime); 2150fa9e4066Sahrens 2151fa9e4066Sahrens if (mask & AT_MTIME) 2152fa9e4066Sahrens ZFS_TIME_ENCODE(&vap->va_mtime, pzp->zp_mtime); 2153fa9e4066Sahrens 21545730cc9aSmaybee if (mask & AT_SIZE) 2155fa9e4066Sahrens zfs_time_stamper_locked(zp, CONTENT_MODIFIED, tx); 21565730cc9aSmaybee else if (mask != 0) 2157fa9e4066Sahrens zfs_time_stamper_locked(zp, STATE_CHANGED, tx); 2158fa9e4066Sahrens 21595730cc9aSmaybee if (mask != 0) 2160b19a79ecSperrin zfs_log_setattr(zilog, tx, TX_SETATTR, zp, vap, mask); 2161fa9e4066Sahrens 2162fa9e4066Sahrens mutex_exit(&zp->z_lock); 2163fa9e4066Sahrens 2164d2443e76Smarks if (attrzp) 2165d2443e76Smarks VN_RELE(ZTOV(attrzp)); 2166d2443e76Smarks 2167fa9e4066Sahrens dmu_tx_commit(tx); 2168fa9e4066Sahrens 2169fa9e4066Sahrens ZFS_EXIT(zfsvfs); 2170fa9e4066Sahrens return (err); 2171fa9e4066Sahrens } 2172fa9e4066Sahrens 2173fa9e4066Sahrens /* 2174fa9e4066Sahrens * Search back through the directory tree, using the ".." entries. 2175fa9e4066Sahrens * Lock each directory in the chain to prevent concurrent renames. 2176fa9e4066Sahrens * Fail any attempt to move a directory into one of its own descendants. 2177fa9e4066Sahrens * XXX - z_parent_lock can overlap with map or grow locks 2178fa9e4066Sahrens */ 2179fa9e4066Sahrens typedef struct zfs_zlock { 2180fa9e4066Sahrens krwlock_t *zl_rwlock; /* lock we acquired */ 2181fa9e4066Sahrens znode_t *zl_znode; /* znode we held */ 2182fa9e4066Sahrens struct zfs_zlock *zl_next; /* next in list */ 2183fa9e4066Sahrens } zfs_zlock_t; 2184fa9e4066Sahrens 2185fa9e4066Sahrens static int 2186fa9e4066Sahrens zfs_rename_lock(znode_t *szp, znode_t *tdzp, znode_t *sdzp, zfs_zlock_t **zlpp) 2187fa9e4066Sahrens { 2188fa9e4066Sahrens zfs_zlock_t *zl; 2189fa9e4066Sahrens znode_t *zp = tdzp; 2190fa9e4066Sahrens uint64_t rootid = zp->z_zfsvfs->z_root; 2191fa9e4066Sahrens uint64_t *oidp = &zp->z_id; 2192fa9e4066Sahrens krwlock_t *rwlp = &szp->z_parent_lock; 2193fa9e4066Sahrens krw_t rw = RW_WRITER; 2194fa9e4066Sahrens 2195fa9e4066Sahrens /* 2196fa9e4066Sahrens * First pass write-locks szp and compares to zp->z_id. 2197fa9e4066Sahrens * Later passes read-lock zp and compare to zp->z_parent. 2198fa9e4066Sahrens */ 2199fa9e4066Sahrens do { 2200fa9e4066Sahrens zl = kmem_alloc(sizeof (*zl), KM_SLEEP); 2201fa9e4066Sahrens zl->zl_rwlock = rwlp; 2202fa9e4066Sahrens zl->zl_znode = NULL; 2203fa9e4066Sahrens zl->zl_next = *zlpp; 2204fa9e4066Sahrens *zlpp = zl; 2205fa9e4066Sahrens 2206fa9e4066Sahrens rw_enter(rwlp, rw); 2207fa9e4066Sahrens 2208fa9e4066Sahrens if (*oidp == szp->z_id) /* We're a descendant of szp */ 2209fa9e4066Sahrens return (EINVAL); 2210fa9e4066Sahrens 2211fa9e4066Sahrens if (*oidp == rootid) /* We've hit the top */ 2212fa9e4066Sahrens return (0); 2213fa9e4066Sahrens 2214fa9e4066Sahrens if (rw == RW_READER) { /* i.e. not the first pass */ 2215fa9e4066Sahrens int error = zfs_zget(zp->z_zfsvfs, *oidp, &zp); 2216fa9e4066Sahrens if (error) 2217fa9e4066Sahrens return (error); 2218fa9e4066Sahrens zl->zl_znode = zp; 2219fa9e4066Sahrens } 2220fa9e4066Sahrens oidp = &zp->z_phys->zp_parent; 2221fa9e4066Sahrens rwlp = &zp->z_parent_lock; 2222fa9e4066Sahrens rw = RW_READER; 2223fa9e4066Sahrens 2224fa9e4066Sahrens } while (zp->z_id != sdzp->z_id); 2225fa9e4066Sahrens 2226fa9e4066Sahrens return (0); 2227fa9e4066Sahrens } 2228fa9e4066Sahrens 2229fa9e4066Sahrens /* 2230fa9e4066Sahrens * Drop locks and release vnodes that were held by zfs_rename_lock(). 2231fa9e4066Sahrens */ 2232fa9e4066Sahrens static void 2233fa9e4066Sahrens zfs_rename_unlock(zfs_zlock_t **zlpp) 2234fa9e4066Sahrens { 2235fa9e4066Sahrens zfs_zlock_t *zl; 2236fa9e4066Sahrens 2237fa9e4066Sahrens while ((zl = *zlpp) != NULL) { 2238fa9e4066Sahrens if (zl->zl_znode != NULL) 2239fa9e4066Sahrens VN_RELE(ZTOV(zl->zl_znode)); 2240fa9e4066Sahrens rw_exit(zl->zl_rwlock); 2241fa9e4066Sahrens *zlpp = zl->zl_next; 2242fa9e4066Sahrens kmem_free(zl, sizeof (*zl)); 2243fa9e4066Sahrens } 2244fa9e4066Sahrens } 2245fa9e4066Sahrens 2246fa9e4066Sahrens /* 2247fa9e4066Sahrens * Move an entry from the provided source directory to the target 2248fa9e4066Sahrens * directory. Change the entry name as indicated. 2249fa9e4066Sahrens * 2250fa9e4066Sahrens * IN: sdvp - Source directory containing the "old entry". 2251fa9e4066Sahrens * snm - Old entry name. 2252fa9e4066Sahrens * tdvp - Target directory to contain the "new entry". 2253fa9e4066Sahrens * tnm - New entry name. 2254fa9e4066Sahrens * cr - credentials of caller. 2255fa9e4066Sahrens * 2256fa9e4066Sahrens * RETURN: 0 if success 2257fa9e4066Sahrens * error code if failure 2258fa9e4066Sahrens * 2259fa9e4066Sahrens * Timestamps: 2260fa9e4066Sahrens * sdvp,tdvp - ctime|mtime updated 2261fa9e4066Sahrens */ 2262fa9e4066Sahrens static int 2263fa9e4066Sahrens zfs_rename(vnode_t *sdvp, char *snm, vnode_t *tdvp, char *tnm, cred_t *cr) 2264fa9e4066Sahrens { 2265fa9e4066Sahrens znode_t *tdzp, *szp, *tzp; 2266fa9e4066Sahrens znode_t *sdzp = VTOZ(sdvp); 2267fa9e4066Sahrens zfsvfs_t *zfsvfs = sdzp->z_zfsvfs; 2268fa9e4066Sahrens zilog_t *zilog = zfsvfs->z_log; 2269fa9e4066Sahrens vnode_t *realvp; 2270fa9e4066Sahrens zfs_dirlock_t *sdl, *tdl; 2271fa9e4066Sahrens dmu_tx_t *tx; 2272fa9e4066Sahrens zfs_zlock_t *zl; 2273fa9e4066Sahrens int cmp, serr, terr, error; 2274fa9e4066Sahrens 2275fa9e4066Sahrens ZFS_ENTER(zfsvfs); 2276fa9e4066Sahrens 2277fa9e4066Sahrens /* 2278fa9e4066Sahrens * Make sure we have the real vp for the target directory. 2279fa9e4066Sahrens */ 2280fa9e4066Sahrens if (VOP_REALVP(tdvp, &realvp) == 0) 2281fa9e4066Sahrens tdvp = realvp; 2282fa9e4066Sahrens 2283fa9e4066Sahrens if (tdvp->v_vfsp != sdvp->v_vfsp) { 2284fa9e4066Sahrens ZFS_EXIT(zfsvfs); 2285fa9e4066Sahrens return (EXDEV); 2286fa9e4066Sahrens } 2287fa9e4066Sahrens 2288fa9e4066Sahrens tdzp = VTOZ(tdvp); 2289fa9e4066Sahrens top: 2290fa9e4066Sahrens szp = NULL; 2291fa9e4066Sahrens tzp = NULL; 2292fa9e4066Sahrens zl = NULL; 2293fa9e4066Sahrens 2294fa9e4066Sahrens /* 2295fa9e4066Sahrens * This is to prevent the creation of links into attribute space 2296fa9e4066Sahrens * by renaming a linked file into/outof an attribute directory. 2297fa9e4066Sahrens * See the comment in zfs_link() for why this is considered bad. 2298fa9e4066Sahrens */ 2299fa9e4066Sahrens if ((tdzp->z_phys->zp_flags & ZFS_XATTR) != 2300fa9e4066Sahrens (sdzp->z_phys->zp_flags & ZFS_XATTR)) { 2301fa9e4066Sahrens ZFS_EXIT(zfsvfs); 2302fa9e4066Sahrens return (EINVAL); 2303fa9e4066Sahrens } 2304fa9e4066Sahrens 2305fa9e4066Sahrens /* 2306fa9e4066Sahrens * Lock source and target directory entries. To prevent deadlock, 2307fa9e4066Sahrens * a lock ordering must be defined. We lock the directory with 2308fa9e4066Sahrens * the smallest object id first, or if it's a tie, the one with 2309fa9e4066Sahrens * the lexically first name. 2310fa9e4066Sahrens */ 2311fa9e4066Sahrens if (sdzp->z_id < tdzp->z_id) { 2312fa9e4066Sahrens cmp = -1; 2313fa9e4066Sahrens } else if (sdzp->z_id > tdzp->z_id) { 2314fa9e4066Sahrens cmp = 1; 2315fa9e4066Sahrens } else { 2316fa9e4066Sahrens cmp = strcmp(snm, tnm); 2317fa9e4066Sahrens if (cmp == 0) { 2318fa9e4066Sahrens /* 2319fa9e4066Sahrens * POSIX: "If the old argument and the new argument 2320fa9e4066Sahrens * both refer to links to the same existing file, 2321fa9e4066Sahrens * the rename() function shall return successfully 2322fa9e4066Sahrens * and perform no other action." 2323fa9e4066Sahrens */ 2324fa9e4066Sahrens ZFS_EXIT(zfsvfs); 2325fa9e4066Sahrens return (0); 2326fa9e4066Sahrens } 2327fa9e4066Sahrens } 2328fa9e4066Sahrens if (cmp < 0) { 2329fa9e4066Sahrens serr = zfs_dirent_lock(&sdl, sdzp, snm, &szp, ZEXISTS); 2330fa9e4066Sahrens terr = zfs_dirent_lock(&tdl, tdzp, tnm, &tzp, 0); 2331fa9e4066Sahrens } else { 2332fa9e4066Sahrens terr = zfs_dirent_lock(&tdl, tdzp, tnm, &tzp, 0); 2333fa9e4066Sahrens serr = zfs_dirent_lock(&sdl, sdzp, snm, &szp, ZEXISTS); 2334fa9e4066Sahrens } 2335fa9e4066Sahrens 2336fa9e4066Sahrens if (serr) { 2337fa9e4066Sahrens /* 2338fa9e4066Sahrens * Source entry invalid or not there. 2339fa9e4066Sahrens */ 2340fa9e4066Sahrens if (!terr) { 2341fa9e4066Sahrens zfs_dirent_unlock(tdl); 2342fa9e4066Sahrens if (tzp) 2343fa9e4066Sahrens VN_RELE(ZTOV(tzp)); 2344fa9e4066Sahrens } 2345fa9e4066Sahrens if (strcmp(snm, "..") == 0) 2346fa9e4066Sahrens serr = EINVAL; 2347fa9e4066Sahrens ZFS_EXIT(zfsvfs); 2348fa9e4066Sahrens return (serr); 2349fa9e4066Sahrens } 2350fa9e4066Sahrens if (terr) { 2351fa9e4066Sahrens zfs_dirent_unlock(sdl); 2352fa9e4066Sahrens VN_RELE(ZTOV(szp)); 2353fa9e4066Sahrens if (strcmp(tnm, "..") == 0) 2354fa9e4066Sahrens terr = EINVAL; 2355fa9e4066Sahrens ZFS_EXIT(zfsvfs); 2356fa9e4066Sahrens return (terr); 2357fa9e4066Sahrens } 2358fa9e4066Sahrens 2359fa9e4066Sahrens /* 2360fa9e4066Sahrens * Must have write access at the source to remove the old entry 2361fa9e4066Sahrens * and write access at the target to create the new entry. 2362fa9e4066Sahrens * Note that if target and source are the same, this can be 2363fa9e4066Sahrens * done in a single check. 2364fa9e4066Sahrens */ 2365fa9e4066Sahrens 2366fa9e4066Sahrens if (error = zfs_zaccess_rename(sdzp, szp, tdzp, tzp, cr)) 2367fa9e4066Sahrens goto out; 2368fa9e4066Sahrens 2369fa9e4066Sahrens if (ZTOV(szp)->v_type == VDIR) { 2370fa9e4066Sahrens /* 2371fa9e4066Sahrens * Check to make sure rename is valid. 2372fa9e4066Sahrens * Can't do a move like this: /usr/a/b to /usr/a/b/c/d 2373fa9e4066Sahrens */ 2374fa9e4066Sahrens if (error = zfs_rename_lock(szp, tdzp, sdzp, &zl)) 2375fa9e4066Sahrens goto out; 2376fa9e4066Sahrens } 2377fa9e4066Sahrens 2378fa9e4066Sahrens /* 2379fa9e4066Sahrens * Does target exist? 2380fa9e4066Sahrens */ 2381fa9e4066Sahrens if (tzp) { 2382fa9e4066Sahrens /* 2383fa9e4066Sahrens * Source and target must be the same type. 2384fa9e4066Sahrens */ 2385fa9e4066Sahrens if (ZTOV(szp)->v_type == VDIR) { 2386fa9e4066Sahrens if (ZTOV(tzp)->v_type != VDIR) { 2387fa9e4066Sahrens error = ENOTDIR; 2388fa9e4066Sahrens goto out; 2389fa9e4066Sahrens } 2390fa9e4066Sahrens } else { 2391fa9e4066Sahrens if (ZTOV(tzp)->v_type == VDIR) { 2392fa9e4066Sahrens error = EISDIR; 2393fa9e4066Sahrens goto out; 2394fa9e4066Sahrens } 2395fa9e4066Sahrens } 2396fa9e4066Sahrens /* 2397fa9e4066Sahrens * POSIX dictates that when the source and target 2398fa9e4066Sahrens * entries refer to the same file object, rename 2399fa9e4066Sahrens * must do nothing and exit without error. 2400fa9e4066Sahrens */ 2401fa9e4066Sahrens if (szp->z_id == tzp->z_id) { 2402fa9e4066Sahrens error = 0; 2403fa9e4066Sahrens goto out; 2404fa9e4066Sahrens } 2405fa9e4066Sahrens } 2406fa9e4066Sahrens 2407fa9e4066Sahrens vnevent_rename_src(ZTOV(szp)); 2408fa9e4066Sahrens if (tzp) 2409fa9e4066Sahrens vnevent_rename_dest(ZTOV(tzp)); 2410fa9e4066Sahrens 2411fa9e4066Sahrens tx = dmu_tx_create(zfsvfs->z_os); 2412fa9e4066Sahrens dmu_tx_hold_bonus(tx, szp->z_id); /* nlink changes */ 2413fa9e4066Sahrens dmu_tx_hold_bonus(tx, sdzp->z_id); /* nlink changes */ 2414ea8dc4b6Seschrock dmu_tx_hold_zap(tx, sdzp->z_id, FALSE, snm); 2415ea8dc4b6Seschrock dmu_tx_hold_zap(tx, tdzp->z_id, TRUE, tnm); 2416ea8dc4b6Seschrock if (sdzp != tdzp) 2417fa9e4066Sahrens dmu_tx_hold_bonus(tx, tdzp->z_id); /* nlink changes */ 2418ea8dc4b6Seschrock if (tzp) 2419ea8dc4b6Seschrock dmu_tx_hold_bonus(tx, tzp->z_id); /* parent changes */ 2420ea8dc4b6Seschrock dmu_tx_hold_zap(tx, zfsvfs->z_dqueue, FALSE, NULL); 2421fa9e4066Sahrens error = dmu_tx_assign(tx, zfsvfs->z_assign); 2422fa9e4066Sahrens if (error) { 2423fa9e4066Sahrens if (zl != NULL) 2424fa9e4066Sahrens zfs_rename_unlock(&zl); 2425fa9e4066Sahrens zfs_dirent_unlock(sdl); 2426fa9e4066Sahrens zfs_dirent_unlock(tdl); 2427fa9e4066Sahrens VN_RELE(ZTOV(szp)); 2428fa9e4066Sahrens if (tzp) 2429fa9e4066Sahrens VN_RELE(ZTOV(tzp)); 2430fa9e4066Sahrens if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) { 24318a2f1b91Sahrens dmu_tx_wait(tx); 24328a2f1b91Sahrens dmu_tx_abort(tx); 2433fa9e4066Sahrens goto top; 2434fa9e4066Sahrens } 24358a2f1b91Sahrens dmu_tx_abort(tx); 2436fa9e4066Sahrens ZFS_EXIT(zfsvfs); 2437fa9e4066Sahrens return (error); 2438fa9e4066Sahrens } 2439fa9e4066Sahrens 2440fa9e4066Sahrens if (tzp) /* Attempt to remove the existing target */ 2441fa9e4066Sahrens error = zfs_link_destroy(tdl, tzp, tx, 0, NULL); 2442fa9e4066Sahrens 2443fa9e4066Sahrens if (error == 0) { 2444fa9e4066Sahrens error = zfs_link_create(tdl, szp, tx, ZRENAMING); 2445fa9e4066Sahrens if (error == 0) { 2446fa9e4066Sahrens error = zfs_link_destroy(sdl, szp, tx, ZRENAMING, NULL); 2447fa9e4066Sahrens ASSERT(error == 0); 2448b19a79ecSperrin zfs_log_rename(zilog, tx, TX_RENAME, sdzp, 2449b19a79ecSperrin sdl->dl_name, tdzp, tdl->dl_name, szp); 2450fa9e4066Sahrens } 2451fa9e4066Sahrens } 2452fa9e4066Sahrens 2453fa9e4066Sahrens dmu_tx_commit(tx); 2454fa9e4066Sahrens out: 2455fa9e4066Sahrens if (zl != NULL) 2456fa9e4066Sahrens zfs_rename_unlock(&zl); 2457fa9e4066Sahrens 2458fa9e4066Sahrens zfs_dirent_unlock(sdl); 2459fa9e4066Sahrens zfs_dirent_unlock(tdl); 2460fa9e4066Sahrens 2461fa9e4066Sahrens VN_RELE(ZTOV(szp)); 2462fa9e4066Sahrens if (tzp) 2463fa9e4066Sahrens VN_RELE(ZTOV(tzp)); 2464fa9e4066Sahrens 2465fa9e4066Sahrens ZFS_EXIT(zfsvfs); 2466fa9e4066Sahrens return (error); 2467fa9e4066Sahrens } 2468fa9e4066Sahrens 2469fa9e4066Sahrens /* 2470fa9e4066Sahrens * Insert the indicated symbolic reference entry into the directory. 2471fa9e4066Sahrens * 2472fa9e4066Sahrens * IN: dvp - Directory to contain new symbolic link. 2473fa9e4066Sahrens * link - Name for new symlink entry. 2474fa9e4066Sahrens * vap - Attributes of new entry. 2475fa9e4066Sahrens * target - Target path of new symlink. 2476fa9e4066Sahrens * cr - credentials of caller. 2477fa9e4066Sahrens * 2478fa9e4066Sahrens * RETURN: 0 if success 2479fa9e4066Sahrens * error code if failure 2480fa9e4066Sahrens * 2481fa9e4066Sahrens * Timestamps: 2482fa9e4066Sahrens * dvp - ctime|mtime updated 2483fa9e4066Sahrens */ 2484fa9e4066Sahrens static int 2485fa9e4066Sahrens zfs_symlink(vnode_t *dvp, char *name, vattr_t *vap, char *link, cred_t *cr) 2486fa9e4066Sahrens { 2487fa9e4066Sahrens znode_t *zp, *dzp = VTOZ(dvp); 2488fa9e4066Sahrens zfs_dirlock_t *dl; 2489fa9e4066Sahrens dmu_tx_t *tx; 2490fa9e4066Sahrens zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 2491fa9e4066Sahrens zilog_t *zilog = zfsvfs->z_log; 2492fa9e4066Sahrens uint64_t zoid; 2493fa9e4066Sahrens int len = strlen(link); 2494fa9e4066Sahrens int error; 2495fa9e4066Sahrens 2496fa9e4066Sahrens ASSERT(vap->va_type == VLNK); 2497fa9e4066Sahrens 2498fa9e4066Sahrens ZFS_ENTER(zfsvfs); 2499fa9e4066Sahrens top: 2500fa9e4066Sahrens if (error = zfs_zaccess(dzp, ACE_ADD_FILE, cr)) { 2501fa9e4066Sahrens ZFS_EXIT(zfsvfs); 2502fa9e4066Sahrens return (error); 2503fa9e4066Sahrens } 2504fa9e4066Sahrens 2505fa9e4066Sahrens if (len > MAXPATHLEN) { 2506fa9e4066Sahrens ZFS_EXIT(zfsvfs); 2507fa9e4066Sahrens return (ENAMETOOLONG); 2508fa9e4066Sahrens } 2509fa9e4066Sahrens 2510fa9e4066Sahrens /* 2511fa9e4066Sahrens * Attempt to lock directory; fail if entry already exists. 2512fa9e4066Sahrens */ 2513fa9e4066Sahrens if (error = zfs_dirent_lock(&dl, dzp, name, &zp, ZNEW)) { 2514fa9e4066Sahrens ZFS_EXIT(zfsvfs); 2515fa9e4066Sahrens return (error); 2516fa9e4066Sahrens } 2517fa9e4066Sahrens 2518fa9e4066Sahrens tx = dmu_tx_create(zfsvfs->z_os); 2519fa9e4066Sahrens dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, MAX(1, len)); 2520fa9e4066Sahrens dmu_tx_hold_bonus(tx, dzp->z_id); 2521ea8dc4b6Seschrock dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); 2522fa9e4066Sahrens if (dzp->z_phys->zp_flags & ZFS_INHERIT_ACE) 2523fa9e4066Sahrens dmu_tx_hold_write(tx, DMU_NEW_OBJECT, 0, SPA_MAXBLOCKSIZE); 2524fa9e4066Sahrens error = dmu_tx_assign(tx, zfsvfs->z_assign); 2525fa9e4066Sahrens if (error) { 2526fa9e4066Sahrens zfs_dirent_unlock(dl); 2527fa9e4066Sahrens if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) { 25288a2f1b91Sahrens dmu_tx_wait(tx); 25298a2f1b91Sahrens dmu_tx_abort(tx); 2530fa9e4066Sahrens goto top; 2531fa9e4066Sahrens } 25328a2f1b91Sahrens dmu_tx_abort(tx); 2533fa9e4066Sahrens ZFS_EXIT(zfsvfs); 2534fa9e4066Sahrens return (error); 2535fa9e4066Sahrens } 2536fa9e4066Sahrens 2537fa9e4066Sahrens dmu_buf_will_dirty(dzp->z_dbuf, tx); 2538fa9e4066Sahrens 2539fa9e4066Sahrens /* 2540fa9e4066Sahrens * Create a new object for the symlink. 2541fa9e4066Sahrens * Put the link content into bonus buffer if it will fit; 2542fa9e4066Sahrens * otherwise, store it just like any other file data. 2543fa9e4066Sahrens */ 2544fa9e4066Sahrens zoid = 0; 2545fa9e4066Sahrens if (sizeof (znode_phys_t) + len <= dmu_bonus_max()) { 2546fa9e4066Sahrens zfs_mknode(dzp, vap, &zoid, tx, cr, 0, &zp, len); 2547fa9e4066Sahrens if (len != 0) 2548fa9e4066Sahrens bcopy(link, zp->z_phys + 1, len); 2549fa9e4066Sahrens } else { 2550fa9e4066Sahrens dmu_buf_t *dbp; 2551104e2ed7Sperrin 2552fa9e4066Sahrens zfs_mknode(dzp, vap, &zoid, tx, cr, 0, &zp, 0); 2553fa9e4066Sahrens 2554104e2ed7Sperrin /* 2555104e2ed7Sperrin * Nothing can access the znode yet so no locking needed 2556104e2ed7Sperrin * for growing the znode's blocksize. 2557104e2ed7Sperrin */ 2558104e2ed7Sperrin zfs_grow_blocksize(zp, len, tx); 2559fa9e4066Sahrens 2560ea8dc4b6Seschrock VERIFY(0 == dmu_buf_hold(zfsvfs->z_os, zoid, 0, FTAG, &dbp)); 2561fa9e4066Sahrens dmu_buf_will_dirty(dbp, tx); 2562fa9e4066Sahrens 2563fa9e4066Sahrens ASSERT3U(len, <=, dbp->db_size); 2564fa9e4066Sahrens bcopy(link, dbp->db_data, len); 2565ea8dc4b6Seschrock dmu_buf_rele(dbp, FTAG); 2566fa9e4066Sahrens } 2567fa9e4066Sahrens zp->z_phys->zp_size = len; 2568fa9e4066Sahrens 2569fa9e4066Sahrens /* 2570fa9e4066Sahrens * Insert the new object into the directory. 2571fa9e4066Sahrens */ 2572fa9e4066Sahrens (void) zfs_link_create(dl, zp, tx, ZNEW); 2573fa9e4066Sahrens out: 2574fa9e4066Sahrens if (error == 0) 2575b19a79ecSperrin zfs_log_symlink(zilog, tx, TX_SYMLINK, dzp, zp, name, link); 2576fa9e4066Sahrens 2577fa9e4066Sahrens dmu_tx_commit(tx); 2578fa9e4066Sahrens 2579fa9e4066Sahrens zfs_dirent_unlock(dl); 2580fa9e4066Sahrens 2581fa9e4066Sahrens VN_RELE(ZTOV(zp)); 2582fa9e4066Sahrens 2583fa9e4066Sahrens ZFS_EXIT(zfsvfs); 2584fa9e4066Sahrens return (error); 2585fa9e4066Sahrens } 2586fa9e4066Sahrens 2587fa9e4066Sahrens /* 2588fa9e4066Sahrens * Return, in the buffer contained in the provided uio structure, 2589fa9e4066Sahrens * the symbolic path referred to by vp. 2590fa9e4066Sahrens * 2591fa9e4066Sahrens * IN: vp - vnode of symbolic link. 2592fa9e4066Sahrens * uoip - structure to contain the link path. 2593fa9e4066Sahrens * cr - credentials of caller. 2594fa9e4066Sahrens * 2595fa9e4066Sahrens * OUT: uio - structure to contain the link path. 2596fa9e4066Sahrens * 2597fa9e4066Sahrens * RETURN: 0 if success 2598fa9e4066Sahrens * error code if failure 2599fa9e4066Sahrens * 2600fa9e4066Sahrens * Timestamps: 2601fa9e4066Sahrens * vp - atime updated 2602fa9e4066Sahrens */ 2603fa9e4066Sahrens /* ARGSUSED */ 2604fa9e4066Sahrens static int 2605fa9e4066Sahrens zfs_readlink(vnode_t *vp, uio_t *uio, cred_t *cr) 2606fa9e4066Sahrens { 2607fa9e4066Sahrens znode_t *zp = VTOZ(vp); 2608fa9e4066Sahrens zfsvfs_t *zfsvfs = zp->z_zfsvfs; 2609fa9e4066Sahrens size_t bufsz; 2610fa9e4066Sahrens int error; 2611fa9e4066Sahrens 2612fa9e4066Sahrens ZFS_ENTER(zfsvfs); 2613fa9e4066Sahrens 2614fa9e4066Sahrens bufsz = (size_t)zp->z_phys->zp_size; 2615fa9e4066Sahrens if (bufsz + sizeof (znode_phys_t) <= zp->z_dbuf->db_size) { 2616fa9e4066Sahrens error = uiomove(zp->z_phys + 1, 2617fa9e4066Sahrens MIN((size_t)bufsz, uio->uio_resid), UIO_READ, uio); 2618fa9e4066Sahrens } else { 2619ea8dc4b6Seschrock dmu_buf_t *dbp; 2620ea8dc4b6Seschrock error = dmu_buf_hold(zfsvfs->z_os, zp->z_id, 0, FTAG, &dbp); 2621ea8dc4b6Seschrock if (error) { 2622fa9e4066Sahrens ZFS_EXIT(zfsvfs); 2623fa9e4066Sahrens return (error); 2624fa9e4066Sahrens } 2625fa9e4066Sahrens error = uiomove(dbp->db_data, 2626fa9e4066Sahrens MIN((size_t)bufsz, uio->uio_resid), UIO_READ, uio); 2627ea8dc4b6Seschrock dmu_buf_rele(dbp, FTAG); 2628fa9e4066Sahrens } 2629fa9e4066Sahrens 2630fa9e4066Sahrens ZFS_ACCESSTIME_STAMP(zfsvfs, zp); 2631fa9e4066Sahrens ZFS_EXIT(zfsvfs); 2632fa9e4066Sahrens return (error); 2633fa9e4066Sahrens } 2634fa9e4066Sahrens 2635fa9e4066Sahrens /* 2636fa9e4066Sahrens * Insert a new entry into directory tdvp referencing svp. 2637fa9e4066Sahrens * 2638fa9e4066Sahrens * IN: tdvp - Directory to contain new entry. 2639fa9e4066Sahrens * svp - vnode of new entry. 2640fa9e4066Sahrens * name - name of new entry. 2641fa9e4066Sahrens * cr - credentials of caller. 2642fa9e4066Sahrens * 2643fa9e4066Sahrens * RETURN: 0 if success 2644fa9e4066Sahrens * error code if failure 2645fa9e4066Sahrens * 2646fa9e4066Sahrens * Timestamps: 2647fa9e4066Sahrens * tdvp - ctime|mtime updated 2648fa9e4066Sahrens * svp - ctime updated 2649fa9e4066Sahrens */ 2650fa9e4066Sahrens /* ARGSUSED */ 2651fa9e4066Sahrens static int 2652fa9e4066Sahrens zfs_link(vnode_t *tdvp, vnode_t *svp, char *name, cred_t *cr) 2653fa9e4066Sahrens { 2654fa9e4066Sahrens znode_t *dzp = VTOZ(tdvp); 2655fa9e4066Sahrens znode_t *tzp, *szp; 2656fa9e4066Sahrens zfsvfs_t *zfsvfs = dzp->z_zfsvfs; 2657fa9e4066Sahrens zilog_t *zilog = zfsvfs->z_log; 2658fa9e4066Sahrens zfs_dirlock_t *dl; 2659fa9e4066Sahrens dmu_tx_t *tx; 2660fa9e4066Sahrens vnode_t *realvp; 2661fa9e4066Sahrens int error; 2662fa9e4066Sahrens 2663fa9e4066Sahrens ASSERT(tdvp->v_type == VDIR); 2664fa9e4066Sahrens 2665fa9e4066Sahrens ZFS_ENTER(zfsvfs); 2666fa9e4066Sahrens 2667fa9e4066Sahrens if (VOP_REALVP(svp, &realvp) == 0) 2668fa9e4066Sahrens svp = realvp; 2669fa9e4066Sahrens 2670fa9e4066Sahrens if (svp->v_vfsp != tdvp->v_vfsp) { 2671fa9e4066Sahrens ZFS_EXIT(zfsvfs); 2672fa9e4066Sahrens return (EXDEV); 2673fa9e4066Sahrens } 2674fa9e4066Sahrens 2675fa9e4066Sahrens szp = VTOZ(svp); 2676fa9e4066Sahrens top: 2677fa9e4066Sahrens /* 2678fa9e4066Sahrens * We do not support links between attributes and non-attributes 2679fa9e4066Sahrens * because of the potential security risk of creating links 2680fa9e4066Sahrens * into "normal" file space in order to circumvent restrictions 2681fa9e4066Sahrens * imposed in attribute space. 2682fa9e4066Sahrens */ 2683fa9e4066Sahrens if ((szp->z_phys->zp_flags & ZFS_XATTR) != 2684fa9e4066Sahrens (dzp->z_phys->zp_flags & ZFS_XATTR)) { 2685fa9e4066Sahrens ZFS_EXIT(zfsvfs); 2686fa9e4066Sahrens return (EINVAL); 2687fa9e4066Sahrens } 2688fa9e4066Sahrens 2689fa9e4066Sahrens /* 2690fa9e4066Sahrens * POSIX dictates that we return EPERM here. 2691fa9e4066Sahrens * Better choices include ENOTSUP or EISDIR. 2692fa9e4066Sahrens */ 2693fa9e4066Sahrens if (svp->v_type == VDIR) { 2694fa9e4066Sahrens ZFS_EXIT(zfsvfs); 2695fa9e4066Sahrens return (EPERM); 2696fa9e4066Sahrens } 2697fa9e4066Sahrens 2698fa9e4066Sahrens if ((uid_t)szp->z_phys->zp_uid != crgetuid(cr) && 2699fa9e4066Sahrens secpolicy_basic_link(cr) != 0) { 2700fa9e4066Sahrens ZFS_EXIT(zfsvfs); 2701fa9e4066Sahrens return (EPERM); 2702fa9e4066Sahrens } 2703fa9e4066Sahrens 2704fa9e4066Sahrens if (error = zfs_zaccess(dzp, ACE_ADD_FILE, cr)) { 2705fa9e4066Sahrens ZFS_EXIT(zfsvfs); 2706fa9e4066Sahrens return (error); 2707fa9e4066Sahrens } 2708fa9e4066Sahrens 2709fa9e4066Sahrens /* 2710fa9e4066Sahrens * Attempt to lock directory; fail if entry already exists. 2711fa9e4066Sahrens */ 2712fa9e4066Sahrens if (error = zfs_dirent_lock(&dl, dzp, name, &tzp, ZNEW)) { 2713fa9e4066Sahrens ZFS_EXIT(zfsvfs); 2714fa9e4066Sahrens return (error); 2715fa9e4066Sahrens } 2716fa9e4066Sahrens 2717fa9e4066Sahrens tx = dmu_tx_create(zfsvfs->z_os); 2718fa9e4066Sahrens dmu_tx_hold_bonus(tx, szp->z_id); 2719ea8dc4b6Seschrock dmu_tx_hold_zap(tx, dzp->z_id, TRUE, name); 2720fa9e4066Sahrens error = dmu_tx_assign(tx, zfsvfs->z_assign); 2721fa9e4066Sahrens if (error) { 2722fa9e4066Sahrens zfs_dirent_unlock(dl); 2723fa9e4066Sahrens if (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) { 27248a2f1b91Sahrens dmu_tx_wait(tx); 27258a2f1b91Sahrens dmu_tx_abort(tx); 2726fa9e4066Sahrens goto top; 2727fa9e4066Sahrens } 27288a2f1b91Sahrens dmu_tx_abort(tx); 2729fa9e4066Sahrens ZFS_EXIT(zfsvfs); 2730fa9e4066Sahrens return (error); 2731fa9e4066Sahrens } 2732fa9e4066Sahrens 2733fa9e4066Sahrens error = zfs_link_create(dl, szp, tx, 0); 2734fa9e4066Sahrens 2735fa9e4066Sahrens if (error == 0) 2736b19a79ecSperrin zfs_log_link(zilog, tx, TX_LINK, dzp, szp, name); 2737fa9e4066Sahrens 2738fa9e4066Sahrens dmu_tx_commit(tx); 2739fa9e4066Sahrens 2740fa9e4066Sahrens zfs_dirent_unlock(dl); 2741fa9e4066Sahrens 2742fa9e4066Sahrens ZFS_EXIT(zfsvfs); 2743fa9e4066Sahrens return (error); 2744fa9e4066Sahrens } 2745fa9e4066Sahrens 2746fa9e4066Sahrens /* 2747fa9e4066Sahrens * zfs_null_putapage() is used when the file system has been force 2748fa9e4066Sahrens * unmounted. It just drops the pages. 2749fa9e4066Sahrens */ 2750fa9e4066Sahrens /* ARGSUSED */ 2751fa9e4066Sahrens static int 2752fa9e4066Sahrens zfs_null_putapage(vnode_t *vp, page_t *pp, u_offset_t *offp, 2753fa9e4066Sahrens size_t *lenp, int flags, cred_t *cr) 2754fa9e4066Sahrens { 2755fa9e4066Sahrens pvn_write_done(pp, B_INVAL|B_FORCE|B_ERROR); 2756fa9e4066Sahrens return (0); 2757fa9e4066Sahrens } 2758fa9e4066Sahrens 2759*44eda4d7Smaybee /* 2760*44eda4d7Smaybee * Push a page out to disk, klustering if possible. 2761*44eda4d7Smaybee * 2762*44eda4d7Smaybee * IN: vp - file to push page to. 2763*44eda4d7Smaybee * pp - page to push. 2764*44eda4d7Smaybee * flags - additional flags. 2765*44eda4d7Smaybee * cr - credentials of caller. 2766*44eda4d7Smaybee * 2767*44eda4d7Smaybee * OUT: offp - start of range pushed. 2768*44eda4d7Smaybee * lenp - len of range pushed. 2769*44eda4d7Smaybee * 2770*44eda4d7Smaybee * RETURN: 0 if success 2771*44eda4d7Smaybee * error code if failure 2772*44eda4d7Smaybee * 2773*44eda4d7Smaybee * NOTE: callers must have locked the page to be pushed. On 2774*44eda4d7Smaybee * exit, the page (and all other pages in the kluster) must be 2775*44eda4d7Smaybee * unlocked. 2776*44eda4d7Smaybee */ 2777fa9e4066Sahrens /* ARGSUSED */ 2778fa9e4066Sahrens static int 2779fa9e4066Sahrens zfs_putapage(vnode_t *vp, page_t *pp, u_offset_t *offp, 2780fa9e4066Sahrens size_t *lenp, int flags, cred_t *cr) 2781fa9e4066Sahrens { 2782fa9e4066Sahrens znode_t *zp = VTOZ(vp); 2783fa9e4066Sahrens zfsvfs_t *zfsvfs = zp->z_zfsvfs; 2784fa9e4066Sahrens zilog_t *zilog = zfsvfs->z_log; 2785fa9e4066Sahrens dmu_tx_t *tx; 2786104e2ed7Sperrin rl_t *rl; 2787*44eda4d7Smaybee u_offset_t off, koff; 2788*44eda4d7Smaybee size_t len, klen; 2789fa9e4066Sahrens int err; 2790fa9e4066Sahrens 2791fa9e4066Sahrens off = pp->p_offset; 2792*44eda4d7Smaybee len = PAGESIZE; 2793*44eda4d7Smaybee /* 2794*44eda4d7Smaybee * If our blocksize is bigger than the page size, try to kluster 2795*44eda4d7Smaybee * muiltiple pages so that we write a full block (thus avoiding 2796*44eda4d7Smaybee * a read-modify-write). 2797*44eda4d7Smaybee */ 2798*44eda4d7Smaybee if (zp->z_blksz > PAGESIZE) { 2799*44eda4d7Smaybee uint64_t filesz = zp->z_phys->zp_size; 2800*44eda4d7Smaybee 2801*44eda4d7Smaybee if (!ISP2(zp->z_blksz)) { 2802*44eda4d7Smaybee /* Only one block in the file. */ 2803*44eda4d7Smaybee klen = P2ROUNDUP((ulong_t)zp->z_blksz, PAGESIZE); 2804*44eda4d7Smaybee koff = 0; 2805*44eda4d7Smaybee } else { 2806*44eda4d7Smaybee klen = zp->z_blksz; 2807*44eda4d7Smaybee koff = P2ALIGN(off, (u_offset_t)klen); 2808*44eda4d7Smaybee } 2809*44eda4d7Smaybee ASSERT(koff <= filesz); 2810*44eda4d7Smaybee if (koff + klen > filesz) 2811*44eda4d7Smaybee klen = P2ROUNDUP(filesz - koff, (uint64_t)PAGESIZE); 2812*44eda4d7Smaybee pp = pvn_write_kluster(vp, pp, &off, &len, koff, klen, flags); 2813*44eda4d7Smaybee } 2814*44eda4d7Smaybee ASSERT3U(btop(len), ==, btopr(len)); 2815*44eda4d7Smaybee top: 2816*44eda4d7Smaybee rl = zfs_range_lock(zp, off, len, RL_WRITER); 2817dd6ef538Smaybee /* 2818dd6ef538Smaybee * Can't push pages past end-of-file. 2819dd6ef538Smaybee */ 2820dd6ef538Smaybee if (off >= zp->z_phys->zp_size) { 2821*44eda4d7Smaybee /* discard all pages */ 2822*44eda4d7Smaybee flags |= B_INVAL; 2823*44eda4d7Smaybee err = 0; 2824*44eda4d7Smaybee goto out; 2825*44eda4d7Smaybee } else if (off + len > zp->z_phys->zp_size) { 2826*44eda4d7Smaybee int npages = btopr(zp->z_phys->zp_size - off); 2827*44eda4d7Smaybee page_t *trunc; 2828*44eda4d7Smaybee 2829*44eda4d7Smaybee page_list_break(&pp, &trunc, npages); 2830*44eda4d7Smaybee /* discard pages past end of file */ 2831*44eda4d7Smaybee if (trunc) 2832*44eda4d7Smaybee pvn_write_done(trunc, B_INVAL | flags); 2833*44eda4d7Smaybee len = zp->z_phys->zp_size - off; 2834dd6ef538Smaybee } 2835fa9e4066Sahrens 2836fa9e4066Sahrens tx = dmu_tx_create(zfsvfs->z_os); 2837fa9e4066Sahrens dmu_tx_hold_write(tx, zp->z_id, off, len); 2838fa9e4066Sahrens dmu_tx_hold_bonus(tx, zp->z_id); 2839fa9e4066Sahrens err = dmu_tx_assign(tx, zfsvfs->z_assign); 2840fa9e4066Sahrens if (err != 0) { 2841fa9e4066Sahrens if (err == ERESTART && zfsvfs->z_assign == TXG_NOWAIT) { 2842*44eda4d7Smaybee zfs_range_unlock(rl); 28438a2f1b91Sahrens dmu_tx_wait(tx); 28448a2f1b91Sahrens dmu_tx_abort(tx); 2845*44eda4d7Smaybee err = 0; 2846fa9e4066Sahrens goto top; 2847fa9e4066Sahrens } 28488a2f1b91Sahrens dmu_tx_abort(tx); 2849fa9e4066Sahrens goto out; 2850fa9e4066Sahrens } 2851fa9e4066Sahrens 2852*44eda4d7Smaybee if (zp->z_blksz <= PAGESIZE) { 2853*44eda4d7Smaybee caddr_t va = ppmapin(pp, PROT_READ, (caddr_t)-1); 2854*44eda4d7Smaybee ASSERT3U(len, <=, PAGESIZE); 2855*44eda4d7Smaybee dmu_write(zfsvfs->z_os, zp->z_id, off, len, va, tx); 2856*44eda4d7Smaybee ppmapout(va); 2857*44eda4d7Smaybee } else { 2858*44eda4d7Smaybee err = dmu_write_pages(zfsvfs->z_os, zp->z_id, off, len, pp, tx); 2859*44eda4d7Smaybee } 2860fa9e4066Sahrens 2861*44eda4d7Smaybee if (err == 0) { 2862*44eda4d7Smaybee zfs_time_stamper(zp, CONTENT_MODIFIED, tx); 2863*44eda4d7Smaybee (void) zfs_log_write( 2864*44eda4d7Smaybee zilog, tx, TX_WRITE, zp, off, len, 0, NULL); 2865*44eda4d7Smaybee dmu_tx_commit(tx); 2866*44eda4d7Smaybee } 2867fa9e4066Sahrens 2868*44eda4d7Smaybee out: 2869c5c6ffa0Smaybee zfs_range_unlock(rl); 2870*44eda4d7Smaybee pvn_write_done(pp, (err ? B_ERROR : 0) | B_WRITE | flags); 2871fa9e4066Sahrens if (offp) 2872fa9e4066Sahrens *offp = off; 2873fa9e4066Sahrens if (lenp) 2874fa9e4066Sahrens *lenp = len; 2875fa9e4066Sahrens 2876fa9e4066Sahrens return (err); 2877fa9e4066Sahrens } 2878fa9e4066Sahrens 2879fa9e4066Sahrens /* 2880fa9e4066Sahrens * Copy the portion of the file indicated from pages into the file. 2881fa9e4066Sahrens * The pages are stored in a page list attached to the files vnode. 2882fa9e4066Sahrens * 2883fa9e4066Sahrens * IN: vp - vnode of file to push page data to. 2884fa9e4066Sahrens * off - position in file to put data. 2885fa9e4066Sahrens * len - amount of data to write. 2886fa9e4066Sahrens * flags - flags to control the operation. 2887fa9e4066Sahrens * cr - credentials of caller. 2888fa9e4066Sahrens * 2889fa9e4066Sahrens * RETURN: 0 if success 2890fa9e4066Sahrens * error code if failure 2891fa9e4066Sahrens * 2892fa9e4066Sahrens * Timestamps: 2893fa9e4066Sahrens * vp - ctime|mtime updated 2894fa9e4066Sahrens */ 2895fa9e4066Sahrens static int 2896fa9e4066Sahrens zfs_putpage(vnode_t *vp, offset_t off, size_t len, int flags, cred_t *cr) 2897fa9e4066Sahrens { 2898fa9e4066Sahrens znode_t *zp = VTOZ(vp); 2899fa9e4066Sahrens zfsvfs_t *zfsvfs = zp->z_zfsvfs; 2900fa9e4066Sahrens page_t *pp; 2901fa9e4066Sahrens size_t io_len; 2902fa9e4066Sahrens u_offset_t io_off; 2903104e2ed7Sperrin uint64_t filesz; 2904fa9e4066Sahrens int error = 0; 2905fa9e4066Sahrens 2906fa9e4066Sahrens ZFS_ENTER(zfsvfs); 2907fa9e4066Sahrens 2908fa9e4066Sahrens ASSERT(zp->z_dbuf_held && zp->z_phys); 2909fa9e4066Sahrens 2910fa9e4066Sahrens if (len == 0) { 2911fa9e4066Sahrens /* 2912fa9e4066Sahrens * Search the entire vp list for pages >= off. 2913fa9e4066Sahrens */ 2914fa9e4066Sahrens error = pvn_vplist_dirty(vp, (u_offset_t)off, zfs_putapage, 2915fa9e4066Sahrens flags, cr); 2916fe9cf88cSperrin goto out; 2917fa9e4066Sahrens } 2918fa9e4066Sahrens 2919104e2ed7Sperrin filesz = zp->z_phys->zp_size; /* get consistent copy of zp_size */ 2920104e2ed7Sperrin if (off > filesz) { 2921fa9e4066Sahrens /* past end of file */ 2922fa9e4066Sahrens ZFS_EXIT(zfsvfs); 2923fa9e4066Sahrens return (0); 2924fa9e4066Sahrens } 2925fa9e4066Sahrens 2926104e2ed7Sperrin len = MIN(len, filesz - off); 2927fa9e4066Sahrens 2928fe9cf88cSperrin for (io_off = off; io_off < off + len; io_off += io_len) { 2929fa9e4066Sahrens if ((flags & B_INVAL) || ((flags & B_ASYNC) == 0)) { 2930104e2ed7Sperrin pp = page_lookup(vp, io_off, 2931fa9e4066Sahrens (flags & (B_INVAL | B_FREE)) ? 2932fa9e4066Sahrens SE_EXCL : SE_SHARED); 2933fa9e4066Sahrens } else { 2934fa9e4066Sahrens pp = page_lookup_nowait(vp, io_off, 2935fa9e4066Sahrens (flags & B_FREE) ? SE_EXCL : SE_SHARED); 2936fa9e4066Sahrens } 2937fa9e4066Sahrens 2938fa9e4066Sahrens if (pp != NULL && pvn_getdirty(pp, flags)) { 2939fa9e4066Sahrens int err; 2940fa9e4066Sahrens 2941fa9e4066Sahrens /* 2942fa9e4066Sahrens * Found a dirty page to push 2943fa9e4066Sahrens */ 2944104e2ed7Sperrin err = zfs_putapage(vp, pp, &io_off, &io_len, flags, cr); 2945104e2ed7Sperrin if (err) 2946fa9e4066Sahrens error = err; 2947fa9e4066Sahrens } else { 2948fa9e4066Sahrens io_len = PAGESIZE; 2949fa9e4066Sahrens } 2950fa9e4066Sahrens } 2951fe9cf88cSperrin out: 2952b19a79ecSperrin if ((flags & B_ASYNC) == 0) 2953b19a79ecSperrin zil_commit(zfsvfs->z_log, UINT64_MAX, zp->z_id); 2954fa9e4066Sahrens ZFS_EXIT(zfsvfs); 2955fa9e4066Sahrens return (error); 2956fa9e4066Sahrens } 2957fa9e4066Sahrens 2958fa9e4066Sahrens void 2959fa9e4066Sahrens zfs_inactive(vnode_t *vp, cred_t *cr) 2960fa9e4066Sahrens { 2961fa9e4066Sahrens znode_t *zp = VTOZ(vp); 2962fa9e4066Sahrens zfsvfs_t *zfsvfs = zp->z_zfsvfs; 2963fa9e4066Sahrens int error; 2964fa9e4066Sahrens 2965fa9e4066Sahrens rw_enter(&zfsvfs->z_um_lock, RW_READER); 2966fa9e4066Sahrens if (zfsvfs->z_unmounted2) { 2967fa9e4066Sahrens ASSERT(zp->z_dbuf_held == 0); 2968fa9e4066Sahrens 2969fa9e4066Sahrens if (vn_has_cached_data(vp)) { 2970fa9e4066Sahrens (void) pvn_vplist_dirty(vp, 0, zfs_null_putapage, 2971fa9e4066Sahrens B_INVAL, cr); 2972fa9e4066Sahrens } 2973fa9e4066Sahrens 2974ea8dc4b6Seschrock mutex_enter(&zp->z_lock); 2975fa9e4066Sahrens vp->v_count = 0; /* count arrives as 1 */ 2976ea8dc4b6Seschrock if (zp->z_dbuf == NULL) { 2977ea8dc4b6Seschrock mutex_exit(&zp->z_lock); 2978ea8dc4b6Seschrock zfs_znode_free(zp); 2979ea8dc4b6Seschrock } else { 2980ea8dc4b6Seschrock mutex_exit(&zp->z_lock); 2981ea8dc4b6Seschrock } 2982fa9e4066Sahrens rw_exit(&zfsvfs->z_um_lock); 2983fa9e4066Sahrens VFS_RELE(zfsvfs->z_vfs); 2984fa9e4066Sahrens return; 2985fa9e4066Sahrens } 2986fa9e4066Sahrens 2987fa9e4066Sahrens /* 2988fa9e4066Sahrens * Attempt to push any data in the page cache. If this fails 2989fa9e4066Sahrens * we will get kicked out later in zfs_zinactive(). 2990fa9e4066Sahrens */ 29918afd4dd6Sperrin if (vn_has_cached_data(vp)) { 29928afd4dd6Sperrin (void) pvn_vplist_dirty(vp, 0, zfs_putapage, B_INVAL|B_ASYNC, 29938afd4dd6Sperrin cr); 29948afd4dd6Sperrin } 2995fa9e4066Sahrens 2996fa9e4066Sahrens if (zp->z_atime_dirty && zp->z_reap == 0) { 2997fa9e4066Sahrens dmu_tx_t *tx = dmu_tx_create(zfsvfs->z_os); 2998fa9e4066Sahrens 2999fa9e4066Sahrens dmu_tx_hold_bonus(tx, zp->z_id); 3000fa9e4066Sahrens error = dmu_tx_assign(tx, TXG_WAIT); 3001fa9e4066Sahrens if (error) { 3002fa9e4066Sahrens dmu_tx_abort(tx); 3003fa9e4066Sahrens } else { 3004fa9e4066Sahrens dmu_buf_will_dirty(zp->z_dbuf, tx); 3005fa9e4066Sahrens mutex_enter(&zp->z_lock); 3006fa9e4066Sahrens zp->z_atime_dirty = 0; 3007fa9e4066Sahrens mutex_exit(&zp->z_lock); 3008fa9e4066Sahrens dmu_tx_commit(tx); 3009fa9e4066Sahrens } 3010fa9e4066Sahrens } 3011fa9e4066Sahrens 3012fa9e4066Sahrens zfs_zinactive(zp); 3013fa9e4066Sahrens rw_exit(&zfsvfs->z_um_lock); 3014fa9e4066Sahrens } 3015fa9e4066Sahrens 3016fa9e4066Sahrens /* 3017fa9e4066Sahrens * Bounds-check the seek operation. 3018fa9e4066Sahrens * 3019fa9e4066Sahrens * IN: vp - vnode seeking within 3020fa9e4066Sahrens * ooff - old file offset 3021fa9e4066Sahrens * noffp - pointer to new file offset 3022fa9e4066Sahrens * 3023fa9e4066Sahrens * RETURN: 0 if success 3024fa9e4066Sahrens * EINVAL if new offset invalid 3025fa9e4066Sahrens */ 3026fa9e4066Sahrens /* ARGSUSED */ 3027fa9e4066Sahrens static int 3028fa9e4066Sahrens zfs_seek(vnode_t *vp, offset_t ooff, offset_t *noffp) 3029fa9e4066Sahrens { 3030fa9e4066Sahrens if (vp->v_type == VDIR) 3031fa9e4066Sahrens return (0); 3032fa9e4066Sahrens return ((*noffp < 0 || *noffp > MAXOFFSET_T) ? EINVAL : 0); 3033fa9e4066Sahrens } 3034fa9e4066Sahrens 3035fa9e4066Sahrens /* 3036fa9e4066Sahrens * Pre-filter the generic locking function to trap attempts to place 3037fa9e4066Sahrens * a mandatory lock on a memory mapped file. 3038fa9e4066Sahrens */ 3039fa9e4066Sahrens static int 3040fa9e4066Sahrens zfs_frlock(vnode_t *vp, int cmd, flock64_t *bfp, int flag, offset_t offset, 3041fa9e4066Sahrens flk_callback_t *flk_cbp, cred_t *cr) 3042fa9e4066Sahrens { 3043fa9e4066Sahrens znode_t *zp = VTOZ(vp); 3044fa9e4066Sahrens zfsvfs_t *zfsvfs = zp->z_zfsvfs; 3045fa9e4066Sahrens int error; 3046fa9e4066Sahrens 3047fa9e4066Sahrens ZFS_ENTER(zfsvfs); 3048fa9e4066Sahrens 3049fa9e4066Sahrens /* 3050ea8dc4b6Seschrock * We are following the UFS semantics with respect to mapcnt 3051ea8dc4b6Seschrock * here: If we see that the file is mapped already, then we will 3052ea8dc4b6Seschrock * return an error, but we don't worry about races between this 3053ea8dc4b6Seschrock * function and zfs_map(). 3054fa9e4066Sahrens */ 3055ea8dc4b6Seschrock if (zp->z_mapcnt > 0 && MANDMODE((mode_t)zp->z_phys->zp_mode)) { 3056fa9e4066Sahrens ZFS_EXIT(zfsvfs); 3057fa9e4066Sahrens return (EAGAIN); 3058fa9e4066Sahrens } 3059fa9e4066Sahrens error = fs_frlock(vp, cmd, bfp, flag, offset, flk_cbp, cr); 3060fa9e4066Sahrens ZFS_EXIT(zfsvfs); 3061fa9e4066Sahrens return (error); 3062fa9e4066Sahrens } 3063fa9e4066Sahrens 3064fa9e4066Sahrens /* 3065fa9e4066Sahrens * If we can't find a page in the cache, we will create a new page 3066fa9e4066Sahrens * and fill it with file data. For efficiency, we may try to fill 3067104e2ed7Sperrin * multiple pages at once (klustering). 3068fa9e4066Sahrens */ 3069fa9e4066Sahrens static int 3070fa9e4066Sahrens zfs_fillpage(vnode_t *vp, u_offset_t off, struct seg *seg, 3071fa9e4066Sahrens caddr_t addr, page_t *pl[], size_t plsz, enum seg_rw rw) 3072fa9e4066Sahrens { 3073fa9e4066Sahrens znode_t *zp = VTOZ(vp); 3074fa9e4066Sahrens page_t *pp, *cur_pp; 3075fa9e4066Sahrens objset_t *os = zp->z_zfsvfs->z_os; 3076fa9e4066Sahrens caddr_t va; 3077fa9e4066Sahrens u_offset_t io_off, total; 3078fa9e4066Sahrens uint64_t oid = zp->z_id; 3079fa9e4066Sahrens size_t io_len; 3080104e2ed7Sperrin uint64_t filesz; 3081fa9e4066Sahrens int err; 3082fa9e4066Sahrens 3083fa9e4066Sahrens /* 3084fa9e4066Sahrens * If we are only asking for a single page don't bother klustering. 3085fa9e4066Sahrens */ 3086104e2ed7Sperrin filesz = zp->z_phys->zp_size; /* get consistent copy of zp_size */ 3087*44eda4d7Smaybee if (off >= filesz) 3088*44eda4d7Smaybee return (EFAULT); 3089*44eda4d7Smaybee if (plsz == PAGESIZE || zp->z_blksz <= PAGESIZE) { 3090fa9e4066Sahrens io_off = off; 3091fa9e4066Sahrens io_len = PAGESIZE; 3092fa9e4066Sahrens pp = page_create_va(vp, io_off, io_len, PG_WAIT, seg, addr); 3093fa9e4066Sahrens } else { 3094fa9e4066Sahrens /* 3095fa9e4066Sahrens * Try to fill a kluster of pages (a blocks worth). 3096fa9e4066Sahrens */ 3097fa9e4066Sahrens size_t klen; 3098fa9e4066Sahrens u_offset_t koff; 3099fa9e4066Sahrens 3100fa9e4066Sahrens if (!ISP2(zp->z_blksz)) { 3101fa9e4066Sahrens /* Only one block in the file. */ 3102fa9e4066Sahrens klen = P2ROUNDUP((ulong_t)zp->z_blksz, PAGESIZE); 3103fa9e4066Sahrens koff = 0; 3104fa9e4066Sahrens } else { 3105fa9e4066Sahrens klen = plsz; 3106fa9e4066Sahrens koff = P2ALIGN(off, (u_offset_t)klen); 3107fa9e4066Sahrens } 3108dd6ef538Smaybee ASSERT(koff <= filesz); 3109dd6ef538Smaybee if (koff + klen > filesz) 3110dd6ef538Smaybee klen = P2ROUNDUP(filesz, (uint64_t)PAGESIZE) - koff; 3111*44eda4d7Smaybee ASSERT3U(off, >=, koff); 3112*44eda4d7Smaybee ASSERT3U(off, <, koff + klen); 3113fa9e4066Sahrens pp = pvn_read_kluster(vp, off, seg, addr, &io_off, 3114fa9e4066Sahrens &io_len, koff, klen, 0); 3115fa9e4066Sahrens } 3116fa9e4066Sahrens if (pp == NULL) { 3117fa9e4066Sahrens /* 3118fa9e4066Sahrens * Some other thread entered the page before us. 3119fa9e4066Sahrens * Return to zfs_getpage to retry the lookup. 3120fa9e4066Sahrens */ 3121fa9e4066Sahrens *pl = NULL; 3122fa9e4066Sahrens return (0); 3123fa9e4066Sahrens } 3124fa9e4066Sahrens 3125fa9e4066Sahrens /* 3126fa9e4066Sahrens * Fill the pages in the kluster. 3127fa9e4066Sahrens */ 3128fa9e4066Sahrens cur_pp = pp; 3129fa9e4066Sahrens for (total = io_off + io_len; io_off < total; io_off += PAGESIZE) { 3130*44eda4d7Smaybee ASSERT3U(io_off, ==, cur_pp->p_offset); 3131fa9e4066Sahrens va = ppmapin(cur_pp, PROT_READ | PROT_WRITE, (caddr_t)-1); 3132ea8dc4b6Seschrock err = dmu_read(os, oid, io_off, PAGESIZE, va); 3133fa9e4066Sahrens ppmapout(va); 3134fa9e4066Sahrens if (err) { 3135fa9e4066Sahrens /* On error, toss the entire kluster */ 3136fa9e4066Sahrens pvn_read_done(pp, B_ERROR); 3137fa9e4066Sahrens return (err); 3138fa9e4066Sahrens } 3139fa9e4066Sahrens cur_pp = cur_pp->p_next; 3140fa9e4066Sahrens } 3141fa9e4066Sahrens out: 3142fa9e4066Sahrens /* 3143fa9e4066Sahrens * Fill in the page list array from the kluster. If 3144fa9e4066Sahrens * there are too many pages in the kluster, return 3145fa9e4066Sahrens * as many pages as possible starting from the desired 3146fa9e4066Sahrens * offset `off'. 3147fa9e4066Sahrens * NOTE: the page list will always be null terminated. 3148fa9e4066Sahrens */ 3149fa9e4066Sahrens pvn_plist_init(pp, pl, plsz, off, io_len, rw); 3150fa9e4066Sahrens 3151fa9e4066Sahrens return (0); 3152fa9e4066Sahrens } 3153fa9e4066Sahrens 3154fa9e4066Sahrens /* 3155fa9e4066Sahrens * Return pointers to the pages for the file region [off, off + len] 3156fa9e4066Sahrens * in the pl array. If plsz is greater than len, this function may 3157fa9e4066Sahrens * also return page pointers from before or after the specified 3158fa9e4066Sahrens * region (i.e. some region [off', off' + plsz]). These additional 3159fa9e4066Sahrens * pages are only returned if they are already in the cache, or were 3160fa9e4066Sahrens * created as part of a klustered read. 3161fa9e4066Sahrens * 3162fa9e4066Sahrens * IN: vp - vnode of file to get data from. 3163fa9e4066Sahrens * off - position in file to get data from. 3164fa9e4066Sahrens * len - amount of data to retrieve. 3165fa9e4066Sahrens * plsz - length of provided page list. 3166fa9e4066Sahrens * seg - segment to obtain pages for. 3167fa9e4066Sahrens * addr - virtual address of fault. 3168fa9e4066Sahrens * rw - mode of created pages. 3169fa9e4066Sahrens * cr - credentials of caller. 3170fa9e4066Sahrens * 3171fa9e4066Sahrens * OUT: protp - protection mode of created pages. 3172fa9e4066Sahrens * pl - list of pages created. 3173fa9e4066Sahrens * 3174fa9e4066Sahrens * RETURN: 0 if success 3175fa9e4066Sahrens * error code if failure 3176fa9e4066Sahrens * 3177fa9e4066Sahrens * Timestamps: 3178fa9e4066Sahrens * vp - atime updated 3179fa9e4066Sahrens */ 3180fa9e4066Sahrens /* ARGSUSED */ 3181fa9e4066Sahrens static int 3182fa9e4066Sahrens zfs_getpage(vnode_t *vp, offset_t off, size_t len, uint_t *protp, 3183fa9e4066Sahrens page_t *pl[], size_t plsz, struct seg *seg, caddr_t addr, 3184fa9e4066Sahrens enum seg_rw rw, cred_t *cr) 3185fa9e4066Sahrens { 3186fa9e4066Sahrens znode_t *zp = VTOZ(vp); 3187fa9e4066Sahrens zfsvfs_t *zfsvfs = zp->z_zfsvfs; 3188fa9e4066Sahrens page_t *pp, **pl0 = pl; 3189104e2ed7Sperrin rl_t *rl; 3190fa9e4066Sahrens int cnt = 0, need_unlock = 0, err = 0; 3191fa9e4066Sahrens 3192fa9e4066Sahrens ZFS_ENTER(zfsvfs); 3193fa9e4066Sahrens 3194fa9e4066Sahrens if (protp) 3195fa9e4066Sahrens *protp = PROT_ALL; 3196fa9e4066Sahrens 3197fa9e4066Sahrens ASSERT(zp->z_dbuf_held && zp->z_phys); 3198fa9e4066Sahrens 3199fa9e4066Sahrens /* no faultahead (for now) */ 3200fa9e4066Sahrens if (pl == NULL) { 3201fa9e4066Sahrens ZFS_EXIT(zfsvfs); 3202fa9e4066Sahrens return (0); 3203fa9e4066Sahrens } 3204fa9e4066Sahrens 3205104e2ed7Sperrin /* 3206104e2ed7Sperrin * Make sure nobody restructures the file in the middle of the getpage. 3207104e2ed7Sperrin */ 3208104e2ed7Sperrin rl = zfs_range_lock(zp, off, len, RL_READER); 3209104e2ed7Sperrin 3210fa9e4066Sahrens /* can't fault past EOF */ 3211fa9e4066Sahrens if (off >= zp->z_phys->zp_size) { 3212c5c6ffa0Smaybee zfs_range_unlock(rl); 3213fa9e4066Sahrens ZFS_EXIT(zfsvfs); 3214fa9e4066Sahrens return (EFAULT); 3215fa9e4066Sahrens } 3216fa9e4066Sahrens 3217fa9e4066Sahrens /* 3218fa9e4066Sahrens * If we already own the lock, then we must be page faulting 3219fa9e4066Sahrens * in the middle of a write to this file (i.e., we are writing 3220fa9e4066Sahrens * to this file using data from a mapped region of the file). 3221fa9e4066Sahrens */ 3222fa9e4066Sahrens if (!rw_owner(&zp->z_map_lock)) { 3223fa9e4066Sahrens rw_enter(&zp->z_map_lock, RW_WRITER); 3224fa9e4066Sahrens need_unlock = TRUE; 3225fa9e4066Sahrens } 3226fa9e4066Sahrens 3227fa9e4066Sahrens /* 3228fa9e4066Sahrens * Loop through the requested range [off, off + len] looking 3229fa9e4066Sahrens * for pages. If we don't find a page, we will need to create 3230fa9e4066Sahrens * a new page and fill it with data from the file. 3231fa9e4066Sahrens */ 3232fa9e4066Sahrens while (len > 0) { 3233fa9e4066Sahrens if (plsz < PAGESIZE) 3234fa9e4066Sahrens break; 3235fa9e4066Sahrens if (pp = page_lookup(vp, off, SE_SHARED)) { 3236fa9e4066Sahrens *pl++ = pp; 3237fa9e4066Sahrens off += PAGESIZE; 3238fa9e4066Sahrens addr += PAGESIZE; 3239fa9e4066Sahrens len -= PAGESIZE; 3240fa9e4066Sahrens plsz -= PAGESIZE; 3241fa9e4066Sahrens } else { 3242fa9e4066Sahrens err = zfs_fillpage(vp, off, seg, addr, pl, plsz, rw); 3243fa9e4066Sahrens /* 3244fa9e4066Sahrens * klustering may have changed our region 3245fa9e4066Sahrens * to be block aligned. 3246fa9e4066Sahrens */ 3247fa9e4066Sahrens if (((pp = *pl) != 0) && (off != pp->p_offset)) { 3248fa9e4066Sahrens int delta = off - pp->p_offset; 3249fa9e4066Sahrens len += delta; 3250fa9e4066Sahrens off -= delta; 3251fa9e4066Sahrens addr -= delta; 3252fa9e4066Sahrens } 3253fa9e4066Sahrens while (*pl) { 3254fa9e4066Sahrens pl++; 3255fa9e4066Sahrens cnt++; 3256fa9e4066Sahrens off += PAGESIZE; 3257fa9e4066Sahrens addr += PAGESIZE; 3258fa9e4066Sahrens plsz -= PAGESIZE; 3259fa9e4066Sahrens if (len > PAGESIZE) 3260fa9e4066Sahrens len -= PAGESIZE; 3261fa9e4066Sahrens else 3262fa9e4066Sahrens len = 0; 3263fa9e4066Sahrens } 3264104e2ed7Sperrin if (err) { 3265104e2ed7Sperrin /* 3266104e2ed7Sperrin * Release any pages we have locked. 3267104e2ed7Sperrin */ 3268104e2ed7Sperrin while (pl > pl0) 3269104e2ed7Sperrin page_unlock(*--pl); 3270104e2ed7Sperrin goto out; 3271104e2ed7Sperrin } 3272fa9e4066Sahrens } 3273fa9e4066Sahrens } 3274fa9e4066Sahrens 3275fa9e4066Sahrens /* 3276fa9e4066Sahrens * Fill out the page array with any pages already in the cache. 3277fa9e4066Sahrens */ 3278fa9e4066Sahrens while (plsz > 0) { 3279fa9e4066Sahrens pp = page_lookup_nowait(vp, off, SE_SHARED); 3280fa9e4066Sahrens if (pp == NULL) 3281fa9e4066Sahrens break; 3282fa9e4066Sahrens *pl++ = pp; 3283fa9e4066Sahrens off += PAGESIZE; 3284fa9e4066Sahrens plsz -= PAGESIZE; 3285fa9e4066Sahrens } 3286fa9e4066Sahrens 3287fa9e4066Sahrens ZFS_ACCESSTIME_STAMP(zfsvfs, zp); 3288fa9e4066Sahrens out: 3289fa9e4066Sahrens *pl = NULL; 3290fa9e4066Sahrens 3291fa9e4066Sahrens if (need_unlock) 3292fa9e4066Sahrens rw_exit(&zp->z_map_lock); 3293c5c6ffa0Smaybee zfs_range_unlock(rl); 3294fa9e4066Sahrens 3295fa9e4066Sahrens ZFS_EXIT(zfsvfs); 3296fa9e4066Sahrens return (err); 3297fa9e4066Sahrens } 3298fa9e4066Sahrens 3299ea8dc4b6Seschrock /* 3300ea8dc4b6Seschrock * Request a memory map for a section of a file. This code interacts 3301ea8dc4b6Seschrock * with common code and the VM system as follows: 3302ea8dc4b6Seschrock * 3303ea8dc4b6Seschrock * common code calls mmap(), which ends up in smmap_common() 3304ea8dc4b6Seschrock * 3305ea8dc4b6Seschrock * this calls VOP_MAP(), which takes you into (say) zfs 3306ea8dc4b6Seschrock * 3307ea8dc4b6Seschrock * zfs_map() calls as_map(), passing segvn_create() as the callback 3308ea8dc4b6Seschrock * 3309ea8dc4b6Seschrock * segvn_create() creates the new segment and calls VOP_ADDMAP() 3310ea8dc4b6Seschrock * 3311ea8dc4b6Seschrock * zfs_addmap() updates z_mapcnt 3312ea8dc4b6Seschrock */ 3313fa9e4066Sahrens static int 3314fa9e4066Sahrens zfs_map(vnode_t *vp, offset_t off, struct as *as, caddr_t *addrp, 3315fa9e4066Sahrens size_t len, uchar_t prot, uchar_t maxprot, uint_t flags, cred_t *cr) 3316fa9e4066Sahrens { 3317fa9e4066Sahrens znode_t *zp = VTOZ(vp); 3318fa9e4066Sahrens zfsvfs_t *zfsvfs = zp->z_zfsvfs; 3319fa9e4066Sahrens segvn_crargs_t vn_a; 3320fa9e4066Sahrens int error; 3321fa9e4066Sahrens 3322fa9e4066Sahrens ZFS_ENTER(zfsvfs); 3323fa9e4066Sahrens 3324fa9e4066Sahrens if (vp->v_flag & VNOMAP) { 3325fa9e4066Sahrens ZFS_EXIT(zfsvfs); 3326fa9e4066Sahrens return (ENOSYS); 3327fa9e4066Sahrens } 3328fa9e4066Sahrens 3329fa9e4066Sahrens if (off < 0 || len > MAXOFFSET_T - off) { 3330fa9e4066Sahrens ZFS_EXIT(zfsvfs); 3331fa9e4066Sahrens return (ENXIO); 3332fa9e4066Sahrens } 3333fa9e4066Sahrens 3334fa9e4066Sahrens if (vp->v_type != VREG) { 3335fa9e4066Sahrens ZFS_EXIT(zfsvfs); 3336fa9e4066Sahrens return (ENODEV); 3337fa9e4066Sahrens } 3338fa9e4066Sahrens 3339fa9e4066Sahrens /* 3340fa9e4066Sahrens * If file is locked, disallow mapping. 3341fa9e4066Sahrens */ 3342ea8dc4b6Seschrock if (MANDMODE((mode_t)zp->z_phys->zp_mode) && vn_has_flocks(vp)) { 3343ea8dc4b6Seschrock ZFS_EXIT(zfsvfs); 3344ea8dc4b6Seschrock return (EAGAIN); 3345fa9e4066Sahrens } 3346fa9e4066Sahrens 3347fa9e4066Sahrens as_rangelock(as); 3348fa9e4066Sahrens if ((flags & MAP_FIXED) == 0) { 3349fa9e4066Sahrens map_addr(addrp, len, off, 1, flags); 3350fa9e4066Sahrens if (*addrp == NULL) { 3351fa9e4066Sahrens as_rangeunlock(as); 3352fa9e4066Sahrens ZFS_EXIT(zfsvfs); 3353fa9e4066Sahrens return (ENOMEM); 3354fa9e4066Sahrens } 3355fa9e4066Sahrens } else { 3356fa9e4066Sahrens /* 3357fa9e4066Sahrens * User specified address - blow away any previous mappings 3358fa9e4066Sahrens */ 3359fa9e4066Sahrens (void) as_unmap(as, *addrp, len); 3360fa9e4066Sahrens } 3361fa9e4066Sahrens 3362fa9e4066Sahrens vn_a.vp = vp; 3363fa9e4066Sahrens vn_a.offset = (u_offset_t)off; 3364fa9e4066Sahrens vn_a.type = flags & MAP_TYPE; 3365fa9e4066Sahrens vn_a.prot = prot; 3366fa9e4066Sahrens vn_a.maxprot = maxprot; 3367fa9e4066Sahrens vn_a.cred = cr; 3368fa9e4066Sahrens vn_a.amp = NULL; 3369fa9e4066Sahrens vn_a.flags = flags & ~MAP_TYPE; 33704944b02eSkchow vn_a.szc = 0; 33714944b02eSkchow vn_a.lgrp_mem_policy_flags = 0; 3372fa9e4066Sahrens 3373fa9e4066Sahrens error = as_map(as, *addrp, len, segvn_create, &vn_a); 3374fa9e4066Sahrens 3375fa9e4066Sahrens as_rangeunlock(as); 3376fa9e4066Sahrens ZFS_EXIT(zfsvfs); 3377fa9e4066Sahrens return (error); 3378fa9e4066Sahrens } 3379fa9e4066Sahrens 3380fa9e4066Sahrens /* ARGSUSED */ 3381fa9e4066Sahrens static int 3382fa9e4066Sahrens zfs_addmap(vnode_t *vp, offset_t off, struct as *as, caddr_t addr, 3383fa9e4066Sahrens size_t len, uchar_t prot, uchar_t maxprot, uint_t flags, cred_t *cr) 3384fa9e4066Sahrens { 3385ea8dc4b6Seschrock uint64_t pages = btopr(len); 3386ea8dc4b6Seschrock 3387ea8dc4b6Seschrock atomic_add_64(&VTOZ(vp)->z_mapcnt, pages); 3388fa9e4066Sahrens return (0); 3389fa9e4066Sahrens } 3390fa9e4066Sahrens 3391b468a217Seschrock /* 3392b468a217Seschrock * The reason we push dirty pages as part of zfs_delmap() is so that we get a 3393b468a217Seschrock * more accurate mtime for the associated file. Since we don't have a way of 3394b468a217Seschrock * detecting when the data was actually modified, we have to resort to 3395b468a217Seschrock * heuristics. If an explicit msync() is done, then we mark the mtime when the 3396b468a217Seschrock * last page is pushed. The problem occurs when the msync() call is omitted, 3397b468a217Seschrock * which by far the most common case: 3398b468a217Seschrock * 3399b468a217Seschrock * open() 3400b468a217Seschrock * mmap() 3401b468a217Seschrock * <modify memory> 3402b468a217Seschrock * munmap() 3403b468a217Seschrock * close() 3404b468a217Seschrock * <time lapse> 3405b468a217Seschrock * putpage() via fsflush 3406b468a217Seschrock * 3407b468a217Seschrock * If we wait until fsflush to come along, we can have a modification time that 3408b468a217Seschrock * is some arbitrary point in the future. In order to prevent this in the 3409b468a217Seschrock * common case, we flush pages whenever a (MAP_SHARED, PROT_WRITE) mapping is 3410b468a217Seschrock * torn down. 3411b468a217Seschrock */ 3412fa9e4066Sahrens /* ARGSUSED */ 3413fa9e4066Sahrens static int 3414fa9e4066Sahrens zfs_delmap(vnode_t *vp, offset_t off, struct as *as, caddr_t addr, 3415fa9e4066Sahrens size_t len, uint_t prot, uint_t maxprot, uint_t flags, cred_t *cr) 3416fa9e4066Sahrens { 3417ea8dc4b6Seschrock uint64_t pages = btopr(len); 3418ea8dc4b6Seschrock 3419ea8dc4b6Seschrock ASSERT3U(VTOZ(vp)->z_mapcnt, >=, pages); 3420ea8dc4b6Seschrock atomic_add_64(&VTOZ(vp)->z_mapcnt, -pages); 3421b468a217Seschrock 3422b468a217Seschrock if ((flags & MAP_SHARED) && (prot & PROT_WRITE) && 3423b468a217Seschrock vn_has_cached_data(vp)) 3424b468a217Seschrock (void) VOP_PUTPAGE(vp, off, len, B_ASYNC, cr); 3425b468a217Seschrock 3426fa9e4066Sahrens return (0); 3427fa9e4066Sahrens } 3428fa9e4066Sahrens 3429fa9e4066Sahrens /* 3430fa9e4066Sahrens * Free or allocate space in a file. Currently, this function only 3431fa9e4066Sahrens * supports the `F_FREESP' command. However, this command is somewhat 3432fa9e4066Sahrens * misnamed, as its functionality includes the ability to allocate as 3433fa9e4066Sahrens * well as free space. 3434fa9e4066Sahrens * 3435fa9e4066Sahrens * IN: vp - vnode of file to free data in. 3436fa9e4066Sahrens * cmd - action to take (only F_FREESP supported). 3437fa9e4066Sahrens * bfp - section of file to free/alloc. 3438fa9e4066Sahrens * flag - current file open mode flags. 3439fa9e4066Sahrens * offset - current file offset. 3440fa9e4066Sahrens * cr - credentials of caller [UNUSED]. 3441fa9e4066Sahrens * 3442fa9e4066Sahrens * RETURN: 0 if success 3443fa9e4066Sahrens * error code if failure 3444fa9e4066Sahrens * 3445fa9e4066Sahrens * Timestamps: 3446fa9e4066Sahrens * vp - ctime|mtime updated 3447fa9e4066Sahrens */ 3448fa9e4066Sahrens /* ARGSUSED */ 3449fa9e4066Sahrens static int 3450fa9e4066Sahrens zfs_space(vnode_t *vp, int cmd, flock64_t *bfp, int flag, 3451fa9e4066Sahrens offset_t offset, cred_t *cr, caller_context_t *ct) 3452fa9e4066Sahrens { 3453fa9e4066Sahrens znode_t *zp = VTOZ(vp); 3454fa9e4066Sahrens zfsvfs_t *zfsvfs = zp->z_zfsvfs; 3455fa9e4066Sahrens uint64_t off, len; 3456fa9e4066Sahrens int error; 3457fa9e4066Sahrens 3458fa9e4066Sahrens ZFS_ENTER(zfsvfs); 3459fa9e4066Sahrens 3460fa9e4066Sahrens top: 3461fa9e4066Sahrens if (cmd != F_FREESP) { 3462fa9e4066Sahrens ZFS_EXIT(zfsvfs); 3463fa9e4066Sahrens return (EINVAL); 3464fa9e4066Sahrens } 3465fa9e4066Sahrens 3466fa9e4066Sahrens if (error = convoff(vp, bfp, 0, offset)) { 3467fa9e4066Sahrens ZFS_EXIT(zfsvfs); 3468fa9e4066Sahrens return (error); 3469fa9e4066Sahrens } 3470fa9e4066Sahrens 3471fa9e4066Sahrens if (bfp->l_len < 0) { 3472fa9e4066Sahrens ZFS_EXIT(zfsvfs); 3473fa9e4066Sahrens return (EINVAL); 3474fa9e4066Sahrens } 3475fa9e4066Sahrens 3476fa9e4066Sahrens off = bfp->l_start; 3477104e2ed7Sperrin len = bfp->l_len; /* 0 means from off to end of file */ 3478104e2ed7Sperrin 34795730cc9aSmaybee do { 34805730cc9aSmaybee error = zfs_freesp(zp, off, len, flag, TRUE); 34818a2f1b91Sahrens /* NB: we already did dmu_tx_wait() if necessary */ 34825730cc9aSmaybee } while (error == ERESTART && zfsvfs->z_assign == TXG_NOWAIT); 3483fa9e4066Sahrens 3484fa9e4066Sahrens ZFS_EXIT(zfsvfs); 3485fa9e4066Sahrens return (error); 3486fa9e4066Sahrens } 3487fa9e4066Sahrens 3488fa9e4066Sahrens static int 3489fa9e4066Sahrens zfs_fid(vnode_t *vp, fid_t *fidp) 3490fa9e4066Sahrens { 3491fa9e4066Sahrens znode_t *zp = VTOZ(vp); 3492fa9e4066Sahrens zfsvfs_t *zfsvfs = zp->z_zfsvfs; 3493fa9e4066Sahrens uint32_t gen = (uint32_t)zp->z_phys->zp_gen; 3494fa9e4066Sahrens uint64_t object = zp->z_id; 3495fa9e4066Sahrens zfid_short_t *zfid; 3496fa9e4066Sahrens int size, i; 3497fa9e4066Sahrens 3498fa9e4066Sahrens ZFS_ENTER(zfsvfs); 3499fa9e4066Sahrens 3500fa9e4066Sahrens size = (zfsvfs->z_parent != zfsvfs) ? LONG_FID_LEN : SHORT_FID_LEN; 3501fa9e4066Sahrens if (fidp->fid_len < size) { 3502fa9e4066Sahrens fidp->fid_len = size; 35030f2dc02eSek ZFS_EXIT(zfsvfs); 3504fa9e4066Sahrens return (ENOSPC); 3505fa9e4066Sahrens } 3506fa9e4066Sahrens 3507fa9e4066Sahrens zfid = (zfid_short_t *)fidp; 3508fa9e4066Sahrens 3509fa9e4066Sahrens zfid->zf_len = size; 3510fa9e4066Sahrens 3511fa9e4066Sahrens for (i = 0; i < sizeof (zfid->zf_object); i++) 3512fa9e4066Sahrens zfid->zf_object[i] = (uint8_t)(object >> (8 * i)); 3513fa9e4066Sahrens 3514fa9e4066Sahrens /* Must have a non-zero generation number to distinguish from .zfs */ 3515fa9e4066Sahrens if (gen == 0) 3516fa9e4066Sahrens gen = 1; 3517fa9e4066Sahrens for (i = 0; i < sizeof (zfid->zf_gen); i++) 3518fa9e4066Sahrens zfid->zf_gen[i] = (uint8_t)(gen >> (8 * i)); 3519fa9e4066Sahrens 3520fa9e4066Sahrens if (size == LONG_FID_LEN) { 3521fa9e4066Sahrens uint64_t objsetid = dmu_objset_id(zfsvfs->z_os); 3522fa9e4066Sahrens zfid_long_t *zlfid; 3523fa9e4066Sahrens 3524fa9e4066Sahrens zlfid = (zfid_long_t *)fidp; 3525fa9e4066Sahrens 3526fa9e4066Sahrens for (i = 0; i < sizeof (zlfid->zf_setid); i++) 3527fa9e4066Sahrens zlfid->zf_setid[i] = (uint8_t)(objsetid >> (8 * i)); 3528fa9e4066Sahrens 3529fa9e4066Sahrens /* XXX - this should be the generation number for the objset */ 3530fa9e4066Sahrens for (i = 0; i < sizeof (zlfid->zf_setgen); i++) 3531fa9e4066Sahrens zlfid->zf_setgen[i] = 0; 3532fa9e4066Sahrens } 3533fa9e4066Sahrens 3534fa9e4066Sahrens ZFS_EXIT(zfsvfs); 3535fa9e4066Sahrens return (0); 3536fa9e4066Sahrens } 3537fa9e4066Sahrens 3538fa9e4066Sahrens static int 3539fa9e4066Sahrens zfs_pathconf(vnode_t *vp, int cmd, ulong_t *valp, cred_t *cr) 3540fa9e4066Sahrens { 3541fa9e4066Sahrens znode_t *zp, *xzp; 3542fa9e4066Sahrens zfsvfs_t *zfsvfs; 3543fa9e4066Sahrens zfs_dirlock_t *dl; 3544fa9e4066Sahrens int error; 3545fa9e4066Sahrens 3546fa9e4066Sahrens switch (cmd) { 3547fa9e4066Sahrens case _PC_LINK_MAX: 3548fa9e4066Sahrens *valp = ULONG_MAX; 3549fa9e4066Sahrens return (0); 3550fa9e4066Sahrens 3551fa9e4066Sahrens case _PC_FILESIZEBITS: 3552fa9e4066Sahrens *valp = 64; 3553fa9e4066Sahrens return (0); 3554fa9e4066Sahrens 3555fa9e4066Sahrens case _PC_XATTR_EXISTS: 3556fa9e4066Sahrens zp = VTOZ(vp); 3557fa9e4066Sahrens zfsvfs = zp->z_zfsvfs; 3558fa9e4066Sahrens ZFS_ENTER(zfsvfs); 3559fa9e4066Sahrens *valp = 0; 3560fa9e4066Sahrens error = zfs_dirent_lock(&dl, zp, "", &xzp, 3561fa9e4066Sahrens ZXATTR | ZEXISTS | ZSHARED); 3562fa9e4066Sahrens if (error == 0) { 3563fa9e4066Sahrens zfs_dirent_unlock(dl); 3564fa9e4066Sahrens if (!zfs_dirempty(xzp)) 3565fa9e4066Sahrens *valp = 1; 3566fa9e4066Sahrens VN_RELE(ZTOV(xzp)); 3567fa9e4066Sahrens } else if (error == ENOENT) { 3568fa9e4066Sahrens /* 3569fa9e4066Sahrens * If there aren't extended attributes, it's the 3570fa9e4066Sahrens * same as having zero of them. 3571fa9e4066Sahrens */ 3572fa9e4066Sahrens error = 0; 3573fa9e4066Sahrens } 3574fa9e4066Sahrens ZFS_EXIT(zfsvfs); 3575fa9e4066Sahrens return (error); 3576fa9e4066Sahrens 3577fa9e4066Sahrens case _PC_ACL_ENABLED: 3578fa9e4066Sahrens *valp = _ACL_ACE_ENABLED; 3579fa9e4066Sahrens return (0); 3580fa9e4066Sahrens 3581fa9e4066Sahrens case _PC_MIN_HOLE_SIZE: 3582fa9e4066Sahrens *valp = (ulong_t)SPA_MINBLOCKSIZE; 3583fa9e4066Sahrens return (0); 3584fa9e4066Sahrens 3585fa9e4066Sahrens default: 3586fa9e4066Sahrens return (fs_pathconf(vp, cmd, valp, cr)); 3587fa9e4066Sahrens } 3588fa9e4066Sahrens } 3589fa9e4066Sahrens 3590fa9e4066Sahrens /*ARGSUSED*/ 3591fa9e4066Sahrens static int 3592fa9e4066Sahrens zfs_getsecattr(vnode_t *vp, vsecattr_t *vsecp, int flag, cred_t *cr) 3593fa9e4066Sahrens { 3594fa9e4066Sahrens znode_t *zp = VTOZ(vp); 3595fa9e4066Sahrens zfsvfs_t *zfsvfs = zp->z_zfsvfs; 3596fa9e4066Sahrens int error; 3597fa9e4066Sahrens 3598fa9e4066Sahrens ZFS_ENTER(zfsvfs); 3599fa9e4066Sahrens error = zfs_getacl(zp, vsecp, cr); 3600fa9e4066Sahrens ZFS_EXIT(zfsvfs); 3601fa9e4066Sahrens 3602fa9e4066Sahrens return (error); 3603fa9e4066Sahrens } 3604fa9e4066Sahrens 3605fa9e4066Sahrens /*ARGSUSED*/ 3606fa9e4066Sahrens static int 3607fa9e4066Sahrens zfs_setsecattr(vnode_t *vp, vsecattr_t *vsecp, int flag, cred_t *cr) 3608fa9e4066Sahrens { 3609fa9e4066Sahrens znode_t *zp = VTOZ(vp); 3610fa9e4066Sahrens zfsvfs_t *zfsvfs = zp->z_zfsvfs; 3611fa9e4066Sahrens int error; 3612fa9e4066Sahrens 3613fa9e4066Sahrens ZFS_ENTER(zfsvfs); 3614fa9e4066Sahrens error = zfs_setacl(zp, vsecp, cr); 3615fa9e4066Sahrens ZFS_EXIT(zfsvfs); 3616fa9e4066Sahrens return (error); 3617fa9e4066Sahrens } 3618fa9e4066Sahrens 3619fa9e4066Sahrens /* 3620fa9e4066Sahrens * Predeclare these here so that the compiler assumes that 3621fa9e4066Sahrens * this is an "old style" function declaration that does 3622fa9e4066Sahrens * not include arguments => we won't get type mismatch errors 3623fa9e4066Sahrens * in the initializations that follow. 3624fa9e4066Sahrens */ 3625fa9e4066Sahrens static int zfs_inval(); 3626fa9e4066Sahrens static int zfs_isdir(); 3627fa9e4066Sahrens 3628fa9e4066Sahrens static int 3629fa9e4066Sahrens zfs_inval() 3630fa9e4066Sahrens { 3631fa9e4066Sahrens return (EINVAL); 3632fa9e4066Sahrens } 3633fa9e4066Sahrens 3634fa9e4066Sahrens static int 3635fa9e4066Sahrens zfs_isdir() 3636fa9e4066Sahrens { 3637fa9e4066Sahrens return (EISDIR); 3638fa9e4066Sahrens } 3639fa9e4066Sahrens /* 3640fa9e4066Sahrens * Directory vnode operations template 3641fa9e4066Sahrens */ 3642fa9e4066Sahrens vnodeops_t *zfs_dvnodeops; 3643fa9e4066Sahrens const fs_operation_def_t zfs_dvnodeops_template[] = { 3644fa9e4066Sahrens VOPNAME_OPEN, zfs_open, 3645fa9e4066Sahrens VOPNAME_CLOSE, zfs_close, 3646fa9e4066Sahrens VOPNAME_READ, zfs_isdir, 3647fa9e4066Sahrens VOPNAME_WRITE, zfs_isdir, 3648fa9e4066Sahrens VOPNAME_IOCTL, zfs_ioctl, 3649fa9e4066Sahrens VOPNAME_GETATTR, zfs_getattr, 3650fa9e4066Sahrens VOPNAME_SETATTR, zfs_setattr, 3651fa9e4066Sahrens VOPNAME_ACCESS, zfs_access, 3652fa9e4066Sahrens VOPNAME_LOOKUP, zfs_lookup, 3653fa9e4066Sahrens VOPNAME_CREATE, zfs_create, 3654fa9e4066Sahrens VOPNAME_REMOVE, zfs_remove, 3655fa9e4066Sahrens VOPNAME_LINK, zfs_link, 3656fa9e4066Sahrens VOPNAME_RENAME, zfs_rename, 3657fa9e4066Sahrens VOPNAME_MKDIR, zfs_mkdir, 3658fa9e4066Sahrens VOPNAME_RMDIR, zfs_rmdir, 3659fa9e4066Sahrens VOPNAME_READDIR, zfs_readdir, 3660fa9e4066Sahrens VOPNAME_SYMLINK, zfs_symlink, 3661fa9e4066Sahrens VOPNAME_FSYNC, zfs_fsync, 3662fa9e4066Sahrens VOPNAME_INACTIVE, (fs_generic_func_p) zfs_inactive, 3663fa9e4066Sahrens VOPNAME_FID, zfs_fid, 3664fa9e4066Sahrens VOPNAME_SEEK, zfs_seek, 3665fa9e4066Sahrens VOPNAME_PATHCONF, zfs_pathconf, 3666fa9e4066Sahrens VOPNAME_GETSECATTR, zfs_getsecattr, 3667fa9e4066Sahrens VOPNAME_SETSECATTR, zfs_setsecattr, 3668fa9e4066Sahrens NULL, NULL 3669fa9e4066Sahrens }; 3670fa9e4066Sahrens 3671fa9e4066Sahrens /* 3672fa9e4066Sahrens * Regular file vnode operations template 3673fa9e4066Sahrens */ 3674fa9e4066Sahrens vnodeops_t *zfs_fvnodeops; 3675fa9e4066Sahrens const fs_operation_def_t zfs_fvnodeops_template[] = { 3676fa9e4066Sahrens VOPNAME_OPEN, zfs_open, 3677fa9e4066Sahrens VOPNAME_CLOSE, zfs_close, 3678fa9e4066Sahrens VOPNAME_READ, zfs_read, 3679fa9e4066Sahrens VOPNAME_WRITE, zfs_write, 3680fa9e4066Sahrens VOPNAME_IOCTL, zfs_ioctl, 3681fa9e4066Sahrens VOPNAME_GETATTR, zfs_getattr, 3682fa9e4066Sahrens VOPNAME_SETATTR, zfs_setattr, 3683fa9e4066Sahrens VOPNAME_ACCESS, zfs_access, 3684fa9e4066Sahrens VOPNAME_LOOKUP, zfs_lookup, 3685fa9e4066Sahrens VOPNAME_RENAME, zfs_rename, 3686fa9e4066Sahrens VOPNAME_FSYNC, zfs_fsync, 3687fa9e4066Sahrens VOPNAME_INACTIVE, (fs_generic_func_p)zfs_inactive, 3688fa9e4066Sahrens VOPNAME_FID, zfs_fid, 3689fa9e4066Sahrens VOPNAME_SEEK, zfs_seek, 3690fa9e4066Sahrens VOPNAME_FRLOCK, zfs_frlock, 3691fa9e4066Sahrens VOPNAME_SPACE, zfs_space, 3692fa9e4066Sahrens VOPNAME_GETPAGE, zfs_getpage, 3693fa9e4066Sahrens VOPNAME_PUTPAGE, zfs_putpage, 3694fa9e4066Sahrens VOPNAME_MAP, (fs_generic_func_p) zfs_map, 3695fa9e4066Sahrens VOPNAME_ADDMAP, (fs_generic_func_p) zfs_addmap, 3696fa9e4066Sahrens VOPNAME_DELMAP, zfs_delmap, 3697fa9e4066Sahrens VOPNAME_PATHCONF, zfs_pathconf, 3698fa9e4066Sahrens VOPNAME_GETSECATTR, zfs_getsecattr, 3699fa9e4066Sahrens VOPNAME_SETSECATTR, zfs_setsecattr, 3700fa9e4066Sahrens VOPNAME_VNEVENT, fs_vnevent_support, 3701fa9e4066Sahrens NULL, NULL 3702fa9e4066Sahrens }; 3703fa9e4066Sahrens 3704fa9e4066Sahrens /* 3705fa9e4066Sahrens * Symbolic link vnode operations template 3706fa9e4066Sahrens */ 3707fa9e4066Sahrens vnodeops_t *zfs_symvnodeops; 3708fa9e4066Sahrens const fs_operation_def_t zfs_symvnodeops_template[] = { 3709fa9e4066Sahrens VOPNAME_GETATTR, zfs_getattr, 3710fa9e4066Sahrens VOPNAME_SETATTR, zfs_setattr, 3711fa9e4066Sahrens VOPNAME_ACCESS, zfs_access, 3712fa9e4066Sahrens VOPNAME_RENAME, zfs_rename, 3713fa9e4066Sahrens VOPNAME_READLINK, zfs_readlink, 3714fa9e4066Sahrens VOPNAME_INACTIVE, (fs_generic_func_p) zfs_inactive, 3715fa9e4066Sahrens VOPNAME_FID, zfs_fid, 3716fa9e4066Sahrens VOPNAME_PATHCONF, zfs_pathconf, 3717fa9e4066Sahrens VOPNAME_VNEVENT, fs_vnevent_support, 3718fa9e4066Sahrens NULL, NULL 3719fa9e4066Sahrens }; 3720fa9e4066Sahrens 3721fa9e4066Sahrens /* 3722fa9e4066Sahrens * Extended attribute directory vnode operations template 3723fa9e4066Sahrens * This template is identical to the directory vnodes 3724fa9e4066Sahrens * operation template except for restricted operations: 3725fa9e4066Sahrens * VOP_MKDIR() 3726fa9e4066Sahrens * VOP_SYMLINK() 3727fa9e4066Sahrens * Note that there are other restrictions embedded in: 3728fa9e4066Sahrens * zfs_create() - restrict type to VREG 3729fa9e4066Sahrens * zfs_link() - no links into/out of attribute space 3730fa9e4066Sahrens * zfs_rename() - no moves into/out of attribute space 3731fa9e4066Sahrens */ 3732fa9e4066Sahrens vnodeops_t *zfs_xdvnodeops; 3733fa9e4066Sahrens const fs_operation_def_t zfs_xdvnodeops_template[] = { 3734fa9e4066Sahrens VOPNAME_OPEN, zfs_open, 3735fa9e4066Sahrens VOPNAME_CLOSE, zfs_close, 3736fa9e4066Sahrens VOPNAME_IOCTL, zfs_ioctl, 3737fa9e4066Sahrens VOPNAME_GETATTR, zfs_getattr, 3738fa9e4066Sahrens VOPNAME_SETATTR, zfs_setattr, 3739fa9e4066Sahrens VOPNAME_ACCESS, zfs_access, 3740fa9e4066Sahrens VOPNAME_LOOKUP, zfs_lookup, 3741fa9e4066Sahrens VOPNAME_CREATE, zfs_create, 3742fa9e4066Sahrens VOPNAME_REMOVE, zfs_remove, 3743fa9e4066Sahrens VOPNAME_LINK, zfs_link, 3744fa9e4066Sahrens VOPNAME_RENAME, zfs_rename, 3745fa9e4066Sahrens VOPNAME_MKDIR, zfs_inval, 3746fa9e4066Sahrens VOPNAME_RMDIR, zfs_rmdir, 3747fa9e4066Sahrens VOPNAME_READDIR, zfs_readdir, 3748fa9e4066Sahrens VOPNAME_SYMLINK, zfs_inval, 3749fa9e4066Sahrens VOPNAME_FSYNC, zfs_fsync, 3750fa9e4066Sahrens VOPNAME_INACTIVE, (fs_generic_func_p) zfs_inactive, 3751fa9e4066Sahrens VOPNAME_FID, zfs_fid, 3752fa9e4066Sahrens VOPNAME_SEEK, zfs_seek, 3753fa9e4066Sahrens VOPNAME_PATHCONF, zfs_pathconf, 3754fa9e4066Sahrens VOPNAME_GETSECATTR, zfs_getsecattr, 3755fa9e4066Sahrens VOPNAME_SETSECATTR, zfs_setsecattr, 3756fa9e4066Sahrens VOPNAME_VNEVENT, fs_vnevent_support, 3757fa9e4066Sahrens NULL, NULL 3758fa9e4066Sahrens }; 3759fa9e4066Sahrens 3760fa9e4066Sahrens /* 3761fa9e4066Sahrens * Error vnode operations template 3762fa9e4066Sahrens */ 3763fa9e4066Sahrens vnodeops_t *zfs_evnodeops; 3764fa9e4066Sahrens const fs_operation_def_t zfs_evnodeops_template[] = { 3765fa9e4066Sahrens VOPNAME_INACTIVE, (fs_generic_func_p) zfs_inactive, 3766fa9e4066Sahrens VOPNAME_PATHCONF, zfs_pathconf, 3767fa9e4066Sahrens NULL, NULL 3768fa9e4066Sahrens }; 3769