17c478bd9Sstevel@tonic-gate /* 27c478bd9Sstevel@tonic-gate * CDDL HEADER START 37c478bd9Sstevel@tonic-gate * 47c478bd9Sstevel@tonic-gate * The contents of this file are subject to the terms of the 57c478bd9Sstevel@tonic-gate * Common Development and Distribution License, Version 1.0 only 67c478bd9Sstevel@tonic-gate * (the "License"). You may not use this file except in compliance 77c478bd9Sstevel@tonic-gate * with the License. 87c478bd9Sstevel@tonic-gate * 97c478bd9Sstevel@tonic-gate * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 107c478bd9Sstevel@tonic-gate * or http://www.opensolaris.org/os/licensing. 117c478bd9Sstevel@tonic-gate * See the License for the specific language governing permissions 127c478bd9Sstevel@tonic-gate * and limitations under the License. 137c478bd9Sstevel@tonic-gate * 147c478bd9Sstevel@tonic-gate * When distributing Covered Code, include this CDDL HEADER in each 157c478bd9Sstevel@tonic-gate * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 167c478bd9Sstevel@tonic-gate * If applicable, add the following below this CDDL HEADER, with the 177c478bd9Sstevel@tonic-gate * fields enclosed by brackets "[]" replaced with your own identifying 187c478bd9Sstevel@tonic-gate * information: Portions Copyright [yyyy] [name of copyright owner] 197c478bd9Sstevel@tonic-gate * 207c478bd9Sstevel@tonic-gate * CDDL HEADER END 217c478bd9Sstevel@tonic-gate */ 227c478bd9Sstevel@tonic-gate /* 237c478bd9Sstevel@tonic-gate * Copyright 2005 Sun Microsystems, Inc. All rights reserved. 247c478bd9Sstevel@tonic-gate * Use is subject to license terms. 257c478bd9Sstevel@tonic-gate */ 267c478bd9Sstevel@tonic-gate 277c478bd9Sstevel@tonic-gate /* Copyright (c) 1983, 1984, 1985, 1986, 1987, 1988, 1989 AT&T */ 287c478bd9Sstevel@tonic-gate /* All Rights Reserved */ 297c478bd9Sstevel@tonic-gate 307c478bd9Sstevel@tonic-gate /* 317c478bd9Sstevel@tonic-gate * Portions of this source code were derived from Berkeley 4.3 BSD 327c478bd9Sstevel@tonic-gate * under license from the Regents of the University of California. 337c478bd9Sstevel@tonic-gate */ 347c478bd9Sstevel@tonic-gate 357c478bd9Sstevel@tonic-gate #pragma ident "%Z%%M% %I% %E% SMI" 367c478bd9Sstevel@tonic-gate 377c478bd9Sstevel@tonic-gate #include <sys/types.h> 387c478bd9Sstevel@tonic-gate #include <sys/t_lock.h> 397c478bd9Sstevel@tonic-gate #include <sys/ksynch.h> 407c478bd9Sstevel@tonic-gate #include <sys/param.h> 417c478bd9Sstevel@tonic-gate #include <sys/time.h> 427c478bd9Sstevel@tonic-gate #include <sys/systm.h> 437c478bd9Sstevel@tonic-gate #include <sys/sysmacros.h> 447c478bd9Sstevel@tonic-gate #include <sys/resource.h> 457c478bd9Sstevel@tonic-gate #include <sys/signal.h> 467c478bd9Sstevel@tonic-gate #include <sys/cred.h> 477c478bd9Sstevel@tonic-gate #include <sys/user.h> 487c478bd9Sstevel@tonic-gate #include <sys/buf.h> 497c478bd9Sstevel@tonic-gate #include <sys/vfs.h> 507c478bd9Sstevel@tonic-gate #include <sys/vnode.h> 517c478bd9Sstevel@tonic-gate #include <sys/proc.h> 527c478bd9Sstevel@tonic-gate #include <sys/disp.h> 537c478bd9Sstevel@tonic-gate #include <sys/file.h> 547c478bd9Sstevel@tonic-gate #include <sys/fcntl.h> 557c478bd9Sstevel@tonic-gate #include <sys/flock.h> 56*bc69f433Saguzovsk #include <sys/atomic.h> 577c478bd9Sstevel@tonic-gate #include <sys/kmem.h> 587c478bd9Sstevel@tonic-gate #include <sys/uio.h> 597c478bd9Sstevel@tonic-gate #include <sys/dnlc.h> 607c478bd9Sstevel@tonic-gate #include <sys/conf.h> 617c478bd9Sstevel@tonic-gate #include <sys/mman.h> 627c478bd9Sstevel@tonic-gate #include <sys/pathname.h> 637c478bd9Sstevel@tonic-gate #include <sys/debug.h> 647c478bd9Sstevel@tonic-gate #include <sys/vmsystm.h> 657c478bd9Sstevel@tonic-gate #include <sys/cmn_err.h> 667c478bd9Sstevel@tonic-gate #include <sys/vtrace.h> 677c478bd9Sstevel@tonic-gate #include <sys/filio.h> 687c478bd9Sstevel@tonic-gate #include <sys/policy.h> 697c478bd9Sstevel@tonic-gate 707c478bd9Sstevel@tonic-gate #include <sys/fs/ufs_fs.h> 717c478bd9Sstevel@tonic-gate #include <sys/fs/ufs_lockfs.h> 727c478bd9Sstevel@tonic-gate #include <sys/fs/ufs_filio.h> 737c478bd9Sstevel@tonic-gate #include <sys/fs/ufs_inode.h> 747c478bd9Sstevel@tonic-gate #include <sys/fs/ufs_fsdir.h> 757c478bd9Sstevel@tonic-gate #include <sys/fs/ufs_quota.h> 767c478bd9Sstevel@tonic-gate #include <sys/fs/ufs_log.h> 777c478bd9Sstevel@tonic-gate #include <sys/fs/ufs_snap.h> 787c478bd9Sstevel@tonic-gate #include <sys/fs/ufs_trans.h> 797c478bd9Sstevel@tonic-gate #include <sys/fs/ufs_panic.h> 807c478bd9Sstevel@tonic-gate #include <sys/fs/ufs_bio.h> 817c478bd9Sstevel@tonic-gate #include <sys/dirent.h> /* must be AFTER <sys/fs/fsdir.h>! */ 827c478bd9Sstevel@tonic-gate #include <sys/errno.h> 837c478bd9Sstevel@tonic-gate #include <sys/fssnap_if.h> 847c478bd9Sstevel@tonic-gate #include <sys/unistd.h> 857c478bd9Sstevel@tonic-gate #include <sys/sunddi.h> 867c478bd9Sstevel@tonic-gate 877c478bd9Sstevel@tonic-gate #include <sys/filio.h> /* _FIOIO */ 887c478bd9Sstevel@tonic-gate 897c478bd9Sstevel@tonic-gate #include <vm/hat.h> 907c478bd9Sstevel@tonic-gate #include <vm/page.h> 917c478bd9Sstevel@tonic-gate #include <vm/pvn.h> 927c478bd9Sstevel@tonic-gate #include <vm/as.h> 937c478bd9Sstevel@tonic-gate #include <vm/seg.h> 947c478bd9Sstevel@tonic-gate #include <vm/seg_map.h> 957c478bd9Sstevel@tonic-gate #include <vm/seg_vn.h> 967c478bd9Sstevel@tonic-gate #include <vm/seg_kmem.h> 977c478bd9Sstevel@tonic-gate #include <vm/rm.h> 987c478bd9Sstevel@tonic-gate #include <sys/swap.h> 997c478bd9Sstevel@tonic-gate 1007c478bd9Sstevel@tonic-gate #include <fs/fs_subr.h> 1017c478bd9Sstevel@tonic-gate 1027c478bd9Sstevel@tonic-gate static struct instats ins; 1037c478bd9Sstevel@tonic-gate 1047c478bd9Sstevel@tonic-gate static int ufs_getpage_ra(struct vnode *, u_offset_t, struct seg *, caddr_t); 1057c478bd9Sstevel@tonic-gate static int ufs_getpage_miss(struct vnode *, u_offset_t, size_t, struct seg *, 1067c478bd9Sstevel@tonic-gate caddr_t, struct page **, size_t, enum seg_rw, int); 1077c478bd9Sstevel@tonic-gate static int ufs_open(struct vnode **, int, struct cred *); 1087c478bd9Sstevel@tonic-gate static int ufs_close(struct vnode *, int, int, offset_t, struct cred *); 1097c478bd9Sstevel@tonic-gate static int ufs_read(struct vnode *, struct uio *, int, struct cred *, 1107c478bd9Sstevel@tonic-gate struct caller_context *); 1117c478bd9Sstevel@tonic-gate static int ufs_write(struct vnode *, struct uio *, int, struct cred *, 1127c478bd9Sstevel@tonic-gate struct caller_context *); 1137c478bd9Sstevel@tonic-gate static int ufs_ioctl(struct vnode *, int, intptr_t, int, struct cred *, int *); 1147c478bd9Sstevel@tonic-gate static int ufs_getattr(struct vnode *, struct vattr *, int, struct cred *); 1157c478bd9Sstevel@tonic-gate static int ufs_setattr(struct vnode *, struct vattr *, int, struct cred *, 1167c478bd9Sstevel@tonic-gate caller_context_t *); 1177c478bd9Sstevel@tonic-gate static int ufs_access(struct vnode *, int, int, struct cred *); 1187c478bd9Sstevel@tonic-gate static int ufs_lookup(struct vnode *, char *, struct vnode **, 1197c478bd9Sstevel@tonic-gate struct pathname *, int, struct vnode *, struct cred *); 1207c478bd9Sstevel@tonic-gate static int ufs_create(struct vnode *, char *, struct vattr *, enum vcexcl, 1217c478bd9Sstevel@tonic-gate int, struct vnode **, struct cred *, int); 1227c478bd9Sstevel@tonic-gate static int ufs_remove(struct vnode *, char *, struct cred *); 1237c478bd9Sstevel@tonic-gate static int ufs_link(struct vnode *, struct vnode *, char *, struct cred *); 1247c478bd9Sstevel@tonic-gate static int ufs_rename(struct vnode *, char *, struct vnode *, char *, 1257c478bd9Sstevel@tonic-gate struct cred *); 1267c478bd9Sstevel@tonic-gate static int ufs_mkdir(struct vnode *, char *, struct vattr *, struct vnode **, 1277c478bd9Sstevel@tonic-gate struct cred *); 1287c478bd9Sstevel@tonic-gate static int ufs_rmdir(struct vnode *, char *, struct vnode *, struct cred *); 1297c478bd9Sstevel@tonic-gate static int ufs_readdir(struct vnode *, struct uio *, struct cred *, int *); 1307c478bd9Sstevel@tonic-gate static int ufs_symlink(struct vnode *, char *, struct vattr *, char *, 1317c478bd9Sstevel@tonic-gate struct cred *); 1327c478bd9Sstevel@tonic-gate static int ufs_readlink(struct vnode *, struct uio *, struct cred *); 1337c478bd9Sstevel@tonic-gate static int ufs_fsync(struct vnode *, int, struct cred *); 1347c478bd9Sstevel@tonic-gate static void ufs_inactive(struct vnode *, struct cred *); 1357c478bd9Sstevel@tonic-gate static int ufs_fid(struct vnode *, struct fid *); 1367c478bd9Sstevel@tonic-gate static int ufs_rwlock(struct vnode *, int, caller_context_t *); 1377c478bd9Sstevel@tonic-gate static void ufs_rwunlock(struct vnode *, int, caller_context_t *); 1387c478bd9Sstevel@tonic-gate static int ufs_seek(struct vnode *, offset_t, offset_t *); 1397c478bd9Sstevel@tonic-gate static int ufs_frlock(struct vnode *, int, struct flock64 *, int, offset_t, 1407c478bd9Sstevel@tonic-gate struct flk_callback *, struct cred *); 1417c478bd9Sstevel@tonic-gate static int ufs_space(struct vnode *, int, struct flock64 *, int, offset_t, 1427c478bd9Sstevel@tonic-gate cred_t *, caller_context_t *); 1437c478bd9Sstevel@tonic-gate static int ufs_getpage(struct vnode *, offset_t, size_t, uint_t *, 1447c478bd9Sstevel@tonic-gate struct page **, size_t, struct seg *, caddr_t, 1457c478bd9Sstevel@tonic-gate enum seg_rw, struct cred *); 1467c478bd9Sstevel@tonic-gate static int ufs_putpage(struct vnode *, offset_t, size_t, int, struct cred *); 1477c478bd9Sstevel@tonic-gate static int ufs_putpages(struct vnode *, offset_t, size_t, int, struct cred *); 1487c478bd9Sstevel@tonic-gate static int ufs_map(struct vnode *, offset_t, struct as *, caddr_t *, size_t, 1497c478bd9Sstevel@tonic-gate uchar_t, uchar_t, uint_t, struct cred *); 1507c478bd9Sstevel@tonic-gate static int ufs_addmap(struct vnode *, offset_t, struct as *, caddr_t, size_t, 1517c478bd9Sstevel@tonic-gate uchar_t, uchar_t, uint_t, struct cred *); 1527c478bd9Sstevel@tonic-gate static int ufs_delmap(struct vnode *, offset_t, struct as *, caddr_t, size_t, 1537c478bd9Sstevel@tonic-gate uint_t, uint_t, uint_t, struct cred *); 1547c478bd9Sstevel@tonic-gate static int ufs_poll(vnode_t *, short, int, short *, struct pollhead **); 1557c478bd9Sstevel@tonic-gate static int ufs_dump(vnode_t *, caddr_t, int, int); 1567c478bd9Sstevel@tonic-gate static int ufs_l_pathconf(struct vnode *, int, ulong_t *, struct cred *); 1577c478bd9Sstevel@tonic-gate static int ufs_pageio(struct vnode *, struct page *, u_offset_t, size_t, int, 1587c478bd9Sstevel@tonic-gate struct cred *); 1597c478bd9Sstevel@tonic-gate static int ufs_dump(vnode_t *, caddr_t, int, int); 1607c478bd9Sstevel@tonic-gate static int ufs_dumpctl(vnode_t *, int, int *); 1617c478bd9Sstevel@tonic-gate static daddr32_t *save_dblks(struct inode *, struct ufsvfs *, daddr32_t *, 1627c478bd9Sstevel@tonic-gate daddr32_t *, int, int); 1637c478bd9Sstevel@tonic-gate static int ufs_getsecattr(struct vnode *, vsecattr_t *, int, struct cred *); 1647c478bd9Sstevel@tonic-gate static int ufs_setsecattr(struct vnode *, vsecattr_t *, int, struct cred *); 1657c478bd9Sstevel@tonic-gate 1667c478bd9Sstevel@tonic-gate /* 1677c478bd9Sstevel@tonic-gate * For lockfs: ulockfs begin/end is now inlined in the ufs_xxx functions. 1687c478bd9Sstevel@tonic-gate * 1697c478bd9Sstevel@tonic-gate * XXX - ULOCKFS in fs_pathconf and ufs_ioctl is not inlined yet. 1707c478bd9Sstevel@tonic-gate */ 1717c478bd9Sstevel@tonic-gate struct vnodeops *ufs_vnodeops; 1727c478bd9Sstevel@tonic-gate 1737c478bd9Sstevel@tonic-gate const fs_operation_def_t ufs_vnodeops_template[] = { 1747c478bd9Sstevel@tonic-gate VOPNAME_OPEN, ufs_open, /* will not be blocked by lockfs */ 1757c478bd9Sstevel@tonic-gate VOPNAME_CLOSE, ufs_close, /* will not be blocked by lockfs */ 1767c478bd9Sstevel@tonic-gate VOPNAME_READ, ufs_read, 1777c478bd9Sstevel@tonic-gate VOPNAME_WRITE, ufs_write, 1787c478bd9Sstevel@tonic-gate VOPNAME_IOCTL, ufs_ioctl, 1797c478bd9Sstevel@tonic-gate VOPNAME_GETATTR, ufs_getattr, 1807c478bd9Sstevel@tonic-gate VOPNAME_SETATTR, ufs_setattr, 1817c478bd9Sstevel@tonic-gate VOPNAME_ACCESS, ufs_access, 1827c478bd9Sstevel@tonic-gate VOPNAME_LOOKUP, ufs_lookup, 1837c478bd9Sstevel@tonic-gate VOPNAME_CREATE, ufs_create, 1847c478bd9Sstevel@tonic-gate VOPNAME_REMOVE, ufs_remove, 1857c478bd9Sstevel@tonic-gate VOPNAME_LINK, ufs_link, 1867c478bd9Sstevel@tonic-gate VOPNAME_RENAME, ufs_rename, 1877c478bd9Sstevel@tonic-gate VOPNAME_MKDIR, ufs_mkdir, 1887c478bd9Sstevel@tonic-gate VOPNAME_RMDIR, ufs_rmdir, 1897c478bd9Sstevel@tonic-gate VOPNAME_READDIR, ufs_readdir, 1907c478bd9Sstevel@tonic-gate VOPNAME_SYMLINK, ufs_symlink, 1917c478bd9Sstevel@tonic-gate VOPNAME_READLINK, ufs_readlink, 1927c478bd9Sstevel@tonic-gate VOPNAME_FSYNC, ufs_fsync, 1937c478bd9Sstevel@tonic-gate VOPNAME_INACTIVE, (fs_generic_func_p) ufs_inactive, /* not blocked */ 1947c478bd9Sstevel@tonic-gate VOPNAME_FID, ufs_fid, 1957c478bd9Sstevel@tonic-gate VOPNAME_RWLOCK, ufs_rwlock, /* not blocked */ 1967c478bd9Sstevel@tonic-gate VOPNAME_RWUNLOCK, (fs_generic_func_p) ufs_rwunlock, /* not blocked */ 1977c478bd9Sstevel@tonic-gate VOPNAME_SEEK, ufs_seek, 1987c478bd9Sstevel@tonic-gate VOPNAME_FRLOCK, ufs_frlock, 1997c478bd9Sstevel@tonic-gate VOPNAME_SPACE, ufs_space, 2007c478bd9Sstevel@tonic-gate VOPNAME_GETPAGE, ufs_getpage, 2017c478bd9Sstevel@tonic-gate VOPNAME_PUTPAGE, ufs_putpage, 2027c478bd9Sstevel@tonic-gate VOPNAME_MAP, (fs_generic_func_p) ufs_map, 2037c478bd9Sstevel@tonic-gate VOPNAME_ADDMAP, (fs_generic_func_p) ufs_addmap, /* not blocked */ 2047c478bd9Sstevel@tonic-gate VOPNAME_DELMAP, ufs_delmap, /* will not be blocked by lockfs */ 2057c478bd9Sstevel@tonic-gate VOPNAME_POLL, (fs_generic_func_p) ufs_poll, /* not blocked */ 2067c478bd9Sstevel@tonic-gate VOPNAME_DUMP, ufs_dump, 2077c478bd9Sstevel@tonic-gate VOPNAME_PATHCONF, ufs_l_pathconf, 2087c478bd9Sstevel@tonic-gate VOPNAME_PAGEIO, ufs_pageio, 2097c478bd9Sstevel@tonic-gate VOPNAME_DUMPCTL, ufs_dumpctl, 2107c478bd9Sstevel@tonic-gate VOPNAME_GETSECATTR, ufs_getsecattr, 2117c478bd9Sstevel@tonic-gate VOPNAME_SETSECATTR, ufs_setsecattr, 2127c478bd9Sstevel@tonic-gate VOPNAME_VNEVENT, fs_vnevent_support, 2137c478bd9Sstevel@tonic-gate NULL, NULL 2147c478bd9Sstevel@tonic-gate }; 2157c478bd9Sstevel@tonic-gate 2167c478bd9Sstevel@tonic-gate #define MAX_BACKFILE_COUNT 9999 2177c478bd9Sstevel@tonic-gate 2187c478bd9Sstevel@tonic-gate /* 2197c478bd9Sstevel@tonic-gate * Created by ufs_dumpctl() to store a file's disk block info into memory. 2207c478bd9Sstevel@tonic-gate * Used by ufs_dump() to dump data to disk directly. 2217c478bd9Sstevel@tonic-gate */ 2227c478bd9Sstevel@tonic-gate struct dump { 2237c478bd9Sstevel@tonic-gate struct inode *ip; /* the file we contain */ 2247c478bd9Sstevel@tonic-gate daddr_t fsbs; /* number of blocks stored */ 2257c478bd9Sstevel@tonic-gate struct timeval32 time; /* time stamp for the struct */ 2267c478bd9Sstevel@tonic-gate daddr32_t dblk[1]; /* place holder for block info */ 2277c478bd9Sstevel@tonic-gate }; 2287c478bd9Sstevel@tonic-gate 2297c478bd9Sstevel@tonic-gate static struct dump *dump_info = NULL; 2307c478bd9Sstevel@tonic-gate 2317c478bd9Sstevel@tonic-gate /* 2327c478bd9Sstevel@tonic-gate * Previously there was no special action required for ordinary files. 2337c478bd9Sstevel@tonic-gate * (Devices are handled through the device file system.) 2347c478bd9Sstevel@tonic-gate * Now we support Large Files and Large File API requires open to 2357c478bd9Sstevel@tonic-gate * fail if file is large. 2367c478bd9Sstevel@tonic-gate * We could take care to prevent data corruption 2377c478bd9Sstevel@tonic-gate * by doing an atomic check of size and truncate if file is opened with 2387c478bd9Sstevel@tonic-gate * FTRUNC flag set but traditionally this is being done by the vfs/vnode 2397c478bd9Sstevel@tonic-gate * layers. So taking care of truncation here is a change in the existing 2407c478bd9Sstevel@tonic-gate * semantics of VOP_OPEN and therefore we chose not to implement any thing 2417c478bd9Sstevel@tonic-gate * here. The check for the size of the file > 2GB is being done at the 2427c478bd9Sstevel@tonic-gate * vfs layer in routine vn_open(). 2437c478bd9Sstevel@tonic-gate */ 2447c478bd9Sstevel@tonic-gate 2457c478bd9Sstevel@tonic-gate /* ARGSUSED */ 2467c478bd9Sstevel@tonic-gate static int 2477c478bd9Sstevel@tonic-gate ufs_open(struct vnode **vpp, int flag, struct cred *cr) 2487c478bd9Sstevel@tonic-gate { 2497c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_OPEN, "ufs_open:vpp %p", vpp); 2507c478bd9Sstevel@tonic-gate return (0); 2517c478bd9Sstevel@tonic-gate } 2527c478bd9Sstevel@tonic-gate 2537c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 2547c478bd9Sstevel@tonic-gate static int 2557c478bd9Sstevel@tonic-gate ufs_close(struct vnode *vp, int flag, int count, offset_t offset, 2567c478bd9Sstevel@tonic-gate struct cred *cr) 2577c478bd9Sstevel@tonic-gate { 2587c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_CLOSE, "ufs_close:vp %p", vp); 2597c478bd9Sstevel@tonic-gate 2607c478bd9Sstevel@tonic-gate cleanlocks(vp, ttoproc(curthread)->p_pid, 0); 2617c478bd9Sstevel@tonic-gate cleanshares(vp, ttoproc(curthread)->p_pid); 2627c478bd9Sstevel@tonic-gate 2637c478bd9Sstevel@tonic-gate /* 2647c478bd9Sstevel@tonic-gate * Push partially filled cluster at last close. 2657c478bd9Sstevel@tonic-gate * ``last close'' is approximated because the dnlc 2667c478bd9Sstevel@tonic-gate * may have a hold on the vnode. 2677c478bd9Sstevel@tonic-gate * Checking for VBAD here will also act as a forced umount check. 2687c478bd9Sstevel@tonic-gate */ 2697c478bd9Sstevel@tonic-gate if (vp->v_count <= 2 && vp->v_type != VBAD) { 2707c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 2717c478bd9Sstevel@tonic-gate if (ip->i_delaylen) { 2727c478bd9Sstevel@tonic-gate ins.in_poc.value.ul++; 2737c478bd9Sstevel@tonic-gate (void) ufs_putpages(vp, ip->i_delayoff, ip->i_delaylen, 2747c478bd9Sstevel@tonic-gate B_ASYNC | B_FREE, cr); 2757c478bd9Sstevel@tonic-gate ip->i_delaylen = 0; 2767c478bd9Sstevel@tonic-gate } 2777c478bd9Sstevel@tonic-gate } 2787c478bd9Sstevel@tonic-gate 2797c478bd9Sstevel@tonic-gate return (0); 2807c478bd9Sstevel@tonic-gate } 2817c478bd9Sstevel@tonic-gate 2827c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 2837c478bd9Sstevel@tonic-gate static int 2847c478bd9Sstevel@tonic-gate ufs_read(struct vnode *vp, struct uio *uiop, int ioflag, struct cred *cr, 2857c478bd9Sstevel@tonic-gate struct caller_context *ct) 2867c478bd9Sstevel@tonic-gate { 2877c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 2887c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp; 2897c478bd9Sstevel@tonic-gate struct ulockfs *ulp = NULL; 2907c478bd9Sstevel@tonic-gate int error = 0; 2917c478bd9Sstevel@tonic-gate int intrans = 0; 2927c478bd9Sstevel@tonic-gate 2937c478bd9Sstevel@tonic-gate ASSERT(RW_READ_HELD(&ip->i_rwlock)); 2947c478bd9Sstevel@tonic-gate TRACE_3(TR_FAC_UFS, TR_UFS_READ_START, 2957c478bd9Sstevel@tonic-gate "ufs_read_start:vp %p uiop %p ioflag %x", 2967c478bd9Sstevel@tonic-gate vp, uiop, ioflag); 2977c478bd9Sstevel@tonic-gate 2987c478bd9Sstevel@tonic-gate /* 2997c478bd9Sstevel@tonic-gate * Mandatory locking needs to be done before ufs_lockfs_begin() 3007c478bd9Sstevel@tonic-gate * and TRANS_BEGIN_SYNC() calls since mandatory locks can sleep. 3017c478bd9Sstevel@tonic-gate */ 3027c478bd9Sstevel@tonic-gate if (MANDLOCK(vp, ip->i_mode)) { 3037c478bd9Sstevel@tonic-gate /* 3047c478bd9Sstevel@tonic-gate * ufs_getattr ends up being called by chklock 3057c478bd9Sstevel@tonic-gate */ 3067c478bd9Sstevel@tonic-gate error = chklock(vp, FREAD, uiop->uio_loffset, 3077c478bd9Sstevel@tonic-gate uiop->uio_resid, uiop->uio_fmode, ct); 3087c478bd9Sstevel@tonic-gate if (error) 3097c478bd9Sstevel@tonic-gate goto out; 3107c478bd9Sstevel@tonic-gate } 3117c478bd9Sstevel@tonic-gate 3127c478bd9Sstevel@tonic-gate ufsvfsp = ip->i_ufsvfs; 3137c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_READ_MASK); 3147c478bd9Sstevel@tonic-gate if (error) 3157c478bd9Sstevel@tonic-gate goto out; 3167c478bd9Sstevel@tonic-gate 3177c478bd9Sstevel@tonic-gate /* 3187c478bd9Sstevel@tonic-gate * In the case that a directory is opened for reading as a file 3197c478bd9Sstevel@tonic-gate * (eg "cat .") with the O_RSYNC, O_SYNC and O_DSYNC flags set. 3207c478bd9Sstevel@tonic-gate * The locking order had to be changed to avoid a deadlock with 3217c478bd9Sstevel@tonic-gate * an update taking place on that directory at the same time. 3227c478bd9Sstevel@tonic-gate */ 3237c478bd9Sstevel@tonic-gate if ((ip->i_mode & IFMT) == IFDIR) { 3247c478bd9Sstevel@tonic-gate 3257c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_READER); 3267c478bd9Sstevel@tonic-gate error = rdip(ip, uiop, ioflag, cr); 3277c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 3287c478bd9Sstevel@tonic-gate 3297c478bd9Sstevel@tonic-gate if (error) { 3307c478bd9Sstevel@tonic-gate if (ulp) 3317c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 3327c478bd9Sstevel@tonic-gate goto out; 3337c478bd9Sstevel@tonic-gate } 3347c478bd9Sstevel@tonic-gate 3357c478bd9Sstevel@tonic-gate if (ulp && (ioflag & FRSYNC) && (ioflag & (FSYNC | FDSYNC)) && 3367c478bd9Sstevel@tonic-gate TRANS_ISTRANS(ufsvfsp)) { 3377c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 3387c478bd9Sstevel@tonic-gate TRANS_BEGIN_SYNC(ufsvfsp, TOP_READ_SYNC, TOP_READ_SIZE, 3397c478bd9Sstevel@tonic-gate error); 3407c478bd9Sstevel@tonic-gate ASSERT(!error); 3417c478bd9Sstevel@tonic-gate TRANS_END_SYNC(ufsvfsp, error, TOP_READ_SYNC, 3427c478bd9Sstevel@tonic-gate TOP_READ_SIZE); 3437c478bd9Sstevel@tonic-gate rw_enter(&ip->i_rwlock, RW_READER); 3447c478bd9Sstevel@tonic-gate } 3457c478bd9Sstevel@tonic-gate } else { 3467c478bd9Sstevel@tonic-gate /* 3477c478bd9Sstevel@tonic-gate * Only transact reads to files opened for sync-read and 3487c478bd9Sstevel@tonic-gate * sync-write on a file system that is not write locked. 3497c478bd9Sstevel@tonic-gate * 3507c478bd9Sstevel@tonic-gate * The ``not write locked'' check prevents problems with 3517c478bd9Sstevel@tonic-gate * enabling/disabling logging on a busy file system. E.g., 3527c478bd9Sstevel@tonic-gate * logging exists at the beginning of the read but does not 3537c478bd9Sstevel@tonic-gate * at the end. 3547c478bd9Sstevel@tonic-gate * 3557c478bd9Sstevel@tonic-gate */ 3567c478bd9Sstevel@tonic-gate if (ulp && (ioflag & FRSYNC) && (ioflag & (FSYNC | FDSYNC)) && 3577c478bd9Sstevel@tonic-gate TRANS_ISTRANS(ufsvfsp)) { 3587c478bd9Sstevel@tonic-gate TRANS_BEGIN_SYNC(ufsvfsp, TOP_READ_SYNC, TOP_READ_SIZE, 3597c478bd9Sstevel@tonic-gate error); 3607c478bd9Sstevel@tonic-gate ASSERT(!error); 3617c478bd9Sstevel@tonic-gate intrans = 1; 3627c478bd9Sstevel@tonic-gate } 3637c478bd9Sstevel@tonic-gate 3647c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_READER); 3657c478bd9Sstevel@tonic-gate error = rdip(ip, uiop, ioflag, cr); 3667c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 3677c478bd9Sstevel@tonic-gate 3687c478bd9Sstevel@tonic-gate if (intrans) { 3697c478bd9Sstevel@tonic-gate TRANS_END_SYNC(ufsvfsp, error, TOP_READ_SYNC, 3707c478bd9Sstevel@tonic-gate TOP_READ_SIZE); 3717c478bd9Sstevel@tonic-gate } 3727c478bd9Sstevel@tonic-gate } 3737c478bd9Sstevel@tonic-gate 3747c478bd9Sstevel@tonic-gate if (ulp) { 3757c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 3767c478bd9Sstevel@tonic-gate } 3777c478bd9Sstevel@tonic-gate out: 3787c478bd9Sstevel@tonic-gate 3797c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_READ_END, 3807c478bd9Sstevel@tonic-gate "ufs_read_end:vp %p error %d", vp, error); 3817c478bd9Sstevel@tonic-gate return (error); 3827c478bd9Sstevel@tonic-gate } 3837c478bd9Sstevel@tonic-gate 3847c478bd9Sstevel@tonic-gate extern int ufs_HW; /* high water mark */ 3857c478bd9Sstevel@tonic-gate extern int ufs_LW; /* low water mark */ 3867c478bd9Sstevel@tonic-gate int ufs_WRITES = 1; /* XXX - enable/disable */ 3877c478bd9Sstevel@tonic-gate int ufs_throttles = 0; /* throttling count */ 3887c478bd9Sstevel@tonic-gate int ufs_allow_shared_writes = 1; /* directio shared writes */ 3897c478bd9Sstevel@tonic-gate 3907c478bd9Sstevel@tonic-gate static int 3917c478bd9Sstevel@tonic-gate ufs_check_rewrite(struct inode *ip, struct uio *uiop, int ioflag) 3927c478bd9Sstevel@tonic-gate { 3937c478bd9Sstevel@tonic-gate 3947c478bd9Sstevel@tonic-gate /* 3957c478bd9Sstevel@tonic-gate * Filter to determine if this request is suitable as a 3967c478bd9Sstevel@tonic-gate * concurrent rewrite. This write must not allocate blocks 3977c478bd9Sstevel@tonic-gate * by extending the file or filling in holes. No use trying 3987c478bd9Sstevel@tonic-gate * through FSYNC descriptors as the inode will be synchronously 3997c478bd9Sstevel@tonic-gate * updated after the write. The uio structure has not yet been 4007c478bd9Sstevel@tonic-gate * checked for sanity, so assume nothing. 4017c478bd9Sstevel@tonic-gate */ 4027c478bd9Sstevel@tonic-gate return (((ip->i_mode & IFMT) == IFREG) && !(ioflag & FAPPEND) && 4037c478bd9Sstevel@tonic-gate (uiop->uio_loffset >= (offset_t)0) && 4047c478bd9Sstevel@tonic-gate (uiop->uio_loffset < ip->i_size) && (uiop->uio_resid > 0) && 4057c478bd9Sstevel@tonic-gate ((ip->i_size - uiop->uio_loffset) >= uiop->uio_resid) && 4067c478bd9Sstevel@tonic-gate !(ioflag & FSYNC) && !bmap_has_holes(ip) && 4077c478bd9Sstevel@tonic-gate ufs_allow_shared_writes); 4087c478bd9Sstevel@tonic-gate } 4097c478bd9Sstevel@tonic-gate 4107c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 4117c478bd9Sstevel@tonic-gate static int 4127c478bd9Sstevel@tonic-gate ufs_write(struct vnode *vp, struct uio *uiop, int ioflag, cred_t *cr, 4137c478bd9Sstevel@tonic-gate caller_context_t *ct) 4147c478bd9Sstevel@tonic-gate { 4157c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 4167c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp; 4177c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 4187c478bd9Sstevel@tonic-gate int retry = 1; 4197c478bd9Sstevel@tonic-gate int error, resv, resid = 0; 4207c478bd9Sstevel@tonic-gate int directio_status; 4217c478bd9Sstevel@tonic-gate int exclusive; 4227c478bd9Sstevel@tonic-gate long start_resid = uiop->uio_resid; 4237c478bd9Sstevel@tonic-gate 4247c478bd9Sstevel@tonic-gate TRACE_3(TR_FAC_UFS, TR_UFS_WRITE_START, 4257c478bd9Sstevel@tonic-gate "ufs_write_start:vp %p uiop %p ioflag %x", 4267c478bd9Sstevel@tonic-gate vp, uiop, ioflag); 4277c478bd9Sstevel@tonic-gate 4287c478bd9Sstevel@tonic-gate ASSERT(RW_LOCK_HELD(&ip->i_rwlock)); 4297c478bd9Sstevel@tonic-gate 4307c478bd9Sstevel@tonic-gate retry_mandlock: 4317c478bd9Sstevel@tonic-gate /* 4327c478bd9Sstevel@tonic-gate * Mandatory locking needs to be done before ufs_lockfs_begin() 4337c478bd9Sstevel@tonic-gate * and TRANS_BEGIN_[A]SYNC() calls since mandatory locks can sleep. 4347c478bd9Sstevel@tonic-gate * Check for forced unmounts normally done in ufs_lockfs_begin(). 4357c478bd9Sstevel@tonic-gate */ 4367c478bd9Sstevel@tonic-gate if ((ufsvfsp = ip->i_ufsvfs) == NULL) { 4377c478bd9Sstevel@tonic-gate error = EIO; 4387c478bd9Sstevel@tonic-gate goto out; 4397c478bd9Sstevel@tonic-gate } 4407c478bd9Sstevel@tonic-gate if (MANDLOCK(vp, ip->i_mode)) { 4417c478bd9Sstevel@tonic-gate 4427c478bd9Sstevel@tonic-gate ASSERT(RW_WRITE_HELD(&ip->i_rwlock)); 4437c478bd9Sstevel@tonic-gate 4447c478bd9Sstevel@tonic-gate /* 4457c478bd9Sstevel@tonic-gate * ufs_getattr ends up being called by chklock 4467c478bd9Sstevel@tonic-gate */ 4477c478bd9Sstevel@tonic-gate error = chklock(vp, FWRITE, uiop->uio_loffset, 4487c478bd9Sstevel@tonic-gate uiop->uio_resid, uiop->uio_fmode, ct); 4497c478bd9Sstevel@tonic-gate if (error) 4507c478bd9Sstevel@tonic-gate goto out; 4517c478bd9Sstevel@tonic-gate } 4527c478bd9Sstevel@tonic-gate 4537c478bd9Sstevel@tonic-gate /* i_rwlock can change in chklock */ 4547c478bd9Sstevel@tonic-gate exclusive = rw_write_held(&ip->i_rwlock); 4557c478bd9Sstevel@tonic-gate 4567c478bd9Sstevel@tonic-gate /* 4577c478bd9Sstevel@tonic-gate * Check for fast-path special case of directio re-writes. 4587c478bd9Sstevel@tonic-gate */ 4597c478bd9Sstevel@tonic-gate if ((ip->i_flag & IDIRECTIO || ufsvfsp->vfs_forcedirectio) && 4607c478bd9Sstevel@tonic-gate !exclusive && ufs_check_rewrite(ip, uiop, ioflag)) { 4617c478bd9Sstevel@tonic-gate 4627c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_WRITE_MASK); 4637c478bd9Sstevel@tonic-gate if (error) 4647c478bd9Sstevel@tonic-gate goto out; 4657c478bd9Sstevel@tonic-gate 4667c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_READER); 4677c478bd9Sstevel@tonic-gate error = ufs_directio_write(ip, uiop, ioflag, 1, cr, 4687c478bd9Sstevel@tonic-gate &directio_status); 4697c478bd9Sstevel@tonic-gate if (directio_status == DIRECTIO_SUCCESS) { 4707c478bd9Sstevel@tonic-gate uint_t i_flag_save; 4717c478bd9Sstevel@tonic-gate 4727c478bd9Sstevel@tonic-gate if (start_resid != uiop->uio_resid) 4737c478bd9Sstevel@tonic-gate error = 0; 4747c478bd9Sstevel@tonic-gate /* 4757c478bd9Sstevel@tonic-gate * Special treatment of access times for re-writes. 4767c478bd9Sstevel@tonic-gate * If IMOD is not already set, then convert it 4777c478bd9Sstevel@tonic-gate * to IMODACC for this operation. This defers 4787c478bd9Sstevel@tonic-gate * entering a delta into the log until the inode 4797c478bd9Sstevel@tonic-gate * is flushed. This mimics what is done for read 4807c478bd9Sstevel@tonic-gate * operations and inode access time. 4817c478bd9Sstevel@tonic-gate */ 4827c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 4837c478bd9Sstevel@tonic-gate i_flag_save = ip->i_flag; 4847c478bd9Sstevel@tonic-gate ip->i_flag |= IUPD | ICHG; 4857c478bd9Sstevel@tonic-gate ip->i_seq++; 4867c478bd9Sstevel@tonic-gate ITIMES_NOLOCK(ip); 4877c478bd9Sstevel@tonic-gate if ((i_flag_save & IMOD) == 0) { 4887c478bd9Sstevel@tonic-gate ip->i_flag &= ~IMOD; 4897c478bd9Sstevel@tonic-gate ip->i_flag |= IMODACC; 4907c478bd9Sstevel@tonic-gate } 4917c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 4927c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 4937c478bd9Sstevel@tonic-gate if (ulp) 4947c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 4957c478bd9Sstevel@tonic-gate goto out; 4967c478bd9Sstevel@tonic-gate } 4977c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 4987c478bd9Sstevel@tonic-gate if (ulp) 4997c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 5007c478bd9Sstevel@tonic-gate } 5017c478bd9Sstevel@tonic-gate 5027c478bd9Sstevel@tonic-gate if (!exclusive && !rw_tryupgrade(&ip->i_rwlock)) { 5037c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 5047c478bd9Sstevel@tonic-gate rw_enter(&ip->i_rwlock, RW_WRITER); 5057c478bd9Sstevel@tonic-gate /* 5067c478bd9Sstevel@tonic-gate * Mandatory locking could have been enabled 5077c478bd9Sstevel@tonic-gate * after dropping the i_rwlock. 5087c478bd9Sstevel@tonic-gate */ 5097c478bd9Sstevel@tonic-gate if (MANDLOCK(vp, ip->i_mode)) 5107c478bd9Sstevel@tonic-gate goto retry_mandlock; 5117c478bd9Sstevel@tonic-gate } 5127c478bd9Sstevel@tonic-gate 5137c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_WRITE_MASK); 5147c478bd9Sstevel@tonic-gate if (error) 5157c478bd9Sstevel@tonic-gate goto out; 5167c478bd9Sstevel@tonic-gate 5177c478bd9Sstevel@tonic-gate /* 5187c478bd9Sstevel@tonic-gate * Amount of log space needed for this write 5197c478bd9Sstevel@tonic-gate */ 5207c478bd9Sstevel@tonic-gate TRANS_WRITE_RESV(ip, uiop, ulp, &resv, &resid); 5217c478bd9Sstevel@tonic-gate 5227c478bd9Sstevel@tonic-gate /* 5237c478bd9Sstevel@tonic-gate * Throttle writes. 5247c478bd9Sstevel@tonic-gate */ 5257c478bd9Sstevel@tonic-gate if (ufs_WRITES && (ip->i_writes > ufs_HW)) { 5267c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 5277c478bd9Sstevel@tonic-gate while (ip->i_writes > ufs_HW) { 5287c478bd9Sstevel@tonic-gate ufs_throttles++; 5297c478bd9Sstevel@tonic-gate cv_wait(&ip->i_wrcv, &ip->i_tlock); 5307c478bd9Sstevel@tonic-gate } 5317c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 5327c478bd9Sstevel@tonic-gate } 5337c478bd9Sstevel@tonic-gate 5347c478bd9Sstevel@tonic-gate /* 5357c478bd9Sstevel@tonic-gate * Enter Transaction 5367c478bd9Sstevel@tonic-gate */ 5377c478bd9Sstevel@tonic-gate if (ioflag & (FSYNC|FDSYNC)) { 5387c478bd9Sstevel@tonic-gate if (ulp) { 5397c478bd9Sstevel@tonic-gate int terr = 0; 5407c478bd9Sstevel@tonic-gate TRANS_BEGIN_SYNC(ufsvfsp, TOP_WRITE_SYNC, resv, terr); 5417c478bd9Sstevel@tonic-gate ASSERT(!terr); 5427c478bd9Sstevel@tonic-gate } 5437c478bd9Sstevel@tonic-gate } else { 5447c478bd9Sstevel@tonic-gate if (ulp) 5457c478bd9Sstevel@tonic-gate TRANS_BEGIN_ASYNC(ufsvfsp, TOP_WRITE, resv); 5467c478bd9Sstevel@tonic-gate } 5477c478bd9Sstevel@tonic-gate 5487c478bd9Sstevel@tonic-gate /* 5497c478bd9Sstevel@tonic-gate * Write the file 5507c478bd9Sstevel@tonic-gate */ 5517c478bd9Sstevel@tonic-gate rw_enter(&ufsvfsp->vfs_dqrwlock, RW_READER); 5527c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_WRITER); 5537c478bd9Sstevel@tonic-gate if ((ioflag & FAPPEND) != 0 && (ip->i_mode & IFMT) == IFREG) { 5547c478bd9Sstevel@tonic-gate /* 5557c478bd9Sstevel@tonic-gate * In append mode start at end of file. 5567c478bd9Sstevel@tonic-gate */ 5577c478bd9Sstevel@tonic-gate uiop->uio_loffset = ip->i_size; 5587c478bd9Sstevel@tonic-gate } 5597c478bd9Sstevel@tonic-gate 5607c478bd9Sstevel@tonic-gate /* 5617c478bd9Sstevel@tonic-gate * Mild optimisation, don't call ufs_trans_write() unless we have to 5627c478bd9Sstevel@tonic-gate * Also, suppress file system full messages if we will retry. 5637c478bd9Sstevel@tonic-gate */ 5647c478bd9Sstevel@tonic-gate if (retry) 5657c478bd9Sstevel@tonic-gate ip->i_flag |= IQUIET; 5667c478bd9Sstevel@tonic-gate if (resid) { 5677c478bd9Sstevel@tonic-gate TRANS_WRITE(ip, uiop, ioflag, error, ulp, cr, resv, resid); 5687c478bd9Sstevel@tonic-gate } else { 5697c478bd9Sstevel@tonic-gate error = wrip(ip, uiop, ioflag, cr); 5707c478bd9Sstevel@tonic-gate } 5717c478bd9Sstevel@tonic-gate ip->i_flag &= ~IQUIET; 5727c478bd9Sstevel@tonic-gate 5737c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 5747c478bd9Sstevel@tonic-gate rw_exit(&ufsvfsp->vfs_dqrwlock); 5757c478bd9Sstevel@tonic-gate 5767c478bd9Sstevel@tonic-gate /* 5777c478bd9Sstevel@tonic-gate * Leave Transaction 5787c478bd9Sstevel@tonic-gate */ 5797c478bd9Sstevel@tonic-gate if (ulp) { 5807c478bd9Sstevel@tonic-gate if (ioflag & (FSYNC|FDSYNC)) { 5817c478bd9Sstevel@tonic-gate int terr = 0; 5827c478bd9Sstevel@tonic-gate TRANS_END_SYNC(ufsvfsp, terr, TOP_WRITE_SYNC, resv); 5837c478bd9Sstevel@tonic-gate if (error == 0) 5847c478bd9Sstevel@tonic-gate error = terr; 5857c478bd9Sstevel@tonic-gate } else { 5867c478bd9Sstevel@tonic-gate TRANS_END_ASYNC(ufsvfsp, TOP_WRITE, resv); 5877c478bd9Sstevel@tonic-gate } 5887c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 5897c478bd9Sstevel@tonic-gate } 5907c478bd9Sstevel@tonic-gate out: 5917c478bd9Sstevel@tonic-gate if ((error == ENOSPC) && retry && TRANS_ISTRANS(ufsvfsp)) { 5927c478bd9Sstevel@tonic-gate /* 5937c478bd9Sstevel@tonic-gate * Any blocks tied up in pending deletes? 5947c478bd9Sstevel@tonic-gate */ 5957c478bd9Sstevel@tonic-gate ufs_delete_drain_wait(ufsvfsp, 1); 5967c478bd9Sstevel@tonic-gate retry = 0; 5977c478bd9Sstevel@tonic-gate goto retry_mandlock; 5987c478bd9Sstevel@tonic-gate } 5997c478bd9Sstevel@tonic-gate 6007c478bd9Sstevel@tonic-gate if (error == ENOSPC && (start_resid != uiop->uio_resid)) 6017c478bd9Sstevel@tonic-gate error = 0; 6027c478bd9Sstevel@tonic-gate 6037c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_WRITE_END, 6047c478bd9Sstevel@tonic-gate "ufs_write_end:vp %p error %d", vp, error); 6057c478bd9Sstevel@tonic-gate return (error); 6067c478bd9Sstevel@tonic-gate } 6077c478bd9Sstevel@tonic-gate 6087c478bd9Sstevel@tonic-gate /* 6097c478bd9Sstevel@tonic-gate * Don't cache write blocks to files with the sticky bit set. 6107c478bd9Sstevel@tonic-gate * Used to keep swap files from blowing the page cache on a server. 6117c478bd9Sstevel@tonic-gate */ 6127c478bd9Sstevel@tonic-gate int stickyhack = 1; 6137c478bd9Sstevel@tonic-gate 6147c478bd9Sstevel@tonic-gate /* 6157c478bd9Sstevel@tonic-gate * Free behind hacks. The pager is busted. 6167c478bd9Sstevel@tonic-gate * XXX - need to pass the information down to writedone() in a flag like B_SEQ 6177c478bd9Sstevel@tonic-gate * or B_FREE_IF_TIGHT_ON_MEMORY. 6187c478bd9Sstevel@tonic-gate */ 6197c478bd9Sstevel@tonic-gate int freebehind = 1; 6207c478bd9Sstevel@tonic-gate int smallfile = 32 * 1024; 6217c478bd9Sstevel@tonic-gate 6227c478bd9Sstevel@tonic-gate /* 6237c478bd9Sstevel@tonic-gate * While we should, in most cases, cache the pages for write, we 6247c478bd9Sstevel@tonic-gate * may also want to cache the pages for read as long as they are 6257c478bd9Sstevel@tonic-gate * frequently re-usable. 6267c478bd9Sstevel@tonic-gate * 6277c478bd9Sstevel@tonic-gate * If cache_read_ahead = 1, the pages for read will go to the tail 6287c478bd9Sstevel@tonic-gate * of the cache list when they are released, otherwise go to the head. 6297c478bd9Sstevel@tonic-gate */ 6307c478bd9Sstevel@tonic-gate int cache_read_ahead = 0; 6317c478bd9Sstevel@tonic-gate 6327c478bd9Sstevel@tonic-gate /* 6337c478bd9Sstevel@tonic-gate * wrip does the real work of write requests for ufs. 6347c478bd9Sstevel@tonic-gate */ 6357c478bd9Sstevel@tonic-gate int 6367c478bd9Sstevel@tonic-gate wrip(struct inode *ip, struct uio *uio, int ioflag, struct cred *cr) 6377c478bd9Sstevel@tonic-gate { 6387c478bd9Sstevel@tonic-gate rlim64_t limit = uio->uio_llimit; 6397c478bd9Sstevel@tonic-gate u_offset_t off; 6407c478bd9Sstevel@tonic-gate u_offset_t old_i_size; 6417c478bd9Sstevel@tonic-gate struct fs *fs; 6427c478bd9Sstevel@tonic-gate struct vnode *vp; 6437c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp; 6447c478bd9Sstevel@tonic-gate caddr_t base; 6457c478bd9Sstevel@tonic-gate long start_resid = uio->uio_resid; /* save starting resid */ 6467c478bd9Sstevel@tonic-gate long premove_resid; /* resid before uiomove() */ 6477c478bd9Sstevel@tonic-gate uint_t flags; 6487c478bd9Sstevel@tonic-gate int newpage; 6497c478bd9Sstevel@tonic-gate int iupdat_flag, directio_status; 6507c478bd9Sstevel@tonic-gate int n, on, mapon; 6517c478bd9Sstevel@tonic-gate int error, pagecreate; 6527c478bd9Sstevel@tonic-gate int do_dqrwlock; /* drop/reacquire vfs_dqrwlock */ 6537c478bd9Sstevel@tonic-gate int32_t iblocks; 6547c478bd9Sstevel@tonic-gate int new_iblocks; 6557c478bd9Sstevel@tonic-gate 6567c478bd9Sstevel@tonic-gate /* 6577c478bd9Sstevel@tonic-gate * ip->i_size is incremented before the uiomove 6587c478bd9Sstevel@tonic-gate * is done on a write. If the move fails (bad user 6597c478bd9Sstevel@tonic-gate * address) reset ip->i_size. 6607c478bd9Sstevel@tonic-gate * The better way would be to increment ip->i_size 6617c478bd9Sstevel@tonic-gate * only if the uiomove succeeds. 6627c478bd9Sstevel@tonic-gate */ 6637c478bd9Sstevel@tonic-gate int i_size_changed = 0; 6647c478bd9Sstevel@tonic-gate o_mode_t type; 6657c478bd9Sstevel@tonic-gate int i_seq_needed = 0; 6667c478bd9Sstevel@tonic-gate 6677c478bd9Sstevel@tonic-gate vp = ITOV(ip); 6687c478bd9Sstevel@tonic-gate 6697c478bd9Sstevel@tonic-gate /* 6707c478bd9Sstevel@tonic-gate * check for forced unmount - should not happen as 6717c478bd9Sstevel@tonic-gate * the request passed the lockfs checks. 6727c478bd9Sstevel@tonic-gate */ 6737c478bd9Sstevel@tonic-gate if ((ufsvfsp = ip->i_ufsvfs) == NULL) 6747c478bd9Sstevel@tonic-gate return (EIO); 6757c478bd9Sstevel@tonic-gate 6767c478bd9Sstevel@tonic-gate fs = ip->i_fs; 6777c478bd9Sstevel@tonic-gate 6787c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_RWIP_START, 6797c478bd9Sstevel@tonic-gate "ufs_wrip_start:vp %p", vp); 6807c478bd9Sstevel@tonic-gate 6817c478bd9Sstevel@tonic-gate ASSERT(RW_WRITE_HELD(&ip->i_contents)); 6827c478bd9Sstevel@tonic-gate 6837c478bd9Sstevel@tonic-gate /* check for valid filetype */ 6847c478bd9Sstevel@tonic-gate type = ip->i_mode & IFMT; 6857c478bd9Sstevel@tonic-gate if ((type != IFREG) && (type != IFDIR) && (type != IFATTRDIR) && 6867c478bd9Sstevel@tonic-gate (type != IFLNK) && (type != IFSHAD)) { 6877c478bd9Sstevel@tonic-gate return (EIO); 6887c478bd9Sstevel@tonic-gate } 6897c478bd9Sstevel@tonic-gate 6907c478bd9Sstevel@tonic-gate /* 6917c478bd9Sstevel@tonic-gate * the actual limit of UFS file size 6927c478bd9Sstevel@tonic-gate * is UFS_MAXOFFSET_T 6937c478bd9Sstevel@tonic-gate */ 6947c478bd9Sstevel@tonic-gate if (limit == RLIM64_INFINITY || limit > MAXOFFSET_T) 6957c478bd9Sstevel@tonic-gate limit = MAXOFFSET_T; 6967c478bd9Sstevel@tonic-gate 6977c478bd9Sstevel@tonic-gate if (uio->uio_loffset >= limit) { 6987c478bd9Sstevel@tonic-gate proc_t *p = ttoproc(curthread); 6997c478bd9Sstevel@tonic-gate 7007c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_RWIP_END, 7017c478bd9Sstevel@tonic-gate "ufs_wrip_end:vp %p error %d", vp, EINVAL); 7027c478bd9Sstevel@tonic-gate 7037c478bd9Sstevel@tonic-gate mutex_enter(&p->p_lock); 7047c478bd9Sstevel@tonic-gate (void) rctl_action(rctlproc_legacy[RLIMIT_FSIZE], p->p_rctls, 7057c478bd9Sstevel@tonic-gate p, RCA_UNSAFE_SIGINFO); 7067c478bd9Sstevel@tonic-gate mutex_exit(&p->p_lock); 7077c478bd9Sstevel@tonic-gate return (EFBIG); 7087c478bd9Sstevel@tonic-gate } 7097c478bd9Sstevel@tonic-gate 7107c478bd9Sstevel@tonic-gate /* 7117c478bd9Sstevel@tonic-gate * if largefiles are disallowed, the limit is 7127c478bd9Sstevel@tonic-gate * the pre-largefiles value of 2GB 7137c478bd9Sstevel@tonic-gate */ 7147c478bd9Sstevel@tonic-gate if (ufsvfsp->vfs_lfflags & UFS_LARGEFILES) 7157c478bd9Sstevel@tonic-gate limit = MIN(UFS_MAXOFFSET_T, limit); 7167c478bd9Sstevel@tonic-gate else 7177c478bd9Sstevel@tonic-gate limit = MIN(MAXOFF32_T, limit); 7187c478bd9Sstevel@tonic-gate 7197c478bd9Sstevel@tonic-gate if (uio->uio_loffset < (offset_t)0) { 7207c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_RWIP_END, 7217c478bd9Sstevel@tonic-gate "ufs_wrip_end:vp %p error %d", vp, EINVAL); 7227c478bd9Sstevel@tonic-gate return (EINVAL); 7237c478bd9Sstevel@tonic-gate } 7247c478bd9Sstevel@tonic-gate if (uio->uio_resid == 0) { 7257c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_RWIP_END, 7267c478bd9Sstevel@tonic-gate "ufs_wrip_end:vp %p error %d", vp, 0); 7277c478bd9Sstevel@tonic-gate return (0); 7287c478bd9Sstevel@tonic-gate } 7297c478bd9Sstevel@tonic-gate 7307c478bd9Sstevel@tonic-gate if (uio->uio_loffset >= limit) 7317c478bd9Sstevel@tonic-gate return (EFBIG); 7327c478bd9Sstevel@tonic-gate 7337c478bd9Sstevel@tonic-gate ip->i_flag |= INOACC; /* don't update ref time in getpage */ 7347c478bd9Sstevel@tonic-gate 7357c478bd9Sstevel@tonic-gate if (ioflag & (FSYNC|FDSYNC)) { 7367c478bd9Sstevel@tonic-gate ip->i_flag |= ISYNC; 7377c478bd9Sstevel@tonic-gate iupdat_flag = 1; 7387c478bd9Sstevel@tonic-gate } 7397c478bd9Sstevel@tonic-gate /* 7407c478bd9Sstevel@tonic-gate * Try to go direct 7417c478bd9Sstevel@tonic-gate */ 7427c478bd9Sstevel@tonic-gate if (ip->i_flag & IDIRECTIO || ufsvfsp->vfs_forcedirectio) { 7437c478bd9Sstevel@tonic-gate uio->uio_llimit = limit; 7447c478bd9Sstevel@tonic-gate error = ufs_directio_write(ip, uio, ioflag, 0, cr, 7457c478bd9Sstevel@tonic-gate &directio_status); 7467c478bd9Sstevel@tonic-gate /* 7477c478bd9Sstevel@tonic-gate * If ufs_directio wrote to the file or set the flags, 7487c478bd9Sstevel@tonic-gate * we need to update i_seq, but it may be deferred. 7497c478bd9Sstevel@tonic-gate */ 7507c478bd9Sstevel@tonic-gate if (start_resid != uio->uio_resid || 7517c478bd9Sstevel@tonic-gate (ip->i_flag & (ICHG|IUPD))) { 7527c478bd9Sstevel@tonic-gate i_seq_needed = 1; 7537c478bd9Sstevel@tonic-gate ip->i_flag |= ISEQ; 7547c478bd9Sstevel@tonic-gate } 7557c478bd9Sstevel@tonic-gate if (directio_status == DIRECTIO_SUCCESS) 7567c478bd9Sstevel@tonic-gate goto out; 7577c478bd9Sstevel@tonic-gate } 7587c478bd9Sstevel@tonic-gate 7597c478bd9Sstevel@tonic-gate /* 7607c478bd9Sstevel@tonic-gate * Behavior with respect to dropping/reacquiring vfs_dqrwlock: 7617c478bd9Sstevel@tonic-gate * 7627c478bd9Sstevel@tonic-gate * o shadow inodes: vfs_dqrwlock is not held at all 7637c478bd9Sstevel@tonic-gate * o quota updates: vfs_dqrwlock is read or write held 7647c478bd9Sstevel@tonic-gate * o other updates: vfs_dqrwlock is read held 7657c478bd9Sstevel@tonic-gate * 7667c478bd9Sstevel@tonic-gate * The first case is the only one where we do not hold 7677c478bd9Sstevel@tonic-gate * vfs_dqrwlock at all while entering wrip(). 7687c478bd9Sstevel@tonic-gate * We must make sure not to downgrade/drop vfs_dqrwlock if we 7697c478bd9Sstevel@tonic-gate * have it as writer, i.e. if we are updating the quota inode. 7707c478bd9Sstevel@tonic-gate * There is no potential deadlock scenario in this case as 7717c478bd9Sstevel@tonic-gate * ufs_getpage() takes care of this and avoids reacquiring 7727c478bd9Sstevel@tonic-gate * vfs_dqrwlock in that case. 7737c478bd9Sstevel@tonic-gate * 7747c478bd9Sstevel@tonic-gate * This check is done here since the above conditions do not change 7757c478bd9Sstevel@tonic-gate * and we possibly loop below, so save a few cycles. 7767c478bd9Sstevel@tonic-gate */ 7777c478bd9Sstevel@tonic-gate if ((type == IFSHAD) || 7787c478bd9Sstevel@tonic-gate (rw_owner(&ufsvfsp->vfs_dqrwlock) == curthread)) { 7797c478bd9Sstevel@tonic-gate do_dqrwlock = 0; 7807c478bd9Sstevel@tonic-gate } else { 7817c478bd9Sstevel@tonic-gate do_dqrwlock = 1; 7827c478bd9Sstevel@tonic-gate } 7837c478bd9Sstevel@tonic-gate 7847c478bd9Sstevel@tonic-gate /* 7857c478bd9Sstevel@tonic-gate * Large Files: We cast MAXBMASK to offset_t 7867c478bd9Sstevel@tonic-gate * inorder to mask out the higher bits. Since offset_t 7877c478bd9Sstevel@tonic-gate * is a signed value, the high order bit set in MAXBMASK 7887c478bd9Sstevel@tonic-gate * value makes it do the right thing by having all bits 1 7897c478bd9Sstevel@tonic-gate * in the higher word. May be removed for _SOLARIS64_. 7907c478bd9Sstevel@tonic-gate */ 7917c478bd9Sstevel@tonic-gate 7927c478bd9Sstevel@tonic-gate fs = ip->i_fs; 7937c478bd9Sstevel@tonic-gate do { 7947c478bd9Sstevel@tonic-gate u_offset_t uoff = uio->uio_loffset; 7957c478bd9Sstevel@tonic-gate off = uoff & (offset_t)MAXBMASK; 7967c478bd9Sstevel@tonic-gate mapon = (int)(uoff & (offset_t)MAXBOFFSET); 7977c478bd9Sstevel@tonic-gate on = (int)blkoff(fs, uoff); 7987c478bd9Sstevel@tonic-gate n = (int)MIN(fs->fs_bsize - on, uio->uio_resid); 7997c478bd9Sstevel@tonic-gate new_iblocks = 1; 8007c478bd9Sstevel@tonic-gate 8017c478bd9Sstevel@tonic-gate if (type == IFREG && uoff + n >= limit) { 8027c478bd9Sstevel@tonic-gate if (uoff >= limit) { 8037c478bd9Sstevel@tonic-gate error = EFBIG; 8047c478bd9Sstevel@tonic-gate goto out; 8057c478bd9Sstevel@tonic-gate } 8067c478bd9Sstevel@tonic-gate /* 8077c478bd9Sstevel@tonic-gate * since uoff + n >= limit, 8087c478bd9Sstevel@tonic-gate * therefore n >= limit - uoff, and n is an int 8097c478bd9Sstevel@tonic-gate * so it is safe to cast it to an int 8107c478bd9Sstevel@tonic-gate */ 8117c478bd9Sstevel@tonic-gate n = (int)(limit - (rlim64_t)uoff); 8127c478bd9Sstevel@tonic-gate } 8137c478bd9Sstevel@tonic-gate if (uoff + n > ip->i_size) { 8147c478bd9Sstevel@tonic-gate /* 8157c478bd9Sstevel@tonic-gate * We are extending the length of the file. 8167c478bd9Sstevel@tonic-gate * bmap is used so that we are sure that 8177c478bd9Sstevel@tonic-gate * if we need to allocate new blocks, that it 8187c478bd9Sstevel@tonic-gate * is done here before we up the file size. 8197c478bd9Sstevel@tonic-gate */ 8207c478bd9Sstevel@tonic-gate error = bmap_write(ip, uoff, (int)(on + n), 8217c478bd9Sstevel@tonic-gate mapon == 0, cr); 8227c478bd9Sstevel@tonic-gate /* 8237c478bd9Sstevel@tonic-gate * bmap_write never drops i_contents so if 8247c478bd9Sstevel@tonic-gate * the flags are set it changed the file. 8257c478bd9Sstevel@tonic-gate */ 8267c478bd9Sstevel@tonic-gate if (ip->i_flag & (ICHG|IUPD)) { 8277c478bd9Sstevel@tonic-gate i_seq_needed = 1; 8287c478bd9Sstevel@tonic-gate ip->i_flag |= ISEQ; 8297c478bd9Sstevel@tonic-gate } 8307c478bd9Sstevel@tonic-gate if (error) 8317c478bd9Sstevel@tonic-gate break; 8327c478bd9Sstevel@tonic-gate /* 8337c478bd9Sstevel@tonic-gate * There is a window of vulnerability here. 8347c478bd9Sstevel@tonic-gate * The sequence of operations: allocate file 8357c478bd9Sstevel@tonic-gate * system blocks, uiomove the data into pages, 8367c478bd9Sstevel@tonic-gate * and then update the size of the file in the 8377c478bd9Sstevel@tonic-gate * inode, must happen atomically. However, due 8387c478bd9Sstevel@tonic-gate * to current locking constraints, this can not 8397c478bd9Sstevel@tonic-gate * be done. 8407c478bd9Sstevel@tonic-gate */ 8417c478bd9Sstevel@tonic-gate ASSERT(ip->i_writer == NULL); 8427c478bd9Sstevel@tonic-gate ip->i_writer = curthread; 8437c478bd9Sstevel@tonic-gate i_size_changed = 1; 8447c478bd9Sstevel@tonic-gate /* 8457c478bd9Sstevel@tonic-gate * If we are writing from the beginning of 8467c478bd9Sstevel@tonic-gate * the mapping, we can just create the 8477c478bd9Sstevel@tonic-gate * pages without having to read them. 8487c478bd9Sstevel@tonic-gate */ 8497c478bd9Sstevel@tonic-gate pagecreate = (mapon == 0); 8507c478bd9Sstevel@tonic-gate } else if (n == MAXBSIZE) { 8517c478bd9Sstevel@tonic-gate /* 8527c478bd9Sstevel@tonic-gate * Going to do a whole mappings worth, 8537c478bd9Sstevel@tonic-gate * so we can just create the pages w/o 8547c478bd9Sstevel@tonic-gate * having to read them in. But before 8557c478bd9Sstevel@tonic-gate * we do that, we need to make sure any 8567c478bd9Sstevel@tonic-gate * needed blocks are allocated first. 8577c478bd9Sstevel@tonic-gate */ 8587c478bd9Sstevel@tonic-gate iblocks = ip->i_blocks; 8597c478bd9Sstevel@tonic-gate error = bmap_write(ip, uoff, (int)(on + n), 1, cr); 8607c478bd9Sstevel@tonic-gate /* 8617c478bd9Sstevel@tonic-gate * bmap_write never drops i_contents so if 8627c478bd9Sstevel@tonic-gate * the flags are set it changed the file. 8637c478bd9Sstevel@tonic-gate */ 8647c478bd9Sstevel@tonic-gate if (ip->i_flag & (ICHG|IUPD)) { 8657c478bd9Sstevel@tonic-gate i_seq_needed = 1; 8667c478bd9Sstevel@tonic-gate ip->i_flag |= ISEQ; 8677c478bd9Sstevel@tonic-gate } 8687c478bd9Sstevel@tonic-gate if (error) 8697c478bd9Sstevel@tonic-gate break; 8707c478bd9Sstevel@tonic-gate pagecreate = 1; 8717c478bd9Sstevel@tonic-gate /* 8727c478bd9Sstevel@tonic-gate * check if the new created page needed the 8737c478bd9Sstevel@tonic-gate * allocation of new disk blocks. 8747c478bd9Sstevel@tonic-gate */ 8757c478bd9Sstevel@tonic-gate if (iblocks == ip->i_blocks) 8767c478bd9Sstevel@tonic-gate new_iblocks = 0; /* no new blocks allocated */ 8777c478bd9Sstevel@tonic-gate } else { 8787c478bd9Sstevel@tonic-gate pagecreate = 0; 8797c478bd9Sstevel@tonic-gate /* 8807c478bd9Sstevel@tonic-gate * In sync mode flush the indirect blocks which 8817c478bd9Sstevel@tonic-gate * may have been allocated and not written on 8827c478bd9Sstevel@tonic-gate * disk. In above cases bmap_write will allocate 8837c478bd9Sstevel@tonic-gate * in sync mode. 8847c478bd9Sstevel@tonic-gate */ 8857c478bd9Sstevel@tonic-gate if (ioflag & (FSYNC|FDSYNC)) { 8867c478bd9Sstevel@tonic-gate error = ufs_indirblk_sync(ip, uoff); 8877c478bd9Sstevel@tonic-gate if (error) 8887c478bd9Sstevel@tonic-gate break; 8897c478bd9Sstevel@tonic-gate } 8907c478bd9Sstevel@tonic-gate } 8917c478bd9Sstevel@tonic-gate 8927c478bd9Sstevel@tonic-gate /* 8937c478bd9Sstevel@tonic-gate * At this point we can enter ufs_getpage() in one 8947c478bd9Sstevel@tonic-gate * of two ways: 8957c478bd9Sstevel@tonic-gate * 1) segmap_getmapflt() calls ufs_getpage() when the 8967c478bd9Sstevel@tonic-gate * forcefault parameter is true (pagecreate == 0) 8977c478bd9Sstevel@tonic-gate * 2) uiomove() causes a page fault. 8987c478bd9Sstevel@tonic-gate * 8997c478bd9Sstevel@tonic-gate * We have to drop the contents lock to prevent the VM 9007c478bd9Sstevel@tonic-gate * system from trying to reaquire it in ufs_getpage() 9017c478bd9Sstevel@tonic-gate * should the uiomove cause a pagefault. 9027c478bd9Sstevel@tonic-gate * 9037c478bd9Sstevel@tonic-gate * We have to drop the reader vfs_dqrwlock here as well. 9047c478bd9Sstevel@tonic-gate */ 9057c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 9067c478bd9Sstevel@tonic-gate if (do_dqrwlock) { 9077c478bd9Sstevel@tonic-gate ASSERT(RW_LOCK_HELD(&ufsvfsp->vfs_dqrwlock)); 9087c478bd9Sstevel@tonic-gate ASSERT(!(RW_WRITE_HELD(&ufsvfsp->vfs_dqrwlock))); 9097c478bd9Sstevel@tonic-gate rw_exit(&ufsvfsp->vfs_dqrwlock); 9107c478bd9Sstevel@tonic-gate } 9117c478bd9Sstevel@tonic-gate 9127c478bd9Sstevel@tonic-gate base = segmap_getmapflt(segkmap, vp, (off + mapon), 9137c478bd9Sstevel@tonic-gate (uint_t)n, !pagecreate, S_WRITE); 9147c478bd9Sstevel@tonic-gate 9157c478bd9Sstevel@tonic-gate /* 9167c478bd9Sstevel@tonic-gate * segmap_pagecreate() returns 1 if it calls 9177c478bd9Sstevel@tonic-gate * page_create_va() to allocate any pages. 9187c478bd9Sstevel@tonic-gate */ 9197c478bd9Sstevel@tonic-gate newpage = 0; 9207c478bd9Sstevel@tonic-gate 9217c478bd9Sstevel@tonic-gate if (pagecreate) 9227c478bd9Sstevel@tonic-gate newpage = segmap_pagecreate(segkmap, base, 9237c478bd9Sstevel@tonic-gate (size_t)n, 0); 9247c478bd9Sstevel@tonic-gate 9257c478bd9Sstevel@tonic-gate premove_resid = uio->uio_resid; 9267c478bd9Sstevel@tonic-gate error = uiomove(base + mapon, (long)n, UIO_WRITE, uio); 9277c478bd9Sstevel@tonic-gate 9287c478bd9Sstevel@tonic-gate /* 9297c478bd9Sstevel@tonic-gate * If "newpage" is set, then a new page was created and it 9307c478bd9Sstevel@tonic-gate * does not contain valid data, so it needs to be initialized 9317c478bd9Sstevel@tonic-gate * at this point. 9327c478bd9Sstevel@tonic-gate * Otherwise the page contains old data, which was overwritten 9337c478bd9Sstevel@tonic-gate * partially or as a whole in uiomove. 9347c478bd9Sstevel@tonic-gate * If there is only one iovec structure within uio, then 9357c478bd9Sstevel@tonic-gate * on error uiomove will not be able to update uio->uio_loffset 9367c478bd9Sstevel@tonic-gate * and we would zero the whole page here! 9377c478bd9Sstevel@tonic-gate * 9387c478bd9Sstevel@tonic-gate * If uiomove fails because of an error, the old valid data 9397c478bd9Sstevel@tonic-gate * is kept instead of filling the rest of the page with zero's. 9407c478bd9Sstevel@tonic-gate */ 9417c478bd9Sstevel@tonic-gate if (newpage && 9427c478bd9Sstevel@tonic-gate uio->uio_loffset < roundup(off + mapon + n, PAGESIZE)) { 9437c478bd9Sstevel@tonic-gate /* 9447c478bd9Sstevel@tonic-gate * We created pages w/o initializing them completely, 9457c478bd9Sstevel@tonic-gate * thus we need to zero the part that wasn't set up. 9467c478bd9Sstevel@tonic-gate * This happens on most EOF write cases and if 9477c478bd9Sstevel@tonic-gate * we had some sort of error during the uiomove. 9487c478bd9Sstevel@tonic-gate */ 9497c478bd9Sstevel@tonic-gate int nzero, nmoved; 9507c478bd9Sstevel@tonic-gate 9517c478bd9Sstevel@tonic-gate nmoved = (int)(uio->uio_loffset - (off + mapon)); 9527c478bd9Sstevel@tonic-gate ASSERT(nmoved >= 0 && nmoved <= n); 9537c478bd9Sstevel@tonic-gate nzero = roundup(on + n, PAGESIZE) - nmoved; 9547c478bd9Sstevel@tonic-gate ASSERT(nzero > 0 && mapon + nmoved + nzero <= MAXBSIZE); 9557c478bd9Sstevel@tonic-gate (void) kzero(base + mapon + nmoved, (uint_t)nzero); 9567c478bd9Sstevel@tonic-gate } 9577c478bd9Sstevel@tonic-gate 9587c478bd9Sstevel@tonic-gate /* 9597c478bd9Sstevel@tonic-gate * Unlock the pages allocated by page_create_va() 9607c478bd9Sstevel@tonic-gate * in segmap_pagecreate() 9617c478bd9Sstevel@tonic-gate */ 9627c478bd9Sstevel@tonic-gate if (newpage) 9637c478bd9Sstevel@tonic-gate segmap_pageunlock(segkmap, base, (size_t)n, S_WRITE); 9647c478bd9Sstevel@tonic-gate 9657c478bd9Sstevel@tonic-gate /* 9667c478bd9Sstevel@tonic-gate * If the size of the file changed, then update the 9677c478bd9Sstevel@tonic-gate * size field in the inode now. This can't be done 9687c478bd9Sstevel@tonic-gate * before the call to segmap_pageunlock or there is 9697c478bd9Sstevel@tonic-gate * a potential deadlock with callers to ufs_putpage(). 9707c478bd9Sstevel@tonic-gate * They will be holding i_contents and trying to lock 9717c478bd9Sstevel@tonic-gate * a page, while this thread is holding a page locked 9727c478bd9Sstevel@tonic-gate * and trying to acquire i_contents. 9737c478bd9Sstevel@tonic-gate */ 9747c478bd9Sstevel@tonic-gate if (i_size_changed) { 9757c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_WRITER); 9767c478bd9Sstevel@tonic-gate old_i_size = ip->i_size; 9777c478bd9Sstevel@tonic-gate UFS_SET_ISIZE(uoff + n, ip); 9787c478bd9Sstevel@tonic-gate TRANS_INODE(ufsvfsp, ip); 9797c478bd9Sstevel@tonic-gate /* 9807c478bd9Sstevel@tonic-gate * file has grown larger than 2GB. Set flag 9817c478bd9Sstevel@tonic-gate * in superblock to indicate this, if it 9827c478bd9Sstevel@tonic-gate * is not already set. 9837c478bd9Sstevel@tonic-gate */ 9847c478bd9Sstevel@tonic-gate if ((ip->i_size > MAXOFF32_T) && 9857c478bd9Sstevel@tonic-gate !(fs->fs_flags & FSLARGEFILES)) { 9867c478bd9Sstevel@tonic-gate ASSERT(ufsvfsp->vfs_lfflags & UFS_LARGEFILES); 9877c478bd9Sstevel@tonic-gate mutex_enter(&ufsvfsp->vfs_lock); 9887c478bd9Sstevel@tonic-gate fs->fs_flags |= FSLARGEFILES; 9897c478bd9Sstevel@tonic-gate ufs_sbwrite(ufsvfsp); 9907c478bd9Sstevel@tonic-gate mutex_exit(&ufsvfsp->vfs_lock); 9917c478bd9Sstevel@tonic-gate } 9927c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 9937c478bd9Sstevel@tonic-gate ip->i_writer = NULL; 9947c478bd9Sstevel@tonic-gate cv_broadcast(&ip->i_wrcv); 9957c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 9967c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 9977c478bd9Sstevel@tonic-gate } 9987c478bd9Sstevel@tonic-gate 9997c478bd9Sstevel@tonic-gate if (error) { 10007c478bd9Sstevel@tonic-gate /* 10017c478bd9Sstevel@tonic-gate * If we failed on a write, we may have already 10027c478bd9Sstevel@tonic-gate * allocated file blocks as well as pages. It's 10037c478bd9Sstevel@tonic-gate * hard to undo the block allocation, but we must 10047c478bd9Sstevel@tonic-gate * be sure to invalidate any pages that may have 10057c478bd9Sstevel@tonic-gate * been allocated. 10067c478bd9Sstevel@tonic-gate * 10077c478bd9Sstevel@tonic-gate * If the page was created without initialization 10087c478bd9Sstevel@tonic-gate * then we must check if it should be possible 10097c478bd9Sstevel@tonic-gate * to destroy the new page and to keep the old data 10107c478bd9Sstevel@tonic-gate * on the disk. 10117c478bd9Sstevel@tonic-gate * 10127c478bd9Sstevel@tonic-gate * It is possible to destroy the page without 10137c478bd9Sstevel@tonic-gate * having to write back its contents only when 10147c478bd9Sstevel@tonic-gate * - the size of the file keeps unchanged 10157c478bd9Sstevel@tonic-gate * - bmap_write() did not allocate new disk blocks 10167c478bd9Sstevel@tonic-gate * it is possible to create big files using "seek" and 10177c478bd9Sstevel@tonic-gate * write to the end of the file. A "write" to a 10187c478bd9Sstevel@tonic-gate * position before the end of the file would not 10197c478bd9Sstevel@tonic-gate * change the size of the file but it would allocate 10207c478bd9Sstevel@tonic-gate * new disk blocks. 10217c478bd9Sstevel@tonic-gate * - uiomove intended to overwrite the whole page. 10227c478bd9Sstevel@tonic-gate * - a new page was created (newpage == 1). 10237c478bd9Sstevel@tonic-gate */ 10247c478bd9Sstevel@tonic-gate 10257c478bd9Sstevel@tonic-gate if (i_size_changed == 0 && new_iblocks == 0 && 10267c478bd9Sstevel@tonic-gate newpage) { 10277c478bd9Sstevel@tonic-gate 10287c478bd9Sstevel@tonic-gate /* unwind what uiomove eventually last did */ 10297c478bd9Sstevel@tonic-gate uio->uio_resid = premove_resid; 10307c478bd9Sstevel@tonic-gate 10317c478bd9Sstevel@tonic-gate /* 10327c478bd9Sstevel@tonic-gate * destroy the page, do not write ambiguous 10337c478bd9Sstevel@tonic-gate * data to the disk. 10347c478bd9Sstevel@tonic-gate */ 10357c478bd9Sstevel@tonic-gate flags = SM_DESTROY; 10367c478bd9Sstevel@tonic-gate } else { 10377c478bd9Sstevel@tonic-gate /* 10387c478bd9Sstevel@tonic-gate * write the page back to the disk, if dirty, 10397c478bd9Sstevel@tonic-gate * and remove the page from the cache. 10407c478bd9Sstevel@tonic-gate */ 10417c478bd9Sstevel@tonic-gate flags = SM_INVAL; 10427c478bd9Sstevel@tonic-gate } 10437c478bd9Sstevel@tonic-gate (void) segmap_release(segkmap, base, flags); 10447c478bd9Sstevel@tonic-gate } else { 10457c478bd9Sstevel@tonic-gate flags = 0; 10467c478bd9Sstevel@tonic-gate /* 10477c478bd9Sstevel@tonic-gate * Force write back for synchronous write cases. 10487c478bd9Sstevel@tonic-gate */ 10497c478bd9Sstevel@tonic-gate if ((ioflag & (FSYNC|FDSYNC)) || type == IFDIR) { 10507c478bd9Sstevel@tonic-gate /* 10517c478bd9Sstevel@tonic-gate * If the sticky bit is set but the 10527c478bd9Sstevel@tonic-gate * execute bit is not set, we do a 10537c478bd9Sstevel@tonic-gate * synchronous write back and free 10547c478bd9Sstevel@tonic-gate * the page when done. We set up swap 10557c478bd9Sstevel@tonic-gate * files to be handled this way to 10567c478bd9Sstevel@tonic-gate * prevent servers from keeping around 10577c478bd9Sstevel@tonic-gate * the client's swap pages too long. 10587c478bd9Sstevel@tonic-gate * XXX - there ought to be a better way. 10597c478bd9Sstevel@tonic-gate */ 10607c478bd9Sstevel@tonic-gate if (IS_SWAPVP(vp)) { 10617c478bd9Sstevel@tonic-gate flags = SM_WRITE | SM_FREE | 10627c478bd9Sstevel@tonic-gate SM_DONTNEED; 10637c478bd9Sstevel@tonic-gate iupdat_flag = 0; 10647c478bd9Sstevel@tonic-gate } else { 10657c478bd9Sstevel@tonic-gate flags = SM_WRITE; 10667c478bd9Sstevel@tonic-gate } 10677c478bd9Sstevel@tonic-gate } else if (n + on == MAXBSIZE || IS_SWAPVP(vp)) { 10687c478bd9Sstevel@tonic-gate /* 10697c478bd9Sstevel@tonic-gate * Have written a whole block. 10707c478bd9Sstevel@tonic-gate * Start an asynchronous write and 10717c478bd9Sstevel@tonic-gate * mark the buffer to indicate that 10727c478bd9Sstevel@tonic-gate * it won't be needed again soon. 10737c478bd9Sstevel@tonic-gate */ 10747c478bd9Sstevel@tonic-gate flags = SM_WRITE | SM_ASYNC | SM_DONTNEED; 10757c478bd9Sstevel@tonic-gate } 10767c478bd9Sstevel@tonic-gate error = segmap_release(segkmap, base, flags); 10777c478bd9Sstevel@tonic-gate /* 10787c478bd9Sstevel@tonic-gate * If the operation failed and is synchronous, 10797c478bd9Sstevel@tonic-gate * then we need to unwind what uiomove() last 10807c478bd9Sstevel@tonic-gate * did so we can potentially return an error to 10817c478bd9Sstevel@tonic-gate * the caller. If this write operation was 10827c478bd9Sstevel@tonic-gate * done in two pieces and the first succeeded, 10837c478bd9Sstevel@tonic-gate * then we won't return an error for the second 10847c478bd9Sstevel@tonic-gate * piece that failed. However, we only want to 10857c478bd9Sstevel@tonic-gate * return a resid value that reflects what was 10867c478bd9Sstevel@tonic-gate * really done. 10877c478bd9Sstevel@tonic-gate * 10887c478bd9Sstevel@tonic-gate * Failures for non-synchronous operations can 10897c478bd9Sstevel@tonic-gate * be ignored since the page subsystem will 10907c478bd9Sstevel@tonic-gate * retry the operation until it succeeds or the 10917c478bd9Sstevel@tonic-gate * file system is unmounted. 10927c478bd9Sstevel@tonic-gate */ 10937c478bd9Sstevel@tonic-gate if (error) { 10947c478bd9Sstevel@tonic-gate if ((ioflag & (FSYNC | FDSYNC)) || 10957c478bd9Sstevel@tonic-gate type == IFDIR) { 10967c478bd9Sstevel@tonic-gate uio->uio_resid = premove_resid; 10977c478bd9Sstevel@tonic-gate } else { 10987c478bd9Sstevel@tonic-gate error = 0; 10997c478bd9Sstevel@tonic-gate } 11007c478bd9Sstevel@tonic-gate } 11017c478bd9Sstevel@tonic-gate } 11027c478bd9Sstevel@tonic-gate 11037c478bd9Sstevel@tonic-gate /* 11047c478bd9Sstevel@tonic-gate * Re-acquire contents lock. 11057c478bd9Sstevel@tonic-gate * If it was dropped, reacquire reader vfs_dqrwlock as well. 11067c478bd9Sstevel@tonic-gate */ 11077c478bd9Sstevel@tonic-gate if (do_dqrwlock) 11087c478bd9Sstevel@tonic-gate rw_enter(&ufsvfsp->vfs_dqrwlock, RW_READER); 11097c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_WRITER); 11107c478bd9Sstevel@tonic-gate 11117c478bd9Sstevel@tonic-gate /* 11127c478bd9Sstevel@tonic-gate * If the uiomove() failed or if a synchronous 11137c478bd9Sstevel@tonic-gate * page push failed, fix up i_size. 11147c478bd9Sstevel@tonic-gate */ 11157c478bd9Sstevel@tonic-gate if (error) { 11167c478bd9Sstevel@tonic-gate if (i_size_changed) { 11177c478bd9Sstevel@tonic-gate /* 11187c478bd9Sstevel@tonic-gate * The uiomove failed, and we 11197c478bd9Sstevel@tonic-gate * allocated blocks,so get rid 11207c478bd9Sstevel@tonic-gate * of them. 11217c478bd9Sstevel@tonic-gate */ 11227c478bd9Sstevel@tonic-gate (void) ufs_itrunc(ip, old_i_size, 0, cr); 11237c478bd9Sstevel@tonic-gate } 11247c478bd9Sstevel@tonic-gate } else { 11257c478bd9Sstevel@tonic-gate /* 11267c478bd9Sstevel@tonic-gate * XXX - Can this be out of the loop? 11277c478bd9Sstevel@tonic-gate */ 11287c478bd9Sstevel@tonic-gate ip->i_flag |= IUPD | ICHG; 11297c478bd9Sstevel@tonic-gate /* 11307c478bd9Sstevel@tonic-gate * Only do one increase of i_seq for multiple 11317c478bd9Sstevel@tonic-gate * pieces. Because we drop locks, record 11327c478bd9Sstevel@tonic-gate * the fact that we changed the timestamp and 11337c478bd9Sstevel@tonic-gate * are deferring the increase in case another thread 11347c478bd9Sstevel@tonic-gate * pushes our timestamp update. 11357c478bd9Sstevel@tonic-gate */ 11367c478bd9Sstevel@tonic-gate i_seq_needed = 1; 11377c478bd9Sstevel@tonic-gate ip->i_flag |= ISEQ; 11387c478bd9Sstevel@tonic-gate if (i_size_changed) 11397c478bd9Sstevel@tonic-gate ip->i_flag |= IATTCHG; 11407c478bd9Sstevel@tonic-gate if ((ip->i_mode & (IEXEC | (IEXEC >> 3) | 11417c478bd9Sstevel@tonic-gate (IEXEC >> 6))) != 0 && 11427c478bd9Sstevel@tonic-gate (ip->i_mode & (ISUID | ISGID)) != 0 && 11437c478bd9Sstevel@tonic-gate secpolicy_vnode_setid_retain(cr, 11447c478bd9Sstevel@tonic-gate (ip->i_mode & ISUID) != 0 && ip->i_uid == 0) != 0) { 11457c478bd9Sstevel@tonic-gate /* 11467c478bd9Sstevel@tonic-gate * Clear Set-UID & Set-GID bits on 11477c478bd9Sstevel@tonic-gate * successful write if not privileged 11487c478bd9Sstevel@tonic-gate * and at least one of the execute bits 11497c478bd9Sstevel@tonic-gate * is set. If we always clear Set-GID, 11507c478bd9Sstevel@tonic-gate * mandatory file and record locking is 11517c478bd9Sstevel@tonic-gate * unuseable. 11527c478bd9Sstevel@tonic-gate */ 11537c478bd9Sstevel@tonic-gate ip->i_mode &= ~(ISUID | ISGID); 11547c478bd9Sstevel@tonic-gate } 11557c478bd9Sstevel@tonic-gate } 11567c478bd9Sstevel@tonic-gate TRANS_INODE(ufsvfsp, ip); 11577c478bd9Sstevel@tonic-gate } while (error == 0 && uio->uio_resid > 0 && n != 0); 11587c478bd9Sstevel@tonic-gate 11597c478bd9Sstevel@tonic-gate out: 11607c478bd9Sstevel@tonic-gate /* 11617c478bd9Sstevel@tonic-gate * Make sure i_seq is increased at least once per write 11627c478bd9Sstevel@tonic-gate */ 11637c478bd9Sstevel@tonic-gate if (i_seq_needed) { 11647c478bd9Sstevel@tonic-gate ip->i_seq++; 11657c478bd9Sstevel@tonic-gate ip->i_flag &= ~ISEQ; /* no longer deferred */ 11667c478bd9Sstevel@tonic-gate } 11677c478bd9Sstevel@tonic-gate 11687c478bd9Sstevel@tonic-gate /* 11697c478bd9Sstevel@tonic-gate * Inode is updated according to this table - 11707c478bd9Sstevel@tonic-gate * 11717c478bd9Sstevel@tonic-gate * FSYNC FDSYNC(posix.4) 11727c478bd9Sstevel@tonic-gate * -------------------------- 11737c478bd9Sstevel@tonic-gate * always@ IATTCHG|IBDWRITE 11747c478bd9Sstevel@tonic-gate * 11757c478bd9Sstevel@tonic-gate * @ - If we are doing synchronous write the only time we should 11767c478bd9Sstevel@tonic-gate * not be sync'ing the ip here is if we have the stickyhack 11777c478bd9Sstevel@tonic-gate * activated, the file is marked with the sticky bit and 11787c478bd9Sstevel@tonic-gate * no exec bit, the file length has not been changed and 11797c478bd9Sstevel@tonic-gate * no new blocks have been allocated during this write. 11807c478bd9Sstevel@tonic-gate */ 11817c478bd9Sstevel@tonic-gate 11827c478bd9Sstevel@tonic-gate if ((ip->i_flag & ISYNC) != 0) { 11837c478bd9Sstevel@tonic-gate /* 11847c478bd9Sstevel@tonic-gate * we have eliminated nosync 11857c478bd9Sstevel@tonic-gate */ 11867c478bd9Sstevel@tonic-gate if ((ip->i_flag & (IATTCHG|IBDWRITE)) || 11877c478bd9Sstevel@tonic-gate ((ioflag & FSYNC) && iupdat_flag)) { 11887c478bd9Sstevel@tonic-gate ufs_iupdat(ip, 1); 11897c478bd9Sstevel@tonic-gate } 11907c478bd9Sstevel@tonic-gate } 11917c478bd9Sstevel@tonic-gate 11927c478bd9Sstevel@tonic-gate /* 11937c478bd9Sstevel@tonic-gate * If we've already done a partial-write, terminate 11947c478bd9Sstevel@tonic-gate * the write but return no error unless the error is ENOSPC 11957c478bd9Sstevel@tonic-gate * because the caller can detect this and free resources and 11967c478bd9Sstevel@tonic-gate * try again. 11977c478bd9Sstevel@tonic-gate */ 11987c478bd9Sstevel@tonic-gate if ((start_resid != uio->uio_resid) && (error != ENOSPC)) 11997c478bd9Sstevel@tonic-gate error = 0; 12007c478bd9Sstevel@tonic-gate 12017c478bd9Sstevel@tonic-gate ip->i_flag &= ~(INOACC | ISYNC); 12027c478bd9Sstevel@tonic-gate ITIMES_NOLOCK(ip); 12037c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_RWIP_END, 12047c478bd9Sstevel@tonic-gate "ufs_wrip_end:vp %p error %d", vp, error); 12057c478bd9Sstevel@tonic-gate return (error); 12067c478bd9Sstevel@tonic-gate } 12077c478bd9Sstevel@tonic-gate 12087c478bd9Sstevel@tonic-gate /* 12097c478bd9Sstevel@tonic-gate * rdip does the real work of read requests for ufs. 12107c478bd9Sstevel@tonic-gate */ 12117c478bd9Sstevel@tonic-gate int 12127c478bd9Sstevel@tonic-gate rdip(struct inode *ip, struct uio *uio, int ioflag, cred_t *cr) 12137c478bd9Sstevel@tonic-gate { 12147c478bd9Sstevel@tonic-gate u_offset_t off; 12157c478bd9Sstevel@tonic-gate caddr_t base; 12167c478bd9Sstevel@tonic-gate struct fs *fs; 12177c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp; 12187c478bd9Sstevel@tonic-gate struct vnode *vp; 12197c478bd9Sstevel@tonic-gate long oresid = uio->uio_resid; 12207c478bd9Sstevel@tonic-gate u_offset_t n, on, mapon; 12217c478bd9Sstevel@tonic-gate int error = 0; 12227c478bd9Sstevel@tonic-gate int doupdate = 1; 12237c478bd9Sstevel@tonic-gate uint_t flags, cachemode; 12247c478bd9Sstevel@tonic-gate int dofree, directio_status; 12257c478bd9Sstevel@tonic-gate krw_t rwtype; 12267c478bd9Sstevel@tonic-gate o_mode_t type; 12277c478bd9Sstevel@tonic-gate 12287c478bd9Sstevel@tonic-gate vp = ITOV(ip); 12297c478bd9Sstevel@tonic-gate 12307c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_RWIP_START, 12317c478bd9Sstevel@tonic-gate "ufs_rdip_start:vp %p", vp); 12327c478bd9Sstevel@tonic-gate 12337c478bd9Sstevel@tonic-gate ASSERT(RW_LOCK_HELD(&ip->i_contents)); 12347c478bd9Sstevel@tonic-gate 12357c478bd9Sstevel@tonic-gate ufsvfsp = ip->i_ufsvfs; 12367c478bd9Sstevel@tonic-gate 12377c478bd9Sstevel@tonic-gate if (ufsvfsp == NULL) 12387c478bd9Sstevel@tonic-gate return (EIO); 12397c478bd9Sstevel@tonic-gate 12407c478bd9Sstevel@tonic-gate fs = ufsvfsp->vfs_fs; 12417c478bd9Sstevel@tonic-gate 12427c478bd9Sstevel@tonic-gate /* check for valid filetype */ 12437c478bd9Sstevel@tonic-gate type = ip->i_mode & IFMT; 12447c478bd9Sstevel@tonic-gate if ((type != IFREG) && (type != IFDIR) && (type != IFATTRDIR) && 12457c478bd9Sstevel@tonic-gate (type != IFLNK) && (type != IFSHAD)) { 12467c478bd9Sstevel@tonic-gate return (EIO); 12477c478bd9Sstevel@tonic-gate } 12487c478bd9Sstevel@tonic-gate 12497c478bd9Sstevel@tonic-gate if (uio->uio_loffset > UFS_MAXOFFSET_T) { 12507c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_RWIP_END, 12517c478bd9Sstevel@tonic-gate "ufs_rdip_end:vp %p error %d", vp, EINVAL); 12527c478bd9Sstevel@tonic-gate error = 0; 12537c478bd9Sstevel@tonic-gate goto out; 12547c478bd9Sstevel@tonic-gate } 12557c478bd9Sstevel@tonic-gate if (uio->uio_loffset < (offset_t)0) { 12567c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_RWIP_END, 12577c478bd9Sstevel@tonic-gate "ufs_rdip_end:vp %p error %d", vp, EINVAL); 12587c478bd9Sstevel@tonic-gate return (EINVAL); 12597c478bd9Sstevel@tonic-gate } 12607c478bd9Sstevel@tonic-gate if (uio->uio_resid == 0) { 12617c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_RWIP_END, 12627c478bd9Sstevel@tonic-gate "ufs_rdip_end:vp %p error %d", vp, 0); 12637c478bd9Sstevel@tonic-gate return (0); 12647c478bd9Sstevel@tonic-gate } 12657c478bd9Sstevel@tonic-gate 12667c478bd9Sstevel@tonic-gate if (!ULOCKFS_IS_NOIACC(ITOUL(ip)) && (fs->fs_ronly == 0) && 12677c478bd9Sstevel@tonic-gate (!ufsvfsp->vfs_noatime)) { 12687c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 12697c478bd9Sstevel@tonic-gate ip->i_flag |= IACC; 12707c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 12717c478bd9Sstevel@tonic-gate } 12727c478bd9Sstevel@tonic-gate /* 12737c478bd9Sstevel@tonic-gate * Try to go direct 12747c478bd9Sstevel@tonic-gate */ 12757c478bd9Sstevel@tonic-gate if (ip->i_flag & IDIRECTIO || ufsvfsp->vfs_forcedirectio) { 12767c478bd9Sstevel@tonic-gate error = ufs_directio_read(ip, uio, cr, &directio_status); 12777c478bd9Sstevel@tonic-gate if (directio_status == DIRECTIO_SUCCESS) 12787c478bd9Sstevel@tonic-gate goto out; 12797c478bd9Sstevel@tonic-gate } 12807c478bd9Sstevel@tonic-gate 12817c478bd9Sstevel@tonic-gate rwtype = (rw_write_held(&ip->i_contents)?RW_WRITER:RW_READER); 12827c478bd9Sstevel@tonic-gate 12837c478bd9Sstevel@tonic-gate /* 12847c478bd9Sstevel@tonic-gate * If cache_read_ahead is enabled, we will 12857c478bd9Sstevel@tonic-gate * release the pages at the tail of the cache 12867c478bd9Sstevel@tonic-gate * list, otherwise we will put them at the head. 12877c478bd9Sstevel@tonic-gate */ 12887c478bd9Sstevel@tonic-gate if (cache_read_ahead) 12897c478bd9Sstevel@tonic-gate cachemode = SM_FREE | SM_ASYNC; 12907c478bd9Sstevel@tonic-gate else 12917c478bd9Sstevel@tonic-gate cachemode = SM_FREE | SM_DONTNEED | SM_ASYNC; 12927c478bd9Sstevel@tonic-gate 12937c478bd9Sstevel@tonic-gate do { 12947c478bd9Sstevel@tonic-gate offset_t diff; 12957c478bd9Sstevel@tonic-gate u_offset_t uoff = uio->uio_loffset; 12967c478bd9Sstevel@tonic-gate off = uoff & (offset_t)MAXBMASK; 12977c478bd9Sstevel@tonic-gate mapon = (u_offset_t)(uoff & (offset_t)MAXBOFFSET); 12987c478bd9Sstevel@tonic-gate on = (u_offset_t)blkoff(fs, uoff); 12997c478bd9Sstevel@tonic-gate n = MIN((u_offset_t)fs->fs_bsize - on, 13007c478bd9Sstevel@tonic-gate (u_offset_t)uio->uio_resid); 13017c478bd9Sstevel@tonic-gate 13027c478bd9Sstevel@tonic-gate diff = ip->i_size - uoff; 13037c478bd9Sstevel@tonic-gate 13047c478bd9Sstevel@tonic-gate if (diff <= (offset_t)0) { 13057c478bd9Sstevel@tonic-gate error = 0; 13067c478bd9Sstevel@tonic-gate goto out; 13077c478bd9Sstevel@tonic-gate } 13087c478bd9Sstevel@tonic-gate if (diff < (offset_t)n) 13097c478bd9Sstevel@tonic-gate n = (int)diff; 13107c478bd9Sstevel@tonic-gate dofree = freebehind && 13117c478bd9Sstevel@tonic-gate ip->i_nextr == (off & PAGEMASK) && off > smallfile; 13127c478bd9Sstevel@tonic-gate 13137c478bd9Sstevel@tonic-gate /* 13147c478bd9Sstevel@tonic-gate * At this point we can enter ufs_getpage() in one of two 13157c478bd9Sstevel@tonic-gate * ways: 13167c478bd9Sstevel@tonic-gate * 1) segmap_getmapflt() calls ufs_getpage() when the 13177c478bd9Sstevel@tonic-gate * forcefault parameter is true (value of 1 is passed) 13187c478bd9Sstevel@tonic-gate * 2) uiomove() causes a page fault. 13197c478bd9Sstevel@tonic-gate * 13207c478bd9Sstevel@tonic-gate * We cannot hold onto an i_contents reader lock without 13217c478bd9Sstevel@tonic-gate * risking deadlock in ufs_getpage() so drop a reader lock. 13227c478bd9Sstevel@tonic-gate * The ufs_getpage() dolock logic already allows for a 13237c478bd9Sstevel@tonic-gate * thread holding i_contents as writer to work properly 13247c478bd9Sstevel@tonic-gate * so we keep a writer lock. 13257c478bd9Sstevel@tonic-gate */ 13267c478bd9Sstevel@tonic-gate if (rwtype == RW_READER) 13277c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 13287c478bd9Sstevel@tonic-gate base = segmap_getmapflt(segkmap, vp, (off + mapon), 13297c478bd9Sstevel@tonic-gate (uint_t)n, 1, S_READ); 13307c478bd9Sstevel@tonic-gate 13317c478bd9Sstevel@tonic-gate error = uiomove(base + mapon, (long)n, UIO_READ, uio); 13327c478bd9Sstevel@tonic-gate 13337c478bd9Sstevel@tonic-gate flags = 0; 13347c478bd9Sstevel@tonic-gate if (!error) { 13357c478bd9Sstevel@tonic-gate /* 13367c478bd9Sstevel@tonic-gate * If reading sequential we won't need 13377c478bd9Sstevel@tonic-gate * this buffer again soon. 13387c478bd9Sstevel@tonic-gate */ 13397c478bd9Sstevel@tonic-gate if (dofree) { 13407c478bd9Sstevel@tonic-gate flags = cachemode; 13417c478bd9Sstevel@tonic-gate } 13427c478bd9Sstevel@tonic-gate /* 13437c478bd9Sstevel@tonic-gate * In POSIX SYNC (FSYNC and FDSYNC) read mode, 13447c478bd9Sstevel@tonic-gate * we want to make sure that the page which has 13457c478bd9Sstevel@tonic-gate * been read, is written on disk if it is dirty. 13467c478bd9Sstevel@tonic-gate * And corresponding indirect blocks should also 13477c478bd9Sstevel@tonic-gate * be flushed out. 13487c478bd9Sstevel@tonic-gate */ 13497c478bd9Sstevel@tonic-gate if ((ioflag & FRSYNC) && (ioflag & (FSYNC|FDSYNC))) { 13507c478bd9Sstevel@tonic-gate flags &= ~SM_ASYNC; 13517c478bd9Sstevel@tonic-gate flags |= SM_WRITE; 13527c478bd9Sstevel@tonic-gate } 13537c478bd9Sstevel@tonic-gate error = segmap_release(segkmap, base, flags); 13547c478bd9Sstevel@tonic-gate } else 13557c478bd9Sstevel@tonic-gate (void) segmap_release(segkmap, base, flags); 13567c478bd9Sstevel@tonic-gate 13577c478bd9Sstevel@tonic-gate if (rwtype == RW_READER) 13587c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, rwtype); 13597c478bd9Sstevel@tonic-gate } while (error == 0 && uio->uio_resid > 0 && n != 0); 13607c478bd9Sstevel@tonic-gate out: 13617c478bd9Sstevel@tonic-gate /* 13627c478bd9Sstevel@tonic-gate * Inode is updated according to this table if FRSYNC is set. 13637c478bd9Sstevel@tonic-gate * 13647c478bd9Sstevel@tonic-gate * FSYNC FDSYNC(posix.4) 13657c478bd9Sstevel@tonic-gate * -------------------------- 13667c478bd9Sstevel@tonic-gate * always IATTCHG|IBDWRITE 13677c478bd9Sstevel@tonic-gate */ 13687c478bd9Sstevel@tonic-gate /* 13697c478bd9Sstevel@tonic-gate * The inode is not updated if we're logging and the inode is a 13707c478bd9Sstevel@tonic-gate * directory with FRSYNC, FSYNC and FDSYNC flags set. 13717c478bd9Sstevel@tonic-gate */ 13727c478bd9Sstevel@tonic-gate if (ioflag & FRSYNC) { 13737c478bd9Sstevel@tonic-gate if (TRANS_ISTRANS(ufsvfsp) && ((ip->i_mode & IFMT) == IFDIR)) { 13747c478bd9Sstevel@tonic-gate doupdate = 0; 13757c478bd9Sstevel@tonic-gate } 13767c478bd9Sstevel@tonic-gate if (doupdate) { 13777c478bd9Sstevel@tonic-gate if ((ioflag & FSYNC) || 13787c478bd9Sstevel@tonic-gate ((ioflag & FDSYNC) && 13797c478bd9Sstevel@tonic-gate (ip->i_flag & (IATTCHG|IBDWRITE)))) { 13807c478bd9Sstevel@tonic-gate ufs_iupdat(ip, 1); 13817c478bd9Sstevel@tonic-gate } 13827c478bd9Sstevel@tonic-gate } 13837c478bd9Sstevel@tonic-gate } 13847c478bd9Sstevel@tonic-gate /* 13857c478bd9Sstevel@tonic-gate * If we've already done a partial read, terminate 13867c478bd9Sstevel@tonic-gate * the read but return no error. 13877c478bd9Sstevel@tonic-gate */ 13887c478bd9Sstevel@tonic-gate if (oresid != uio->uio_resid) 13897c478bd9Sstevel@tonic-gate error = 0; 13907c478bd9Sstevel@tonic-gate ITIMES(ip); 13917c478bd9Sstevel@tonic-gate 13927c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_RWIP_END, 13937c478bd9Sstevel@tonic-gate "ufs_rdip_end:vp %p error %d", vp, error); 13947c478bd9Sstevel@tonic-gate return (error); 13957c478bd9Sstevel@tonic-gate } 13967c478bd9Sstevel@tonic-gate 13977c478bd9Sstevel@tonic-gate /* ARGSUSED */ 13987c478bd9Sstevel@tonic-gate static int 13997c478bd9Sstevel@tonic-gate ufs_ioctl( 14007c478bd9Sstevel@tonic-gate struct vnode *vp, 14017c478bd9Sstevel@tonic-gate int cmd, 14027c478bd9Sstevel@tonic-gate intptr_t arg, 14037c478bd9Sstevel@tonic-gate int flag, 14047c478bd9Sstevel@tonic-gate struct cred *cr, 14057c478bd9Sstevel@tonic-gate int *rvalp) 14067c478bd9Sstevel@tonic-gate { 14077c478bd9Sstevel@tonic-gate struct lockfs lockfs, lockfs_out; 14087c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp = VTOI(vp)->i_ufsvfs; 14097c478bd9Sstevel@tonic-gate char *comment, *original_comment; 14107c478bd9Sstevel@tonic-gate struct fs *fs; 14117c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 14127c478bd9Sstevel@tonic-gate offset_t off; 14137c478bd9Sstevel@tonic-gate extern int maxphys; 14147c478bd9Sstevel@tonic-gate int error; 14157c478bd9Sstevel@tonic-gate int issync; 14167c478bd9Sstevel@tonic-gate int trans_size; 14177c478bd9Sstevel@tonic-gate 14187c478bd9Sstevel@tonic-gate 14197c478bd9Sstevel@tonic-gate /* 14207c478bd9Sstevel@tonic-gate * forcibly unmounted 14217c478bd9Sstevel@tonic-gate */ 14227c478bd9Sstevel@tonic-gate if (ufsvfsp == NULL) { 14237c478bd9Sstevel@tonic-gate return (EIO); 14247c478bd9Sstevel@tonic-gate } 14257c478bd9Sstevel@tonic-gate 14267c478bd9Sstevel@tonic-gate fs = ufsvfsp->vfs_fs; 14277c478bd9Sstevel@tonic-gate 14287c478bd9Sstevel@tonic-gate if (cmd == Q_QUOTACTL) { 14297c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_QUOTA_MASK); 14307c478bd9Sstevel@tonic-gate if (error) 14317c478bd9Sstevel@tonic-gate return (error); 14327c478bd9Sstevel@tonic-gate 14337c478bd9Sstevel@tonic-gate if (ulp) { 14347c478bd9Sstevel@tonic-gate TRANS_BEGIN_ASYNC(ufsvfsp, TOP_QUOTA, 14357c478bd9Sstevel@tonic-gate TOP_SETQUOTA_SIZE(fs)); 14367c478bd9Sstevel@tonic-gate } 14377c478bd9Sstevel@tonic-gate 14387c478bd9Sstevel@tonic-gate error = quotactl(vp, arg, flag, cr); 14397c478bd9Sstevel@tonic-gate 14407c478bd9Sstevel@tonic-gate if (ulp) { 14417c478bd9Sstevel@tonic-gate TRANS_END_ASYNC(ufsvfsp, TOP_QUOTA, 14427c478bd9Sstevel@tonic-gate TOP_SETQUOTA_SIZE(fs)); 14437c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 14447c478bd9Sstevel@tonic-gate } 14457c478bd9Sstevel@tonic-gate return (error); 14467c478bd9Sstevel@tonic-gate } 14477c478bd9Sstevel@tonic-gate 14487c478bd9Sstevel@tonic-gate switch (cmd) { 14497c478bd9Sstevel@tonic-gate case _FIOLFS: 14507c478bd9Sstevel@tonic-gate /* 14517c478bd9Sstevel@tonic-gate * file system locking 14527c478bd9Sstevel@tonic-gate */ 14537c478bd9Sstevel@tonic-gate if (secpolicy_fs_config(cr, ufsvfsp->vfs_vfs) != 0) 14547c478bd9Sstevel@tonic-gate return (EPERM); 14557c478bd9Sstevel@tonic-gate 14567c478bd9Sstevel@tonic-gate if ((flag & DATAMODEL_MASK) == DATAMODEL_NATIVE) { 14577c478bd9Sstevel@tonic-gate if (copyin((caddr_t)arg, &lockfs, 14587c478bd9Sstevel@tonic-gate sizeof (struct lockfs))) 14597c478bd9Sstevel@tonic-gate return (EFAULT); 14607c478bd9Sstevel@tonic-gate } 14617c478bd9Sstevel@tonic-gate #ifdef _SYSCALL32_IMPL 14627c478bd9Sstevel@tonic-gate else { 14637c478bd9Sstevel@tonic-gate struct lockfs32 lockfs32; 14647c478bd9Sstevel@tonic-gate /* Translate ILP32 lockfs to LP64 lockfs */ 14657c478bd9Sstevel@tonic-gate if (copyin((caddr_t)arg, &lockfs32, 14667c478bd9Sstevel@tonic-gate sizeof (struct lockfs32))) 14677c478bd9Sstevel@tonic-gate return (EFAULT); 14687c478bd9Sstevel@tonic-gate lockfs.lf_lock = (ulong_t)lockfs32.lf_lock; 14697c478bd9Sstevel@tonic-gate lockfs.lf_flags = (ulong_t)lockfs32.lf_flags; 14707c478bd9Sstevel@tonic-gate lockfs.lf_key = (ulong_t)lockfs32.lf_key; 14717c478bd9Sstevel@tonic-gate lockfs.lf_comlen = (ulong_t)lockfs32.lf_comlen; 14727c478bd9Sstevel@tonic-gate lockfs.lf_comment = 14737c478bd9Sstevel@tonic-gate (caddr_t)(uintptr_t)lockfs32.lf_comment; 14747c478bd9Sstevel@tonic-gate } 14757c478bd9Sstevel@tonic-gate #endif /* _SYSCALL32_IMPL */ 14767c478bd9Sstevel@tonic-gate 14777c478bd9Sstevel@tonic-gate if (lockfs.lf_comlen) { 14787c478bd9Sstevel@tonic-gate if (lockfs.lf_comlen > LOCKFS_MAXCOMMENTLEN) 14797c478bd9Sstevel@tonic-gate return (ENAMETOOLONG); 14807c478bd9Sstevel@tonic-gate comment = kmem_alloc(lockfs.lf_comlen, 14817c478bd9Sstevel@tonic-gate KM_SLEEP); 14827c478bd9Sstevel@tonic-gate if (copyin(lockfs.lf_comment, comment, 14837c478bd9Sstevel@tonic-gate lockfs.lf_comlen)) { 14847c478bd9Sstevel@tonic-gate kmem_free(comment, lockfs.lf_comlen); 14857c478bd9Sstevel@tonic-gate return (EFAULT); 14867c478bd9Sstevel@tonic-gate } 14877c478bd9Sstevel@tonic-gate original_comment = lockfs.lf_comment; 14887c478bd9Sstevel@tonic-gate lockfs.lf_comment = comment; 14897c478bd9Sstevel@tonic-gate } 14907c478bd9Sstevel@tonic-gate if ((error = ufs_fiolfs(vp, &lockfs, 0)) == 0) { 14917c478bd9Sstevel@tonic-gate lockfs.lf_comment = original_comment; 14927c478bd9Sstevel@tonic-gate 14937c478bd9Sstevel@tonic-gate if ((flag & DATAMODEL_MASK) == 14947c478bd9Sstevel@tonic-gate DATAMODEL_NATIVE) { 14957c478bd9Sstevel@tonic-gate (void) copyout(&lockfs, (caddr_t)arg, 14967c478bd9Sstevel@tonic-gate sizeof (struct lockfs)); 14977c478bd9Sstevel@tonic-gate } 14987c478bd9Sstevel@tonic-gate #ifdef _SYSCALL32_IMPL 14997c478bd9Sstevel@tonic-gate else { 15007c478bd9Sstevel@tonic-gate struct lockfs32 lockfs32; 15017c478bd9Sstevel@tonic-gate /* Translate LP64 to ILP32 lockfs */ 15027c478bd9Sstevel@tonic-gate lockfs32.lf_lock = 15037c478bd9Sstevel@tonic-gate (uint32_t)lockfs.lf_lock; 15047c478bd9Sstevel@tonic-gate lockfs32.lf_flags = 15057c478bd9Sstevel@tonic-gate (uint32_t)lockfs.lf_flags; 15067c478bd9Sstevel@tonic-gate lockfs32.lf_key = 15077c478bd9Sstevel@tonic-gate (uint32_t)lockfs.lf_key; 15087c478bd9Sstevel@tonic-gate lockfs32.lf_comlen = 15097c478bd9Sstevel@tonic-gate (uint32_t)lockfs.lf_comlen; 15107c478bd9Sstevel@tonic-gate lockfs32.lf_comment = 15117c478bd9Sstevel@tonic-gate (uint32_t)(uintptr_t)lockfs.lf_comment; 15127c478bd9Sstevel@tonic-gate (void) copyout(&lockfs32, (caddr_t)arg, 15137c478bd9Sstevel@tonic-gate sizeof (struct lockfs32)); 15147c478bd9Sstevel@tonic-gate } 15157c478bd9Sstevel@tonic-gate #endif /* _SYSCALL32_IMPL */ 15167c478bd9Sstevel@tonic-gate 15177c478bd9Sstevel@tonic-gate } else { 15187c478bd9Sstevel@tonic-gate if (lockfs.lf_comlen) 15197c478bd9Sstevel@tonic-gate kmem_free(comment, lockfs.lf_comlen); 15207c478bd9Sstevel@tonic-gate } 15217c478bd9Sstevel@tonic-gate return (error); 15227c478bd9Sstevel@tonic-gate 15237c478bd9Sstevel@tonic-gate case _FIOLFSS: 15247c478bd9Sstevel@tonic-gate /* 15257c478bd9Sstevel@tonic-gate * get file system locking status 15267c478bd9Sstevel@tonic-gate */ 15277c478bd9Sstevel@tonic-gate 15287c478bd9Sstevel@tonic-gate if ((flag & DATAMODEL_MASK) == DATAMODEL_NATIVE) { 15297c478bd9Sstevel@tonic-gate if (copyin((caddr_t)arg, &lockfs, 15307c478bd9Sstevel@tonic-gate sizeof (struct lockfs))) 15317c478bd9Sstevel@tonic-gate return (EFAULT); 15327c478bd9Sstevel@tonic-gate } 15337c478bd9Sstevel@tonic-gate #ifdef _SYSCALL32_IMPL 15347c478bd9Sstevel@tonic-gate else { 15357c478bd9Sstevel@tonic-gate struct lockfs32 lockfs32; 15367c478bd9Sstevel@tonic-gate /* Translate ILP32 lockfs to LP64 lockfs */ 15377c478bd9Sstevel@tonic-gate if (copyin((caddr_t)arg, &lockfs32, 15387c478bd9Sstevel@tonic-gate sizeof (struct lockfs32))) 15397c478bd9Sstevel@tonic-gate return (EFAULT); 15407c478bd9Sstevel@tonic-gate lockfs.lf_lock = (ulong_t)lockfs32.lf_lock; 15417c478bd9Sstevel@tonic-gate lockfs.lf_flags = (ulong_t)lockfs32.lf_flags; 15427c478bd9Sstevel@tonic-gate lockfs.lf_key = (ulong_t)lockfs32.lf_key; 15437c478bd9Sstevel@tonic-gate lockfs.lf_comlen = (ulong_t)lockfs32.lf_comlen; 15447c478bd9Sstevel@tonic-gate lockfs.lf_comment = 15457c478bd9Sstevel@tonic-gate (caddr_t)(uintptr_t)lockfs32.lf_comment; 15467c478bd9Sstevel@tonic-gate } 15477c478bd9Sstevel@tonic-gate #endif /* _SYSCALL32_IMPL */ 15487c478bd9Sstevel@tonic-gate 15497c478bd9Sstevel@tonic-gate if (error = ufs_fiolfss(vp, &lockfs_out)) 15507c478bd9Sstevel@tonic-gate return (error); 15517c478bd9Sstevel@tonic-gate lockfs.lf_lock = lockfs_out.lf_lock; 15527c478bd9Sstevel@tonic-gate lockfs.lf_key = lockfs_out.lf_key; 15537c478bd9Sstevel@tonic-gate lockfs.lf_flags = lockfs_out.lf_flags; 15547c478bd9Sstevel@tonic-gate lockfs.lf_comlen = MIN(lockfs.lf_comlen, 15557c478bd9Sstevel@tonic-gate lockfs_out.lf_comlen); 15567c478bd9Sstevel@tonic-gate 15577c478bd9Sstevel@tonic-gate if ((flag & DATAMODEL_MASK) == DATAMODEL_NATIVE) { 15587c478bd9Sstevel@tonic-gate if (copyout(&lockfs, (caddr_t)arg, 15597c478bd9Sstevel@tonic-gate sizeof (struct lockfs))) 15607c478bd9Sstevel@tonic-gate return (EFAULT); 15617c478bd9Sstevel@tonic-gate } 15627c478bd9Sstevel@tonic-gate #ifdef _SYSCALL32_IMPL 15637c478bd9Sstevel@tonic-gate else { 15647c478bd9Sstevel@tonic-gate /* Translate LP64 to ILP32 lockfs */ 15657c478bd9Sstevel@tonic-gate struct lockfs32 lockfs32; 15667c478bd9Sstevel@tonic-gate lockfs32.lf_lock = (uint32_t)lockfs.lf_lock; 15677c478bd9Sstevel@tonic-gate lockfs32.lf_flags = (uint32_t)lockfs.lf_flags; 15687c478bd9Sstevel@tonic-gate lockfs32.lf_key = (uint32_t)lockfs.lf_key; 15697c478bd9Sstevel@tonic-gate lockfs32.lf_comlen = (uint32_t)lockfs.lf_comlen; 15707c478bd9Sstevel@tonic-gate lockfs32.lf_comment = 15717c478bd9Sstevel@tonic-gate (uint32_t)(uintptr_t)lockfs.lf_comment; 15727c478bd9Sstevel@tonic-gate if (copyout(&lockfs32, (caddr_t)arg, 15737c478bd9Sstevel@tonic-gate sizeof (struct lockfs32))) 15747c478bd9Sstevel@tonic-gate return (EFAULT); 15757c478bd9Sstevel@tonic-gate } 15767c478bd9Sstevel@tonic-gate #endif /* _SYSCALL32_IMPL */ 15777c478bd9Sstevel@tonic-gate 15787c478bd9Sstevel@tonic-gate if (lockfs.lf_comlen && 15797c478bd9Sstevel@tonic-gate lockfs.lf_comment && lockfs_out.lf_comment) 15807c478bd9Sstevel@tonic-gate if (copyout(lockfs_out.lf_comment, 15817c478bd9Sstevel@tonic-gate lockfs.lf_comment, 15827c478bd9Sstevel@tonic-gate lockfs.lf_comlen)) 15837c478bd9Sstevel@tonic-gate return (EFAULT); 15847c478bd9Sstevel@tonic-gate return (0); 15857c478bd9Sstevel@tonic-gate 15867c478bd9Sstevel@tonic-gate case _FIOSATIME: 15877c478bd9Sstevel@tonic-gate /* 15887c478bd9Sstevel@tonic-gate * set access time 15897c478bd9Sstevel@tonic-gate */ 15907c478bd9Sstevel@tonic-gate 15917c478bd9Sstevel@tonic-gate /* 15927c478bd9Sstevel@tonic-gate * if mounted w/o atime, return quietly. 15937c478bd9Sstevel@tonic-gate * I briefly thought about returning ENOSYS, but 15947c478bd9Sstevel@tonic-gate * figured that most apps would consider this fatal 15957c478bd9Sstevel@tonic-gate * but the idea is to make this as seamless as poss. 15967c478bd9Sstevel@tonic-gate */ 15977c478bd9Sstevel@tonic-gate if (ufsvfsp->vfs_noatime) 15987c478bd9Sstevel@tonic-gate return (0); 15997c478bd9Sstevel@tonic-gate 16007c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, 16017c478bd9Sstevel@tonic-gate ULOCKFS_SETATTR_MASK); 16027c478bd9Sstevel@tonic-gate if (error) 16037c478bd9Sstevel@tonic-gate return (error); 16047c478bd9Sstevel@tonic-gate 16057c478bd9Sstevel@tonic-gate if (ulp) { 16067c478bd9Sstevel@tonic-gate trans_size = (int)TOP_SETATTR_SIZE(VTOI(vp)); 16077c478bd9Sstevel@tonic-gate TRANS_BEGIN_CSYNC(ufsvfsp, issync, 16087c478bd9Sstevel@tonic-gate TOP_SETATTR, trans_size); 16097c478bd9Sstevel@tonic-gate } 16107c478bd9Sstevel@tonic-gate 16117c478bd9Sstevel@tonic-gate error = ufs_fiosatime(vp, (struct timeval *)arg, 16127c478bd9Sstevel@tonic-gate flag, cr); 16137c478bd9Sstevel@tonic-gate 16147c478bd9Sstevel@tonic-gate if (ulp) { 16157c478bd9Sstevel@tonic-gate TRANS_END_CSYNC(ufsvfsp, error, issync, 16167c478bd9Sstevel@tonic-gate TOP_SETATTR, trans_size); 16177c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 16187c478bd9Sstevel@tonic-gate } 16197c478bd9Sstevel@tonic-gate return (error); 16207c478bd9Sstevel@tonic-gate 16217c478bd9Sstevel@tonic-gate case _FIOSDIO: 16227c478bd9Sstevel@tonic-gate /* 16237c478bd9Sstevel@tonic-gate * set delayed-io 16247c478bd9Sstevel@tonic-gate */ 16257c478bd9Sstevel@tonic-gate return (ufs_fiosdio(vp, (uint_t *)arg, flag, cr)); 16267c478bd9Sstevel@tonic-gate 16277c478bd9Sstevel@tonic-gate case _FIOGDIO: 16287c478bd9Sstevel@tonic-gate /* 16297c478bd9Sstevel@tonic-gate * get delayed-io 16307c478bd9Sstevel@tonic-gate */ 16317c478bd9Sstevel@tonic-gate return (ufs_fiogdio(vp, (uint_t *)arg, flag, cr)); 16327c478bd9Sstevel@tonic-gate 16337c478bd9Sstevel@tonic-gate case _FIOIO: 16347c478bd9Sstevel@tonic-gate /* 16357c478bd9Sstevel@tonic-gate * inode open 16367c478bd9Sstevel@tonic-gate */ 16377c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, 16387c478bd9Sstevel@tonic-gate ULOCKFS_VGET_MASK); 16397c478bd9Sstevel@tonic-gate if (error) 16407c478bd9Sstevel@tonic-gate return (error); 16417c478bd9Sstevel@tonic-gate 16427c478bd9Sstevel@tonic-gate error = ufs_fioio(vp, (struct fioio *)arg, flag, cr); 16437c478bd9Sstevel@tonic-gate 16447c478bd9Sstevel@tonic-gate if (ulp) { 16457c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 16467c478bd9Sstevel@tonic-gate } 16477c478bd9Sstevel@tonic-gate return (error); 16487c478bd9Sstevel@tonic-gate 16497c478bd9Sstevel@tonic-gate case _FIOFFS: 16507c478bd9Sstevel@tonic-gate /* 16517c478bd9Sstevel@tonic-gate * file system flush (push w/invalidate) 16527c478bd9Sstevel@tonic-gate */ 16537c478bd9Sstevel@tonic-gate if ((caddr_t)arg != NULL) 16547c478bd9Sstevel@tonic-gate return (EINVAL); 16557c478bd9Sstevel@tonic-gate return (ufs_fioffs(vp, NULL, cr)); 16567c478bd9Sstevel@tonic-gate 16577c478bd9Sstevel@tonic-gate case _FIOISBUSY: 16587c478bd9Sstevel@tonic-gate /* 16597c478bd9Sstevel@tonic-gate * Contract-private interface for Legato 16607c478bd9Sstevel@tonic-gate * Purge this vnode from the DNLC and decide 16617c478bd9Sstevel@tonic-gate * if this vnode is busy (*arg == 1) or not 16627c478bd9Sstevel@tonic-gate * (*arg == 0) 16637c478bd9Sstevel@tonic-gate */ 16647c478bd9Sstevel@tonic-gate if (secpolicy_fs_config(cr, ufsvfsp->vfs_vfs) != 0) 16657c478bd9Sstevel@tonic-gate return (EPERM); 16667c478bd9Sstevel@tonic-gate error = ufs_fioisbusy(vp, (int *)arg, cr); 16677c478bd9Sstevel@tonic-gate return (error); 16687c478bd9Sstevel@tonic-gate 16697c478bd9Sstevel@tonic-gate case _FIODIRECTIO: 16707c478bd9Sstevel@tonic-gate return (ufs_fiodirectio(vp, (int)arg, cr)); 16717c478bd9Sstevel@tonic-gate 16727c478bd9Sstevel@tonic-gate case _FIOTUNE: 16737c478bd9Sstevel@tonic-gate /* 16747c478bd9Sstevel@tonic-gate * Tune the file system (aka setting fs attributes) 16757c478bd9Sstevel@tonic-gate */ 16767c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, 16777c478bd9Sstevel@tonic-gate ULOCKFS_SETATTR_MASK); 16787c478bd9Sstevel@tonic-gate if (error) 16797c478bd9Sstevel@tonic-gate return (error); 16807c478bd9Sstevel@tonic-gate 16817c478bd9Sstevel@tonic-gate error = ufs_fiotune(vp, (struct fiotune *)arg, cr); 16827c478bd9Sstevel@tonic-gate 16837c478bd9Sstevel@tonic-gate if (ulp) 16847c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 16857c478bd9Sstevel@tonic-gate return (error); 16867c478bd9Sstevel@tonic-gate 16877c478bd9Sstevel@tonic-gate case _FIOLOGENABLE: 16887c478bd9Sstevel@tonic-gate if (secpolicy_fs_config(cr, ufsvfsp->vfs_vfs) != 0) 16897c478bd9Sstevel@tonic-gate return (EPERM); 16907c478bd9Sstevel@tonic-gate return (ufs_fiologenable(vp, (void *)arg, cr, flag)); 16917c478bd9Sstevel@tonic-gate 16927c478bd9Sstevel@tonic-gate case _FIOLOGDISABLE: 16937c478bd9Sstevel@tonic-gate if (secpolicy_fs_config(cr, ufsvfsp->vfs_vfs) != 0) 16947c478bd9Sstevel@tonic-gate return (EPERM); 16957c478bd9Sstevel@tonic-gate return (ufs_fiologdisable(vp, (void *)arg, cr, flag)); 16967c478bd9Sstevel@tonic-gate 16977c478bd9Sstevel@tonic-gate case _FIOISLOG: 16987c478bd9Sstevel@tonic-gate return (ufs_fioislog(vp, (void *)arg, cr, flag)); 16997c478bd9Sstevel@tonic-gate 17007c478bd9Sstevel@tonic-gate case _FIOSNAPSHOTCREATE_MULTI: 17017c478bd9Sstevel@tonic-gate { 17027c478bd9Sstevel@tonic-gate struct fiosnapcreate_multi fc, *fcp; 17037c478bd9Sstevel@tonic-gate size_t fcm_size; 17047c478bd9Sstevel@tonic-gate 17057c478bd9Sstevel@tonic-gate if (copyin((void *)arg, &fc, sizeof (fc))) 17067c478bd9Sstevel@tonic-gate return (EFAULT); 17077c478bd9Sstevel@tonic-gate if (fc.backfilecount > MAX_BACKFILE_COUNT) 17087c478bd9Sstevel@tonic-gate return (EINVAL); 17097c478bd9Sstevel@tonic-gate fcm_size = sizeof (struct fiosnapcreate_multi) + 17107c478bd9Sstevel@tonic-gate (fc.backfilecount - 1) * sizeof (int); 17117c478bd9Sstevel@tonic-gate fcp = (struct fiosnapcreate_multi *) 17127c478bd9Sstevel@tonic-gate kmem_alloc(fcm_size, KM_SLEEP); 17137c478bd9Sstevel@tonic-gate if (copyin((void *)arg, fcp, fcm_size)) { 17147c478bd9Sstevel@tonic-gate kmem_free(fcp, fcm_size); 17157c478bd9Sstevel@tonic-gate return (EFAULT); 17167c478bd9Sstevel@tonic-gate } 17177c478bd9Sstevel@tonic-gate error = ufs_snap_create(vp, fcp, cr); 17187c478bd9Sstevel@tonic-gate if (!error && copyout(fcp, (void *)arg, fcm_size)) 17197c478bd9Sstevel@tonic-gate error = EFAULT; 17207c478bd9Sstevel@tonic-gate kmem_free(fcp, fcm_size); 17217c478bd9Sstevel@tonic-gate return (error); 17227c478bd9Sstevel@tonic-gate } 17237c478bd9Sstevel@tonic-gate 17247c478bd9Sstevel@tonic-gate case _FIOSNAPSHOTDELETE: 17257c478bd9Sstevel@tonic-gate { 17267c478bd9Sstevel@tonic-gate struct fiosnapdelete fc; 17277c478bd9Sstevel@tonic-gate 17287c478bd9Sstevel@tonic-gate if (copyin((void *)arg, &fc, sizeof (fc))) 17297c478bd9Sstevel@tonic-gate return (EFAULT); 17307c478bd9Sstevel@tonic-gate error = ufs_snap_delete(vp, &fc, cr); 17317c478bd9Sstevel@tonic-gate if (!error && copyout(&fc, (void *)arg, sizeof (fc))) 17327c478bd9Sstevel@tonic-gate error = EFAULT; 17337c478bd9Sstevel@tonic-gate return (error); 17347c478bd9Sstevel@tonic-gate } 17357c478bd9Sstevel@tonic-gate 17367c478bd9Sstevel@tonic-gate case _FIOGETSUPERBLOCK: 17377c478bd9Sstevel@tonic-gate if (copyout(fs, (void *)arg, SBSIZE)) 17387c478bd9Sstevel@tonic-gate return (EFAULT); 17397c478bd9Sstevel@tonic-gate return (0); 17407c478bd9Sstevel@tonic-gate 17417c478bd9Sstevel@tonic-gate case _FIOGETMAXPHYS: 17427c478bd9Sstevel@tonic-gate if (copyout(&maxphys, (void *)arg, sizeof (maxphys))) 17437c478bd9Sstevel@tonic-gate return (EFAULT); 17447c478bd9Sstevel@tonic-gate return (0); 17457c478bd9Sstevel@tonic-gate 17467c478bd9Sstevel@tonic-gate /* 17477c478bd9Sstevel@tonic-gate * The following 3 ioctls are for TSufs support 17487c478bd9Sstevel@tonic-gate * although could potentially be used elsewhere 17497c478bd9Sstevel@tonic-gate */ 17507c478bd9Sstevel@tonic-gate case _FIO_SET_LUFS_DEBUG: 17517c478bd9Sstevel@tonic-gate if (secpolicy_fs_config(cr, ufsvfsp->vfs_vfs) != 0) 17527c478bd9Sstevel@tonic-gate return (EPERM); 17537c478bd9Sstevel@tonic-gate lufs_debug = (uint32_t)arg; 17547c478bd9Sstevel@tonic-gate return (0); 17557c478bd9Sstevel@tonic-gate 17567c478bd9Sstevel@tonic-gate case _FIO_SET_LUFS_ERROR: 17577c478bd9Sstevel@tonic-gate if (secpolicy_fs_config(cr, ufsvfsp->vfs_vfs) != 0) 17587c478bd9Sstevel@tonic-gate return (EPERM); 17597c478bd9Sstevel@tonic-gate TRANS_SETERROR(ufsvfsp); 17607c478bd9Sstevel@tonic-gate return (0); 17617c478bd9Sstevel@tonic-gate 17627c478bd9Sstevel@tonic-gate case _FIO_GET_TOP_STATS: 17637c478bd9Sstevel@tonic-gate { 17647c478bd9Sstevel@tonic-gate fio_lufs_stats_t *ls; 17657c478bd9Sstevel@tonic-gate ml_unit_t *ul = ufsvfsp->vfs_log; 17667c478bd9Sstevel@tonic-gate 17677c478bd9Sstevel@tonic-gate ls = kmem_zalloc(sizeof (*ls), KM_SLEEP); 17687c478bd9Sstevel@tonic-gate ls->ls_debug = ul->un_debug; /* return debug value */ 17697c478bd9Sstevel@tonic-gate /* Copy stucture if statistics are being kept */ 17707c478bd9Sstevel@tonic-gate if (ul->un_logmap->mtm_tops) { 17717c478bd9Sstevel@tonic-gate ls->ls_topstats = *(ul->un_logmap->mtm_tops); 17727c478bd9Sstevel@tonic-gate } 17737c478bd9Sstevel@tonic-gate error = 0; 17747c478bd9Sstevel@tonic-gate if (copyout(ls, (void *)arg, sizeof (*ls))) 17757c478bd9Sstevel@tonic-gate error = EFAULT; 17767c478bd9Sstevel@tonic-gate kmem_free(ls, sizeof (*ls)); 17777c478bd9Sstevel@tonic-gate return (error); 17787c478bd9Sstevel@tonic-gate } 17797c478bd9Sstevel@tonic-gate 17807c478bd9Sstevel@tonic-gate case _FIO_SEEK_DATA: 17817c478bd9Sstevel@tonic-gate case _FIO_SEEK_HOLE: 17827c478bd9Sstevel@tonic-gate if (ddi_copyin((void *)arg, &off, sizeof (off), flag)) 17837c478bd9Sstevel@tonic-gate return (EFAULT); 17847c478bd9Sstevel@tonic-gate /* offset paramater is in/out */ 17857c478bd9Sstevel@tonic-gate error = ufs_fio_holey(vp, cmd, &off); 17867c478bd9Sstevel@tonic-gate if (error) 17877c478bd9Sstevel@tonic-gate return (error); 17887c478bd9Sstevel@tonic-gate if (ddi_copyout(&off, (void *)arg, sizeof (off), flag)) 17897c478bd9Sstevel@tonic-gate return (EFAULT); 17907c478bd9Sstevel@tonic-gate return (0); 17917c478bd9Sstevel@tonic-gate 17927c478bd9Sstevel@tonic-gate default: 17937c478bd9Sstevel@tonic-gate return (ENOTTY); 17947c478bd9Sstevel@tonic-gate } 17957c478bd9Sstevel@tonic-gate } 17967c478bd9Sstevel@tonic-gate 17977c478bd9Sstevel@tonic-gate /* ARGSUSED */ 17987c478bd9Sstevel@tonic-gate static int 17997c478bd9Sstevel@tonic-gate ufs_getattr(struct vnode *vp, struct vattr *vap, int flags, 18007c478bd9Sstevel@tonic-gate struct cred *cr) 18017c478bd9Sstevel@tonic-gate { 18027c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 18037c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp; 18047c478bd9Sstevel@tonic-gate int err; 18057c478bd9Sstevel@tonic-gate 18067c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_GETATTR_START, 18077c478bd9Sstevel@tonic-gate "ufs_getattr_start:vp %p flags %x", vp, flags); 18087c478bd9Sstevel@tonic-gate 18097c478bd9Sstevel@tonic-gate if (vap->va_mask == AT_SIZE) { 18107c478bd9Sstevel@tonic-gate /* 18117c478bd9Sstevel@tonic-gate * for performance, if only the size is requested don't bother 18127c478bd9Sstevel@tonic-gate * with anything else. 18137c478bd9Sstevel@tonic-gate */ 18147c478bd9Sstevel@tonic-gate UFS_GET_ISIZE(&vap->va_size, ip); 18157c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_GETATTR_END, 18167c478bd9Sstevel@tonic-gate "ufs_getattr_end:vp %p", vp); 18177c478bd9Sstevel@tonic-gate return (0); 18187c478bd9Sstevel@tonic-gate } 18197c478bd9Sstevel@tonic-gate 18207c478bd9Sstevel@tonic-gate /* 18217c478bd9Sstevel@tonic-gate * inlined lockfs checks 18227c478bd9Sstevel@tonic-gate */ 18237c478bd9Sstevel@tonic-gate ufsvfsp = ip->i_ufsvfs; 18247c478bd9Sstevel@tonic-gate if ((ufsvfsp == NULL) || ULOCKFS_IS_HLOCK(&ufsvfsp->vfs_ulockfs)) { 18257c478bd9Sstevel@tonic-gate err = EIO; 18267c478bd9Sstevel@tonic-gate goto out; 18277c478bd9Sstevel@tonic-gate } 18287c478bd9Sstevel@tonic-gate 18297c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_READER); 18307c478bd9Sstevel@tonic-gate /* 18317c478bd9Sstevel@tonic-gate * Return all the attributes. This should be refined so 18327c478bd9Sstevel@tonic-gate * that it only returns what's asked for. 18337c478bd9Sstevel@tonic-gate */ 18347c478bd9Sstevel@tonic-gate 18357c478bd9Sstevel@tonic-gate /* 18367c478bd9Sstevel@tonic-gate * Copy from inode table. 18377c478bd9Sstevel@tonic-gate */ 18387c478bd9Sstevel@tonic-gate vap->va_type = vp->v_type; 18397c478bd9Sstevel@tonic-gate vap->va_mode = ip->i_mode & MODEMASK; 18407c478bd9Sstevel@tonic-gate /* 18417c478bd9Sstevel@tonic-gate * If there is an ACL and there is a mask entry, then do the 18427c478bd9Sstevel@tonic-gate * extra work that completes the equivalent of an acltomode(3) 18437c478bd9Sstevel@tonic-gate * call. According to POSIX P1003.1e, the acl mask should be 18447c478bd9Sstevel@tonic-gate * returned in the group permissions field. 18457c478bd9Sstevel@tonic-gate * 18467c478bd9Sstevel@tonic-gate * - start with the original permission and mode bits (from above) 18477c478bd9Sstevel@tonic-gate * - clear the group owner bits 18487c478bd9Sstevel@tonic-gate * - add in the mask bits. 18497c478bd9Sstevel@tonic-gate */ 18507c478bd9Sstevel@tonic-gate if (ip->i_ufs_acl && ip->i_ufs_acl->aclass.acl_ismask) { 18517c478bd9Sstevel@tonic-gate vap->va_mode &= ~((VREAD | VWRITE | VEXEC) >> 3); 18527c478bd9Sstevel@tonic-gate vap->va_mode |= 18537c478bd9Sstevel@tonic-gate (ip->i_ufs_acl->aclass.acl_maskbits & PERMMASK) << 3; 18547c478bd9Sstevel@tonic-gate } 18557c478bd9Sstevel@tonic-gate vap->va_uid = ip->i_uid; 18567c478bd9Sstevel@tonic-gate vap->va_gid = ip->i_gid; 18577c478bd9Sstevel@tonic-gate vap->va_fsid = ip->i_dev; 18587c478bd9Sstevel@tonic-gate vap->va_nodeid = (ino64_t)ip->i_number; 18597c478bd9Sstevel@tonic-gate vap->va_nlink = ip->i_nlink; 18607c478bd9Sstevel@tonic-gate vap->va_size = ip->i_size; 18617c478bd9Sstevel@tonic-gate if (vp->v_type == VCHR || vp->v_type == VBLK) 18627c478bd9Sstevel@tonic-gate vap->va_rdev = ip->i_rdev; 18637c478bd9Sstevel@tonic-gate else 18647c478bd9Sstevel@tonic-gate vap->va_rdev = 0; /* not a b/c spec. */ 18657c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 18667c478bd9Sstevel@tonic-gate ITIMES_NOLOCK(ip); /* mark correct time in inode */ 18677c478bd9Sstevel@tonic-gate vap->va_seq = ip->i_seq; 18687c478bd9Sstevel@tonic-gate vap->va_atime.tv_sec = (time_t)ip->i_atime.tv_sec; 18697c478bd9Sstevel@tonic-gate vap->va_atime.tv_nsec = ip->i_atime.tv_usec*1000; 18707c478bd9Sstevel@tonic-gate vap->va_mtime.tv_sec = (time_t)ip->i_mtime.tv_sec; 18717c478bd9Sstevel@tonic-gate vap->va_mtime.tv_nsec = ip->i_mtime.tv_usec*1000; 18727c478bd9Sstevel@tonic-gate vap->va_ctime.tv_sec = (time_t)ip->i_ctime.tv_sec; 18737c478bd9Sstevel@tonic-gate vap->va_ctime.tv_nsec = ip->i_ctime.tv_usec*1000; 18747c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 18757c478bd9Sstevel@tonic-gate 18767c478bd9Sstevel@tonic-gate switch (ip->i_mode & IFMT) { 18777c478bd9Sstevel@tonic-gate 18787c478bd9Sstevel@tonic-gate case IFBLK: 18797c478bd9Sstevel@tonic-gate vap->va_blksize = MAXBSIZE; /* was BLKDEV_IOSIZE */ 18807c478bd9Sstevel@tonic-gate break; 18817c478bd9Sstevel@tonic-gate 18827c478bd9Sstevel@tonic-gate case IFCHR: 18837c478bd9Sstevel@tonic-gate vap->va_blksize = MAXBSIZE; 18847c478bd9Sstevel@tonic-gate break; 18857c478bd9Sstevel@tonic-gate 18867c478bd9Sstevel@tonic-gate default: 18877c478bd9Sstevel@tonic-gate vap->va_blksize = ip->i_fs->fs_bsize; 18887c478bd9Sstevel@tonic-gate break; 18897c478bd9Sstevel@tonic-gate } 18907c478bd9Sstevel@tonic-gate vap->va_nblocks = (fsblkcnt64_t)ip->i_blocks; 18917c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 18927c478bd9Sstevel@tonic-gate err = 0; 18937c478bd9Sstevel@tonic-gate 18947c478bd9Sstevel@tonic-gate out: 18957c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_GETATTR_END, "ufs_getattr_end:vp %p", vp); 18967c478bd9Sstevel@tonic-gate 18977c478bd9Sstevel@tonic-gate return (err); 18987c478bd9Sstevel@tonic-gate } 18997c478bd9Sstevel@tonic-gate 19007c478bd9Sstevel@tonic-gate /*ARGSUSED4*/ 19017c478bd9Sstevel@tonic-gate static int 19027c478bd9Sstevel@tonic-gate ufs_setattr( 19037c478bd9Sstevel@tonic-gate struct vnode *vp, 19047c478bd9Sstevel@tonic-gate struct vattr *vap, 19057c478bd9Sstevel@tonic-gate int flags, 19067c478bd9Sstevel@tonic-gate struct cred *cr, 19077c478bd9Sstevel@tonic-gate caller_context_t *ct) 19087c478bd9Sstevel@tonic-gate { 19097c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 19107c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp = ip->i_ufsvfs; 19117c478bd9Sstevel@tonic-gate struct fs *fs; 19127c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 19137c478bd9Sstevel@tonic-gate char *errmsg1; 19147c478bd9Sstevel@tonic-gate char *errmsg2; 19157c478bd9Sstevel@tonic-gate long blocks; 19167c478bd9Sstevel@tonic-gate long int mask = vap->va_mask; 19177c478bd9Sstevel@tonic-gate size_t len1, len2; 19187c478bd9Sstevel@tonic-gate int issync; 19197c478bd9Sstevel@tonic-gate int trans_size; 19207c478bd9Sstevel@tonic-gate int dotrans; 19217c478bd9Sstevel@tonic-gate int dorwlock; 19227c478bd9Sstevel@tonic-gate int error; 19237c478bd9Sstevel@tonic-gate int owner_change; 19247c478bd9Sstevel@tonic-gate int dodqlock; 19257c478bd9Sstevel@tonic-gate timestruc_t now; 19267c478bd9Sstevel@tonic-gate vattr_t oldva; 19277c478bd9Sstevel@tonic-gate int retry = 1; 19287c478bd9Sstevel@tonic-gate 19297c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_SETATTR_START, 19307c478bd9Sstevel@tonic-gate "ufs_setattr_start:vp %p flags %x", vp, flags); 19317c478bd9Sstevel@tonic-gate 19327c478bd9Sstevel@tonic-gate /* 19337c478bd9Sstevel@tonic-gate * Cannot set these attributes. 19347c478bd9Sstevel@tonic-gate */ 19357c478bd9Sstevel@tonic-gate if (mask & AT_NOSET) { 19367c478bd9Sstevel@tonic-gate error = EINVAL; 19377c478bd9Sstevel@tonic-gate goto out; 19387c478bd9Sstevel@tonic-gate } 19397c478bd9Sstevel@tonic-gate 19407c478bd9Sstevel@tonic-gate /* 19417c478bd9Sstevel@tonic-gate * check for forced unmount 19427c478bd9Sstevel@tonic-gate */ 19437c478bd9Sstevel@tonic-gate if (ufsvfsp == NULL) 19447c478bd9Sstevel@tonic-gate return (EIO); 19457c478bd9Sstevel@tonic-gate 19467c478bd9Sstevel@tonic-gate fs = ufsvfsp->vfs_fs; 19477c478bd9Sstevel@tonic-gate if (fs->fs_ronly != 0) 19487c478bd9Sstevel@tonic-gate return (EROFS); 19497c478bd9Sstevel@tonic-gate 19507c478bd9Sstevel@tonic-gate again: 19517c478bd9Sstevel@tonic-gate errmsg1 = NULL; 19527c478bd9Sstevel@tonic-gate errmsg2 = NULL; 19537c478bd9Sstevel@tonic-gate dotrans = 0; 19547c478bd9Sstevel@tonic-gate dorwlock = 0; 19557c478bd9Sstevel@tonic-gate dodqlock = 0; 19567c478bd9Sstevel@tonic-gate 19577c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_SETATTR_MASK); 19587c478bd9Sstevel@tonic-gate if (error) 19597c478bd9Sstevel@tonic-gate goto out; 19607c478bd9Sstevel@tonic-gate 19617c478bd9Sstevel@tonic-gate /* 19627c478bd9Sstevel@tonic-gate * Acquire i_rwlock before TRANS_BEGIN_CSYNC() if this is a file. 19637c478bd9Sstevel@tonic-gate * This follows the protocol for read()/write(). 19647c478bd9Sstevel@tonic-gate */ 19657c478bd9Sstevel@tonic-gate if (vp->v_type != VDIR) { 19667c478bd9Sstevel@tonic-gate rw_enter(&ip->i_rwlock, RW_WRITER); 19677c478bd9Sstevel@tonic-gate dorwlock = 1; 19687c478bd9Sstevel@tonic-gate } 19697c478bd9Sstevel@tonic-gate 19707c478bd9Sstevel@tonic-gate /* 19717c478bd9Sstevel@tonic-gate * Truncate file. Must have write permission and not be a directory. 19727c478bd9Sstevel@tonic-gate */ 19737c478bd9Sstevel@tonic-gate if (mask & AT_SIZE) { 19747c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_WRITER); 19757c478bd9Sstevel@tonic-gate if (vp->v_type == VDIR) { 19767c478bd9Sstevel@tonic-gate error = EISDIR; 19777c478bd9Sstevel@tonic-gate goto update_inode; 19787c478bd9Sstevel@tonic-gate } 19797c478bd9Sstevel@tonic-gate if (error = ufs_iaccess(ip, IWRITE, cr)) 19807c478bd9Sstevel@tonic-gate goto update_inode; 19817c478bd9Sstevel@tonic-gate 19827c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 19837c478bd9Sstevel@tonic-gate error = TRANS_ITRUNC(ip, vap->va_size, 0, cr); 19847c478bd9Sstevel@tonic-gate if (error) { 19857c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_WRITER); 19867c478bd9Sstevel@tonic-gate goto update_inode; 19877c478bd9Sstevel@tonic-gate } 19887c478bd9Sstevel@tonic-gate } 19897c478bd9Sstevel@tonic-gate 19907c478bd9Sstevel@tonic-gate if (ulp) { 19917c478bd9Sstevel@tonic-gate trans_size = (int)TOP_SETATTR_SIZE(ip); 19927c478bd9Sstevel@tonic-gate TRANS_BEGIN_CSYNC(ufsvfsp, issync, TOP_SETATTR, trans_size); 19937c478bd9Sstevel@tonic-gate ++dotrans; 19947c478bd9Sstevel@tonic-gate } 19957c478bd9Sstevel@tonic-gate 19967c478bd9Sstevel@tonic-gate /* 19977c478bd9Sstevel@tonic-gate * Acquire i_rwlock after TRANS_BEGIN_CSYNC() if this is a directory. 19987c478bd9Sstevel@tonic-gate * This follows the protocol established by 19997c478bd9Sstevel@tonic-gate * ufs_link/create/remove/rename/mkdir/rmdir/symlink. 20007c478bd9Sstevel@tonic-gate */ 20017c478bd9Sstevel@tonic-gate if (vp->v_type == VDIR) { 20027c478bd9Sstevel@tonic-gate rw_enter(&ip->i_rwlock, RW_WRITER); 20037c478bd9Sstevel@tonic-gate dorwlock = 1; 20047c478bd9Sstevel@tonic-gate } 20057c478bd9Sstevel@tonic-gate 20067c478bd9Sstevel@tonic-gate /* 20077c478bd9Sstevel@tonic-gate * Grab quota lock if we are changing the file's owner. 20087c478bd9Sstevel@tonic-gate */ 20097c478bd9Sstevel@tonic-gate if (mask & AT_UID) { 20107c478bd9Sstevel@tonic-gate rw_enter(&ufsvfsp->vfs_dqrwlock, RW_READER); 20117c478bd9Sstevel@tonic-gate dodqlock = 1; 20127c478bd9Sstevel@tonic-gate } 20137c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_WRITER); 20147c478bd9Sstevel@tonic-gate 20157c478bd9Sstevel@tonic-gate oldva.va_mode = ip->i_mode; 20167c478bd9Sstevel@tonic-gate oldva.va_uid = ip->i_uid; 20177c478bd9Sstevel@tonic-gate oldva.va_gid = ip->i_gid; 20187c478bd9Sstevel@tonic-gate 20197c478bd9Sstevel@tonic-gate vap->va_mask &= ~AT_SIZE; 20207c478bd9Sstevel@tonic-gate /* 20217c478bd9Sstevel@tonic-gate * ufs_iaccess is "close enough"; that's because it doesn't 20227c478bd9Sstevel@tonic-gate * map the defines. 20237c478bd9Sstevel@tonic-gate */ 20247c478bd9Sstevel@tonic-gate error = secpolicy_vnode_setattr(cr, vp, vap, &oldva, flags, 20257c478bd9Sstevel@tonic-gate ufs_iaccess, ip); 20267c478bd9Sstevel@tonic-gate if (error) 20277c478bd9Sstevel@tonic-gate goto update_inode; 20287c478bd9Sstevel@tonic-gate 20297c478bd9Sstevel@tonic-gate mask = vap->va_mask; 20307c478bd9Sstevel@tonic-gate 20317c478bd9Sstevel@tonic-gate /* 20327c478bd9Sstevel@tonic-gate * Change file access modes. 20337c478bd9Sstevel@tonic-gate */ 20347c478bd9Sstevel@tonic-gate if (mask & AT_MODE) { 20357c478bd9Sstevel@tonic-gate ip->i_mode = (ip->i_mode & IFMT) | (vap->va_mode & ~IFMT); 20367c478bd9Sstevel@tonic-gate TRANS_INODE(ufsvfsp, ip); 20377c478bd9Sstevel@tonic-gate ip->i_flag |= ICHG; 20387c478bd9Sstevel@tonic-gate if (stickyhack) { 20397c478bd9Sstevel@tonic-gate mutex_enter(&vp->v_lock); 20407c478bd9Sstevel@tonic-gate if ((ip->i_mode & (ISVTX | IEXEC | IFDIR)) == ISVTX) 20417c478bd9Sstevel@tonic-gate vp->v_flag |= VSWAPLIKE; 20427c478bd9Sstevel@tonic-gate else 20437c478bd9Sstevel@tonic-gate vp->v_flag &= ~VSWAPLIKE; 20447c478bd9Sstevel@tonic-gate mutex_exit(&vp->v_lock); 20457c478bd9Sstevel@tonic-gate } 20467c478bd9Sstevel@tonic-gate } 20477c478bd9Sstevel@tonic-gate if (mask & (AT_UID|AT_GID)) { 20487c478bd9Sstevel@tonic-gate if (mask & AT_UID) { 20497c478bd9Sstevel@tonic-gate /* 20507c478bd9Sstevel@tonic-gate * Don't change ownership of the quota inode. 20517c478bd9Sstevel@tonic-gate */ 20527c478bd9Sstevel@tonic-gate if (ufsvfsp->vfs_qinod == ip) { 20537c478bd9Sstevel@tonic-gate ASSERT(ufsvfsp->vfs_qflags & MQ_ENABLED); 20547c478bd9Sstevel@tonic-gate error = EINVAL; 20557c478bd9Sstevel@tonic-gate goto update_inode; 20567c478bd9Sstevel@tonic-gate } 20577c478bd9Sstevel@tonic-gate 20587c478bd9Sstevel@tonic-gate /* 20597c478bd9Sstevel@tonic-gate * No real ownership change. 20607c478bd9Sstevel@tonic-gate */ 20617c478bd9Sstevel@tonic-gate if (ip->i_uid == vap->va_uid) { 20627c478bd9Sstevel@tonic-gate blocks = 0; 20637c478bd9Sstevel@tonic-gate owner_change = 0; 20647c478bd9Sstevel@tonic-gate } 20657c478bd9Sstevel@tonic-gate /* 20667c478bd9Sstevel@tonic-gate * Remove the blocks and the file, from the old user's 20677c478bd9Sstevel@tonic-gate * quota. 20687c478bd9Sstevel@tonic-gate */ 20697c478bd9Sstevel@tonic-gate else { 20707c478bd9Sstevel@tonic-gate blocks = ip->i_blocks; 20717c478bd9Sstevel@tonic-gate owner_change = 1; 20727c478bd9Sstevel@tonic-gate 20737c478bd9Sstevel@tonic-gate (void) chkdq(ip, -blocks, /* force */ 1, cr, 20747c478bd9Sstevel@tonic-gate (char **)NULL, (size_t *)NULL); 20757c478bd9Sstevel@tonic-gate (void) chkiq(ufsvfsp, /* change */ -1, ip, 20767c478bd9Sstevel@tonic-gate (uid_t)ip->i_uid, 20777c478bd9Sstevel@tonic-gate /* force */ 1, cr, 20787c478bd9Sstevel@tonic-gate (char **)NULL, (size_t *)NULL); 20797c478bd9Sstevel@tonic-gate dqrele(ip->i_dquot); 20807c478bd9Sstevel@tonic-gate } 20817c478bd9Sstevel@tonic-gate 20827c478bd9Sstevel@tonic-gate ip->i_uid = vap->va_uid; 20837c478bd9Sstevel@tonic-gate 20847c478bd9Sstevel@tonic-gate /* 20857c478bd9Sstevel@tonic-gate * There is a real ownership change. 20867c478bd9Sstevel@tonic-gate */ 20877c478bd9Sstevel@tonic-gate if (owner_change) { 20887c478bd9Sstevel@tonic-gate /* 20897c478bd9Sstevel@tonic-gate * Add the blocks and the file to the new 20907c478bd9Sstevel@tonic-gate * user's quota. 20917c478bd9Sstevel@tonic-gate */ 20927c478bd9Sstevel@tonic-gate ip->i_dquot = getinoquota(ip); 20937c478bd9Sstevel@tonic-gate (void) chkdq(ip, blocks, /* force */ 1, cr, 20947c478bd9Sstevel@tonic-gate &errmsg1, &len1); 20957c478bd9Sstevel@tonic-gate (void) chkiq(ufsvfsp, /* change */ 1, 20967c478bd9Sstevel@tonic-gate (struct inode *)NULL, 20977c478bd9Sstevel@tonic-gate (uid_t)ip->i_uid, 20987c478bd9Sstevel@tonic-gate /* force */ 1, cr, 20997c478bd9Sstevel@tonic-gate &errmsg2, &len2); 21007c478bd9Sstevel@tonic-gate } 21017c478bd9Sstevel@tonic-gate } 21027c478bd9Sstevel@tonic-gate if (mask & AT_GID) { 21037c478bd9Sstevel@tonic-gate ip->i_gid = vap->va_gid; 21047c478bd9Sstevel@tonic-gate } 21057c478bd9Sstevel@tonic-gate TRANS_INODE(ufsvfsp, ip); 21067c478bd9Sstevel@tonic-gate ip->i_flag |= ICHG; 21077c478bd9Sstevel@tonic-gate } 21087c478bd9Sstevel@tonic-gate /* 21097c478bd9Sstevel@tonic-gate * Change file access or modified times. 21107c478bd9Sstevel@tonic-gate */ 21117c478bd9Sstevel@tonic-gate if (mask & (AT_ATIME|AT_MTIME)) { 21127c478bd9Sstevel@tonic-gate /* Check that the time value is within ufs range */ 21137c478bd9Sstevel@tonic-gate if (((mask & AT_ATIME) && TIMESPEC_OVERFLOW(&vap->va_atime)) || 21147c478bd9Sstevel@tonic-gate ((mask & AT_MTIME) && TIMESPEC_OVERFLOW(&vap->va_mtime))) { 21157c478bd9Sstevel@tonic-gate error = EOVERFLOW; 21167c478bd9Sstevel@tonic-gate goto update_inode; 21177c478bd9Sstevel@tonic-gate } 21187c478bd9Sstevel@tonic-gate 21197c478bd9Sstevel@tonic-gate /* 21207c478bd9Sstevel@tonic-gate * if the "noaccess" mount option is set and only atime 21217c478bd9Sstevel@tonic-gate * update is requested, do nothing. No error is returned. 21227c478bd9Sstevel@tonic-gate */ 21237c478bd9Sstevel@tonic-gate if ((ufsvfsp->vfs_noatime) && 21247c478bd9Sstevel@tonic-gate ((mask & (AT_ATIME|AT_MTIME)) == AT_ATIME)) 21257c478bd9Sstevel@tonic-gate goto skip_atime; 21267c478bd9Sstevel@tonic-gate 21277c478bd9Sstevel@tonic-gate if (mask & AT_ATIME) { 21287c478bd9Sstevel@tonic-gate ip->i_atime.tv_sec = vap->va_atime.tv_sec; 21297c478bd9Sstevel@tonic-gate ip->i_atime.tv_usec = vap->va_atime.tv_nsec / 1000; 21307c478bd9Sstevel@tonic-gate ip->i_flag &= ~IACC; 21317c478bd9Sstevel@tonic-gate } 21327c478bd9Sstevel@tonic-gate if (mask & AT_MTIME) { 21337c478bd9Sstevel@tonic-gate ip->i_mtime.tv_sec = vap->va_mtime.tv_sec; 21347c478bd9Sstevel@tonic-gate ip->i_mtime.tv_usec = vap->va_mtime.tv_nsec / 1000; 21357c478bd9Sstevel@tonic-gate gethrestime(&now); 21367c478bd9Sstevel@tonic-gate if (now.tv_sec > TIME32_MAX) { 21377c478bd9Sstevel@tonic-gate /* 21387c478bd9Sstevel@tonic-gate * In 2038, ctime sticks forever.. 21397c478bd9Sstevel@tonic-gate */ 21407c478bd9Sstevel@tonic-gate ip->i_ctime.tv_sec = TIME32_MAX; 21417c478bd9Sstevel@tonic-gate ip->i_ctime.tv_usec = 0; 21427c478bd9Sstevel@tonic-gate } else { 21437c478bd9Sstevel@tonic-gate ip->i_ctime.tv_sec = now.tv_sec; 21447c478bd9Sstevel@tonic-gate ip->i_ctime.tv_usec = now.tv_nsec / 1000; 21457c478bd9Sstevel@tonic-gate } 21467c478bd9Sstevel@tonic-gate ip->i_flag &= ~(IUPD|ICHG); 21477c478bd9Sstevel@tonic-gate ip->i_flag |= IMODTIME; 21487c478bd9Sstevel@tonic-gate } 21497c478bd9Sstevel@tonic-gate TRANS_INODE(ufsvfsp, ip); 21507c478bd9Sstevel@tonic-gate ip->i_flag |= IMOD; 21517c478bd9Sstevel@tonic-gate } 21527c478bd9Sstevel@tonic-gate 21537c478bd9Sstevel@tonic-gate skip_atime: 21547c478bd9Sstevel@tonic-gate /* 21557c478bd9Sstevel@tonic-gate * The presence of a shadow inode may indicate an ACL, but does 21567c478bd9Sstevel@tonic-gate * not imply an ACL. Future FSD types should be handled here too 21577c478bd9Sstevel@tonic-gate * and check for the presence of the attribute-specific data 21587c478bd9Sstevel@tonic-gate * before referencing it. 21597c478bd9Sstevel@tonic-gate */ 21607c478bd9Sstevel@tonic-gate if (ip->i_shadow) { 21617c478bd9Sstevel@tonic-gate /* 21627c478bd9Sstevel@tonic-gate * XXX if ufs_iupdat is changed to sandbagged write fix 21637c478bd9Sstevel@tonic-gate * ufs_acl_setattr to push ip to keep acls consistent 21647c478bd9Sstevel@tonic-gate * 21657c478bd9Sstevel@tonic-gate * Suppress out of inodes messages if we will retry. 21667c478bd9Sstevel@tonic-gate */ 21677c478bd9Sstevel@tonic-gate if (retry) 21687c478bd9Sstevel@tonic-gate ip->i_flag |= IQUIET; 21697c478bd9Sstevel@tonic-gate error = ufs_acl_setattr(ip, vap, cr); 21707c478bd9Sstevel@tonic-gate ip->i_flag &= ~IQUIET; 21717c478bd9Sstevel@tonic-gate } 21727c478bd9Sstevel@tonic-gate 21737c478bd9Sstevel@tonic-gate update_inode: 21747c478bd9Sstevel@tonic-gate /* 21757c478bd9Sstevel@tonic-gate * Setattr always increases the sequence number 21767c478bd9Sstevel@tonic-gate */ 21777c478bd9Sstevel@tonic-gate ip->i_seq++; 21787c478bd9Sstevel@tonic-gate 21797c478bd9Sstevel@tonic-gate /* 21807c478bd9Sstevel@tonic-gate * if nfsd and not logging; push synchronously 21817c478bd9Sstevel@tonic-gate */ 21827c478bd9Sstevel@tonic-gate if ((curthread->t_flag & T_DONTPEND) && !TRANS_ISTRANS(ufsvfsp)) { 21837c478bd9Sstevel@tonic-gate ufs_iupdat(ip, 1); 21847c478bd9Sstevel@tonic-gate } else { 21857c478bd9Sstevel@tonic-gate ITIMES_NOLOCK(ip); 21867c478bd9Sstevel@tonic-gate } 21877c478bd9Sstevel@tonic-gate 21887c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 21897c478bd9Sstevel@tonic-gate if (dodqlock) { 21907c478bd9Sstevel@tonic-gate rw_exit(&ufsvfsp->vfs_dqrwlock); 21917c478bd9Sstevel@tonic-gate } 21927c478bd9Sstevel@tonic-gate if (dorwlock) 21937c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 21947c478bd9Sstevel@tonic-gate 21957c478bd9Sstevel@tonic-gate if (ulp) { 21967c478bd9Sstevel@tonic-gate if (dotrans) { 21977c478bd9Sstevel@tonic-gate int terr = 0; 21987c478bd9Sstevel@tonic-gate TRANS_END_CSYNC(ufsvfsp, terr, issync, TOP_SETATTR, 21997c478bd9Sstevel@tonic-gate trans_size); 22007c478bd9Sstevel@tonic-gate if (error == 0) 22017c478bd9Sstevel@tonic-gate error = terr; 22027c478bd9Sstevel@tonic-gate } 22037c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 22047c478bd9Sstevel@tonic-gate } 22057c478bd9Sstevel@tonic-gate out: 22067c478bd9Sstevel@tonic-gate /* 22077c478bd9Sstevel@tonic-gate * If out of inodes or blocks, see if we can free something 22087c478bd9Sstevel@tonic-gate * up from the delete queue. 22097c478bd9Sstevel@tonic-gate */ 22107c478bd9Sstevel@tonic-gate if ((error == ENOSPC) && retry && TRANS_ISTRANS(ufsvfsp)) { 22117c478bd9Sstevel@tonic-gate ufs_delete_drain_wait(ufsvfsp, 1); 22127c478bd9Sstevel@tonic-gate retry = 0; 22137c478bd9Sstevel@tonic-gate if (errmsg1 != NULL) 22147c478bd9Sstevel@tonic-gate kmem_free(errmsg1, len1); 22157c478bd9Sstevel@tonic-gate if (errmsg2 != NULL) 22167c478bd9Sstevel@tonic-gate kmem_free(errmsg2, len2); 22177c478bd9Sstevel@tonic-gate goto again; 22187c478bd9Sstevel@tonic-gate } 22197c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_SETATTR_END, 22207c478bd9Sstevel@tonic-gate "ufs_setattr_end:vp %p error %d", vp, error); 22217c478bd9Sstevel@tonic-gate if (errmsg1 != NULL) { 22227c478bd9Sstevel@tonic-gate uprintf(errmsg1); 22237c478bd9Sstevel@tonic-gate kmem_free(errmsg1, len1); 22247c478bd9Sstevel@tonic-gate } 22257c478bd9Sstevel@tonic-gate if (errmsg2 != NULL) { 22267c478bd9Sstevel@tonic-gate uprintf(errmsg2); 22277c478bd9Sstevel@tonic-gate kmem_free(errmsg2, len2); 22287c478bd9Sstevel@tonic-gate } 22297c478bd9Sstevel@tonic-gate return (error); 22307c478bd9Sstevel@tonic-gate } 22317c478bd9Sstevel@tonic-gate 22327c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 22337c478bd9Sstevel@tonic-gate static int 22347c478bd9Sstevel@tonic-gate ufs_access(struct vnode *vp, int mode, int flags, struct cred *cr) 22357c478bd9Sstevel@tonic-gate { 22367c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 22377c478bd9Sstevel@tonic-gate int error; 22387c478bd9Sstevel@tonic-gate 22397c478bd9Sstevel@tonic-gate TRACE_3(TR_FAC_UFS, TR_UFS_ACCESS_START, 22407c478bd9Sstevel@tonic-gate "ufs_access_start:vp %p mode %x flags %x", vp, mode, flags); 22417c478bd9Sstevel@tonic-gate 22427c478bd9Sstevel@tonic-gate if (ip->i_ufsvfs == NULL) 22437c478bd9Sstevel@tonic-gate return (EIO); 22447c478bd9Sstevel@tonic-gate 22457c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_READER); 22467c478bd9Sstevel@tonic-gate 22477c478bd9Sstevel@tonic-gate /* 22487c478bd9Sstevel@tonic-gate * The ufs_iaccess function wants to be called with 22497c478bd9Sstevel@tonic-gate * mode bits expressed as "ufs specific" bits. 22507c478bd9Sstevel@tonic-gate * I.e., VWRITE|VREAD|VEXEC do not make sense to 22517c478bd9Sstevel@tonic-gate * ufs_iaccess() but IWRITE|IREAD|IEXEC do. 22527c478bd9Sstevel@tonic-gate * But since they're the same we just pass the vnode mode 22537c478bd9Sstevel@tonic-gate * bit but just verify that assumption at compile time. 22547c478bd9Sstevel@tonic-gate */ 22557c478bd9Sstevel@tonic-gate #if IWRITE != VWRITE || IREAD != VREAD || IEXEC != VEXEC 22567c478bd9Sstevel@tonic-gate #error "ufs_access needs to map Vmodes to Imodes" 22577c478bd9Sstevel@tonic-gate #endif 22587c478bd9Sstevel@tonic-gate error = ufs_iaccess(ip, mode, cr); 22597c478bd9Sstevel@tonic-gate 22607c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 22617c478bd9Sstevel@tonic-gate 22627c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_ACCESS_END, 22637c478bd9Sstevel@tonic-gate "ufs_access_end:vp %p error %d", vp, error); 22647c478bd9Sstevel@tonic-gate return (error); 22657c478bd9Sstevel@tonic-gate } 22667c478bd9Sstevel@tonic-gate 22677c478bd9Sstevel@tonic-gate /* ARGSUSED */ 22687c478bd9Sstevel@tonic-gate static int 22697c478bd9Sstevel@tonic-gate ufs_readlink(struct vnode *vp, struct uio *uiop, struct cred *cr) 22707c478bd9Sstevel@tonic-gate { 22717c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 22727c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp; 22737c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 22747c478bd9Sstevel@tonic-gate int error; 22757c478bd9Sstevel@tonic-gate int fastsymlink; 22767c478bd9Sstevel@tonic-gate 22777c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_READLINK_START, 22787c478bd9Sstevel@tonic-gate "ufs_readlink_start:vp %p uiop %p", uiop, vp); 22797c478bd9Sstevel@tonic-gate 22807c478bd9Sstevel@tonic-gate if (vp->v_type != VLNK) { 22817c478bd9Sstevel@tonic-gate error = EINVAL; 22827c478bd9Sstevel@tonic-gate goto nolockout; 22837c478bd9Sstevel@tonic-gate } 22847c478bd9Sstevel@tonic-gate 22857c478bd9Sstevel@tonic-gate /* 22867c478bd9Sstevel@tonic-gate * If the symbolic link is empty there is nothing to read. 22877c478bd9Sstevel@tonic-gate * Fast-track these empty symbolic links 22887c478bd9Sstevel@tonic-gate */ 22897c478bd9Sstevel@tonic-gate if (ip->i_size == 0) { 22907c478bd9Sstevel@tonic-gate error = 0; 22917c478bd9Sstevel@tonic-gate goto nolockout; 22927c478bd9Sstevel@tonic-gate } 22937c478bd9Sstevel@tonic-gate 22947c478bd9Sstevel@tonic-gate ufsvfsp = ip->i_ufsvfs; 22957c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_READLINK_MASK); 22967c478bd9Sstevel@tonic-gate if (error) 22977c478bd9Sstevel@tonic-gate goto nolockout; 22987c478bd9Sstevel@tonic-gate /* 22997c478bd9Sstevel@tonic-gate * The ip->i_rwlock protects the data blocks used for FASTSYMLINK 23007c478bd9Sstevel@tonic-gate */ 23017c478bd9Sstevel@tonic-gate again: 23027c478bd9Sstevel@tonic-gate fastsymlink = 0; 23037c478bd9Sstevel@tonic-gate if (ip->i_flag & IFASTSYMLNK) { 23047c478bd9Sstevel@tonic-gate rw_enter(&ip->i_rwlock, RW_READER); 23057c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_READER); 23067c478bd9Sstevel@tonic-gate if (ip->i_flag & IFASTSYMLNK) { 23077c478bd9Sstevel@tonic-gate if (!ULOCKFS_IS_NOIACC(ITOUL(ip)) && 23087c478bd9Sstevel@tonic-gate (ip->i_fs->fs_ronly == 0) && 23097c478bd9Sstevel@tonic-gate (!ufsvfsp->vfs_noatime)) { 23107c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 23117c478bd9Sstevel@tonic-gate ip->i_flag |= IACC; 23127c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 23137c478bd9Sstevel@tonic-gate } 23147c478bd9Sstevel@tonic-gate error = uiomove((caddr_t)&ip->i_db[1], 23157c478bd9Sstevel@tonic-gate MIN(ip->i_size, uiop->uio_resid), 23167c478bd9Sstevel@tonic-gate UIO_READ, uiop); 23177c478bd9Sstevel@tonic-gate ITIMES(ip); 23187c478bd9Sstevel@tonic-gate ++fastsymlink; 23197c478bd9Sstevel@tonic-gate } 23207c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 23217c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 23227c478bd9Sstevel@tonic-gate } 23237c478bd9Sstevel@tonic-gate if (!fastsymlink) { 23247c478bd9Sstevel@tonic-gate ssize_t size; /* number of bytes read */ 23257c478bd9Sstevel@tonic-gate caddr_t basep; /* pointer to input data */ 23267c478bd9Sstevel@tonic-gate ino_t ino; 23277c478bd9Sstevel@tonic-gate long igen; 23287c478bd9Sstevel@tonic-gate struct uio tuio; /* temp uio struct */ 23297c478bd9Sstevel@tonic-gate struct uio *tuiop; 23307c478bd9Sstevel@tonic-gate iovec_t tiov; /* temp iovec struct */ 23317c478bd9Sstevel@tonic-gate char kbuf[FSL_SIZE]; /* buffer to hold fast symlink */ 23327c478bd9Sstevel@tonic-gate int tflag = 0; /* flag to indicate temp vars used */ 23337c478bd9Sstevel@tonic-gate 23347c478bd9Sstevel@tonic-gate ino = ip->i_number; 23357c478bd9Sstevel@tonic-gate igen = ip->i_gen; 23367c478bd9Sstevel@tonic-gate size = uiop->uio_resid; 23377c478bd9Sstevel@tonic-gate basep = uiop->uio_iov->iov_base; 23387c478bd9Sstevel@tonic-gate tuiop = uiop; 23397c478bd9Sstevel@tonic-gate 23407c478bd9Sstevel@tonic-gate rw_enter(&ip->i_rwlock, RW_WRITER); 23417c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_WRITER); 23427c478bd9Sstevel@tonic-gate if (ip->i_flag & IFASTSYMLNK) { 23437c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 23447c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 23457c478bd9Sstevel@tonic-gate goto again; 23467c478bd9Sstevel@tonic-gate } 23477c478bd9Sstevel@tonic-gate 23487c478bd9Sstevel@tonic-gate /* can this be a fast symlink and is it a user buffer? */ 23497c478bd9Sstevel@tonic-gate if (ip->i_size <= FSL_SIZE && 23507c478bd9Sstevel@tonic-gate (uiop->uio_segflg == UIO_USERSPACE || 23517c478bd9Sstevel@tonic-gate uiop->uio_segflg == UIO_USERISPACE)) { 23527c478bd9Sstevel@tonic-gate 23537c478bd9Sstevel@tonic-gate bzero(&tuio, sizeof (struct uio)); 23547c478bd9Sstevel@tonic-gate /* 23557c478bd9Sstevel@tonic-gate * setup a kernel buffer to read link into. this 23567c478bd9Sstevel@tonic-gate * is to fix a race condition where the user buffer 23577c478bd9Sstevel@tonic-gate * got corrupted before copying it into the inode. 23587c478bd9Sstevel@tonic-gate */ 23597c478bd9Sstevel@tonic-gate size = ip->i_size; 23607c478bd9Sstevel@tonic-gate tiov.iov_len = size; 23617c478bd9Sstevel@tonic-gate tiov.iov_base = kbuf; 23627c478bd9Sstevel@tonic-gate tuio.uio_iov = &tiov; 23637c478bd9Sstevel@tonic-gate tuio.uio_iovcnt = 1; 23647c478bd9Sstevel@tonic-gate tuio.uio_offset = uiop->uio_offset; 23657c478bd9Sstevel@tonic-gate tuio.uio_segflg = UIO_SYSSPACE; 23667c478bd9Sstevel@tonic-gate tuio.uio_fmode = uiop->uio_fmode; 23677c478bd9Sstevel@tonic-gate tuio.uio_extflg = uiop->uio_extflg; 23687c478bd9Sstevel@tonic-gate tuio.uio_limit = uiop->uio_limit; 23697c478bd9Sstevel@tonic-gate tuio.uio_resid = size; 23707c478bd9Sstevel@tonic-gate 23717c478bd9Sstevel@tonic-gate basep = tuio.uio_iov->iov_base; 23727c478bd9Sstevel@tonic-gate tuiop = &tuio; 23737c478bd9Sstevel@tonic-gate tflag = 1; 23747c478bd9Sstevel@tonic-gate } 23757c478bd9Sstevel@tonic-gate 23767c478bd9Sstevel@tonic-gate error = rdip(ip, tuiop, 0, cr); 23777c478bd9Sstevel@tonic-gate if (!(error == 0 && ip->i_number == ino && ip->i_gen == igen)) { 23787c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 23797c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 23807c478bd9Sstevel@tonic-gate goto out; 23817c478bd9Sstevel@tonic-gate } 23827c478bd9Sstevel@tonic-gate 23837c478bd9Sstevel@tonic-gate if (tflag == 0) 23847c478bd9Sstevel@tonic-gate size -= uiop->uio_resid; 23857c478bd9Sstevel@tonic-gate 23867c478bd9Sstevel@tonic-gate if ((tflag == 0 && ip->i_size <= FSL_SIZE && 23877c478bd9Sstevel@tonic-gate ip->i_size == size) || (tflag == 1 && 23887c478bd9Sstevel@tonic-gate tuio.uio_resid == 0)) { 23897c478bd9Sstevel@tonic-gate error = kcopy(basep, &ip->i_db[1], ip->i_size); 23907c478bd9Sstevel@tonic-gate if (error == 0) { 23917c478bd9Sstevel@tonic-gate ip->i_flag |= IFASTSYMLNK; 23927c478bd9Sstevel@tonic-gate /* 23937c478bd9Sstevel@tonic-gate * free page 23947c478bd9Sstevel@tonic-gate */ 23957c478bd9Sstevel@tonic-gate (void) VOP_PUTPAGE(ITOV(ip), 23967c478bd9Sstevel@tonic-gate (offset_t)0, PAGESIZE, 23977c478bd9Sstevel@tonic-gate (B_DONTNEED | B_FREE | B_FORCE | B_ASYNC), 23987c478bd9Sstevel@tonic-gate cr); 23997c478bd9Sstevel@tonic-gate } else { 24007c478bd9Sstevel@tonic-gate int i; 24017c478bd9Sstevel@tonic-gate /* error, clear garbage left behind */ 24027c478bd9Sstevel@tonic-gate for (i = 1; i < NDADDR; i++) 24037c478bd9Sstevel@tonic-gate ip->i_db[i] = 0; 24047c478bd9Sstevel@tonic-gate for (i = 0; i < NIADDR; i++) 24057c478bd9Sstevel@tonic-gate ip->i_ib[i] = 0; 24067c478bd9Sstevel@tonic-gate } 24077c478bd9Sstevel@tonic-gate } 24087c478bd9Sstevel@tonic-gate if (tflag == 1) { 24097c478bd9Sstevel@tonic-gate /* now, copy it into the user buffer */ 24107c478bd9Sstevel@tonic-gate error = uiomove((caddr_t)kbuf, 24117c478bd9Sstevel@tonic-gate MIN(size, uiop->uio_resid), 24127c478bd9Sstevel@tonic-gate UIO_READ, uiop); 24137c478bd9Sstevel@tonic-gate } 24147c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 24157c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 24167c478bd9Sstevel@tonic-gate } 24177c478bd9Sstevel@tonic-gate out: 24187c478bd9Sstevel@tonic-gate if (ulp) { 24197c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 24207c478bd9Sstevel@tonic-gate } 24217c478bd9Sstevel@tonic-gate nolockout: 24227c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_READLINK_END, 24237c478bd9Sstevel@tonic-gate "ufs_readlink_end:vp %p error %d", vp, error); 24247c478bd9Sstevel@tonic-gate 24257c478bd9Sstevel@tonic-gate return (error); 24267c478bd9Sstevel@tonic-gate } 24277c478bd9Sstevel@tonic-gate 24287c478bd9Sstevel@tonic-gate /* ARGSUSED */ 24297c478bd9Sstevel@tonic-gate static int 24307c478bd9Sstevel@tonic-gate ufs_fsync(struct vnode *vp, int syncflag, struct cred *cr) 24317c478bd9Sstevel@tonic-gate { 24327c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 24337c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp = ip->i_ufsvfs; 24347c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 24357c478bd9Sstevel@tonic-gate int error; 24367c478bd9Sstevel@tonic-gate 24377c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_FSYNC_START, 24387c478bd9Sstevel@tonic-gate "ufs_fsync_start:vp %p", vp); 24397c478bd9Sstevel@tonic-gate 24407c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_FSYNC_MASK); 24417c478bd9Sstevel@tonic-gate if (error) 24427c478bd9Sstevel@tonic-gate return (error); 24437c478bd9Sstevel@tonic-gate 24447c478bd9Sstevel@tonic-gate if (TRANS_ISTRANS(ufsvfsp)) { 24457c478bd9Sstevel@tonic-gate /* 24467c478bd9Sstevel@tonic-gate * First push out any data pages 24477c478bd9Sstevel@tonic-gate */ 24487c478bd9Sstevel@tonic-gate if (vn_has_cached_data(vp) && !(syncflag & FNODSYNC) && 24497c478bd9Sstevel@tonic-gate (vp->v_type != VCHR) && !(IS_SWAPVP(vp))) { 24507c478bd9Sstevel@tonic-gate error = VOP_PUTPAGE(vp, (offset_t)0, (size_t)0, 24517c478bd9Sstevel@tonic-gate 0, CRED()); 24527c478bd9Sstevel@tonic-gate if (error) 24537c478bd9Sstevel@tonic-gate goto out; 24547c478bd9Sstevel@tonic-gate } 24557c478bd9Sstevel@tonic-gate 24567c478bd9Sstevel@tonic-gate /* 24577c478bd9Sstevel@tonic-gate * Delta any delayed inode times updates 24587c478bd9Sstevel@tonic-gate * and push inode to log. 24597c478bd9Sstevel@tonic-gate * All other inode deltas will have already been delta'd 24607c478bd9Sstevel@tonic-gate * and will be pushed during the commit. 24617c478bd9Sstevel@tonic-gate */ 24627c478bd9Sstevel@tonic-gate if (!(syncflag & FDSYNC) && 24637c478bd9Sstevel@tonic-gate ((ip->i_flag & (IMOD|IMODACC)) == IMODACC)) { 24647c478bd9Sstevel@tonic-gate if (ulp) { 24657c478bd9Sstevel@tonic-gate TRANS_BEGIN_ASYNC(ufsvfsp, TOP_FSYNC, 24667c478bd9Sstevel@tonic-gate TOP_SYNCIP_SIZE); 24677c478bd9Sstevel@tonic-gate } 24687c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_READER); 24697c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 24707c478bd9Sstevel@tonic-gate ip->i_flag &= ~IMODTIME; 24717c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 24727c478bd9Sstevel@tonic-gate ufs_iupdat(ip, I_SYNC); 24737c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 24747c478bd9Sstevel@tonic-gate if (ulp) { 24757c478bd9Sstevel@tonic-gate TRANS_END_ASYNC(ufsvfsp, TOP_FSYNC, 24767c478bd9Sstevel@tonic-gate TOP_SYNCIP_SIZE); 24777c478bd9Sstevel@tonic-gate } 24787c478bd9Sstevel@tonic-gate } 24797c478bd9Sstevel@tonic-gate 24807c478bd9Sstevel@tonic-gate /* 24817c478bd9Sstevel@tonic-gate * Commit the Moby transaction 24827c478bd9Sstevel@tonic-gate * 24837c478bd9Sstevel@tonic-gate * Deltas have already been made so we just need to 24847c478bd9Sstevel@tonic-gate * commit them with a synchronous transaction. 24857c478bd9Sstevel@tonic-gate * TRANS_BEGIN_SYNC() will return an error 24867c478bd9Sstevel@tonic-gate * if there are no deltas to commit, for an 24877c478bd9Sstevel@tonic-gate * empty transaction. 24887c478bd9Sstevel@tonic-gate */ 24897c478bd9Sstevel@tonic-gate if (ulp) { 24907c478bd9Sstevel@tonic-gate TRANS_BEGIN_SYNC(ufsvfsp, TOP_FSYNC, TOP_COMMIT_SIZE, 24917c478bd9Sstevel@tonic-gate error); 24927c478bd9Sstevel@tonic-gate if (error) { 24937c478bd9Sstevel@tonic-gate error = 0; /* commit wasn't needed */ 24947c478bd9Sstevel@tonic-gate goto out; 24957c478bd9Sstevel@tonic-gate } 24967c478bd9Sstevel@tonic-gate TRANS_END_SYNC(ufsvfsp, error, TOP_FSYNC, 24977c478bd9Sstevel@tonic-gate TOP_COMMIT_SIZE); 24987c478bd9Sstevel@tonic-gate } 24997c478bd9Sstevel@tonic-gate } else { /* not logging */ 25007c478bd9Sstevel@tonic-gate if (!(IS_SWAPVP(vp))) 25017c478bd9Sstevel@tonic-gate if (syncflag & FNODSYNC) { 25027c478bd9Sstevel@tonic-gate /* Just update the inode only */ 25037c478bd9Sstevel@tonic-gate TRANS_IUPDAT(ip, 1); 25047c478bd9Sstevel@tonic-gate error = 0; 25057c478bd9Sstevel@tonic-gate } else if (syncflag & FDSYNC) 25067c478bd9Sstevel@tonic-gate /* Do data-synchronous writes */ 25077c478bd9Sstevel@tonic-gate error = TRANS_SYNCIP(ip, 0, I_DSYNC, TOP_FSYNC); 25087c478bd9Sstevel@tonic-gate else 25097c478bd9Sstevel@tonic-gate /* Do synchronous writes */ 25107c478bd9Sstevel@tonic-gate error = TRANS_SYNCIP(ip, 0, I_SYNC, TOP_FSYNC); 25117c478bd9Sstevel@tonic-gate 25127c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_WRITER); 25137c478bd9Sstevel@tonic-gate if (!error) 25147c478bd9Sstevel@tonic-gate error = ufs_sync_indir(ip); 25157c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 25167c478bd9Sstevel@tonic-gate } 25177c478bd9Sstevel@tonic-gate out: 25187c478bd9Sstevel@tonic-gate if (ulp) { 25197c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 25207c478bd9Sstevel@tonic-gate } 25217c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_FSYNC_END, 25227c478bd9Sstevel@tonic-gate "ufs_fsync_end:vp %p error %d", vp, error); 25237c478bd9Sstevel@tonic-gate return (error); 25247c478bd9Sstevel@tonic-gate } 25257c478bd9Sstevel@tonic-gate 25267c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 25277c478bd9Sstevel@tonic-gate static void 25287c478bd9Sstevel@tonic-gate ufs_inactive(struct vnode *vp, struct cred *cr) 25297c478bd9Sstevel@tonic-gate { 25307c478bd9Sstevel@tonic-gate ufs_iinactive(VTOI(vp)); 25317c478bd9Sstevel@tonic-gate } 25327c478bd9Sstevel@tonic-gate 25337c478bd9Sstevel@tonic-gate /* 25347c478bd9Sstevel@tonic-gate * Unix file system operations having to do with directory manipulation. 25357c478bd9Sstevel@tonic-gate */ 25367c478bd9Sstevel@tonic-gate int ufs_lookup_idle_count = 2; /* Number of inodes to idle each time */ 25377c478bd9Sstevel@tonic-gate /* ARGSUSED */ 25387c478bd9Sstevel@tonic-gate static int 25397c478bd9Sstevel@tonic-gate ufs_lookup(struct vnode *dvp, char *nm, struct vnode **vpp, 25407c478bd9Sstevel@tonic-gate struct pathname *pnp, int flags, struct vnode *rdir, struct cred *cr) 25417c478bd9Sstevel@tonic-gate { 25427c478bd9Sstevel@tonic-gate struct inode *ip; 25437c478bd9Sstevel@tonic-gate struct inode *sip; 25447c478bd9Sstevel@tonic-gate struct inode *xip; 25457c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp; 25467c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 25477c478bd9Sstevel@tonic-gate struct vnode *vp; 25487c478bd9Sstevel@tonic-gate int error; 25497c478bd9Sstevel@tonic-gate 25507c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_LOOKUP_START, 25517c478bd9Sstevel@tonic-gate "ufs_lookup_start:dvp %p name %s", dvp, nm); 25527c478bd9Sstevel@tonic-gate 25537c478bd9Sstevel@tonic-gate 25547c478bd9Sstevel@tonic-gate /* 25557c478bd9Sstevel@tonic-gate * Check flags for type of lookup (regular file or attribute file) 25567c478bd9Sstevel@tonic-gate */ 25577c478bd9Sstevel@tonic-gate 25587c478bd9Sstevel@tonic-gate ip = VTOI(dvp); 25597c478bd9Sstevel@tonic-gate 25607c478bd9Sstevel@tonic-gate if (flags & LOOKUP_XATTR) { 25617c478bd9Sstevel@tonic-gate 25627c478bd9Sstevel@tonic-gate /* 25637c478bd9Sstevel@tonic-gate * We don't allow recursive attributes... 25647c478bd9Sstevel@tonic-gate * Maybe someday we will. 25657c478bd9Sstevel@tonic-gate */ 25667c478bd9Sstevel@tonic-gate if ((ip->i_cflags & IXATTR)) { 25677c478bd9Sstevel@tonic-gate return (EINVAL); 25687c478bd9Sstevel@tonic-gate } 25697c478bd9Sstevel@tonic-gate 25707c478bd9Sstevel@tonic-gate if ((vp = dnlc_lookup(dvp, XATTR_DIR_NAME)) == NULL) { 25717c478bd9Sstevel@tonic-gate error = ufs_xattr_getattrdir(dvp, &sip, flags, cr); 25727c478bd9Sstevel@tonic-gate if (error) { 25737c478bd9Sstevel@tonic-gate *vpp = NULL; 25747c478bd9Sstevel@tonic-gate goto out; 25757c478bd9Sstevel@tonic-gate } 25767c478bd9Sstevel@tonic-gate 25777c478bd9Sstevel@tonic-gate vp = ITOV(sip); 25787c478bd9Sstevel@tonic-gate dnlc_update(dvp, XATTR_DIR_NAME, vp); 25797c478bd9Sstevel@tonic-gate } 25807c478bd9Sstevel@tonic-gate 25817c478bd9Sstevel@tonic-gate /* 25827c478bd9Sstevel@tonic-gate * Check accessibility of directory. 25837c478bd9Sstevel@tonic-gate */ 25847c478bd9Sstevel@tonic-gate if (vp == DNLC_NO_VNODE) { 25857c478bd9Sstevel@tonic-gate VN_RELE(vp); 25867c478bd9Sstevel@tonic-gate error = ENOENT; 25877c478bd9Sstevel@tonic-gate goto out; 25887c478bd9Sstevel@tonic-gate } 25897c478bd9Sstevel@tonic-gate if ((error = ufs_iaccess(VTOI(vp), IEXEC, cr)) != 0) { 25907c478bd9Sstevel@tonic-gate VN_RELE(vp); 25917c478bd9Sstevel@tonic-gate goto out; 25927c478bd9Sstevel@tonic-gate } 25937c478bd9Sstevel@tonic-gate 25947c478bd9Sstevel@tonic-gate *vpp = vp; 25957c478bd9Sstevel@tonic-gate return (0); 25967c478bd9Sstevel@tonic-gate } 25977c478bd9Sstevel@tonic-gate 25987c478bd9Sstevel@tonic-gate /* 25997c478bd9Sstevel@tonic-gate * Check for a null component, which we should treat as 26007c478bd9Sstevel@tonic-gate * looking at dvp from within it's parent, so we don't 26017c478bd9Sstevel@tonic-gate * need a call to ufs_iaccess(), as it has already been 26027c478bd9Sstevel@tonic-gate * done. 26037c478bd9Sstevel@tonic-gate */ 26047c478bd9Sstevel@tonic-gate if (nm[0] == 0) { 26057c478bd9Sstevel@tonic-gate VN_HOLD(dvp); 26067c478bd9Sstevel@tonic-gate error = 0; 26077c478bd9Sstevel@tonic-gate *vpp = dvp; 26087c478bd9Sstevel@tonic-gate goto out; 26097c478bd9Sstevel@tonic-gate } 26107c478bd9Sstevel@tonic-gate 26117c478bd9Sstevel@tonic-gate /* 26127c478bd9Sstevel@tonic-gate * Check for "." ie itself. this is a quick check and 26137c478bd9Sstevel@tonic-gate * avoids adding "." into the dnlc (which have been seen 26147c478bd9Sstevel@tonic-gate * to occupy >10% of the cache). 26157c478bd9Sstevel@tonic-gate */ 26167c478bd9Sstevel@tonic-gate if ((nm[0] == '.') && (nm[1] == 0)) { 26177c478bd9Sstevel@tonic-gate /* 26187c478bd9Sstevel@tonic-gate * Don't return without checking accessibility 26197c478bd9Sstevel@tonic-gate * of the directory. We only need the lock if 26207c478bd9Sstevel@tonic-gate * we are going to return it. 26217c478bd9Sstevel@tonic-gate */ 26227c478bd9Sstevel@tonic-gate if ((error = ufs_iaccess(ip, IEXEC, cr)) == 0) { 26237c478bd9Sstevel@tonic-gate VN_HOLD(dvp); 26247c478bd9Sstevel@tonic-gate *vpp = dvp; 26257c478bd9Sstevel@tonic-gate } 26267c478bd9Sstevel@tonic-gate goto out; 26277c478bd9Sstevel@tonic-gate } 26287c478bd9Sstevel@tonic-gate 26297c478bd9Sstevel@tonic-gate /* 26307c478bd9Sstevel@tonic-gate * Fast path: Check the directory name lookup cache. 26317c478bd9Sstevel@tonic-gate */ 26327c478bd9Sstevel@tonic-gate if (vp = dnlc_lookup(dvp, nm)) { 26337c478bd9Sstevel@tonic-gate /* 26347c478bd9Sstevel@tonic-gate * Check accessibility of directory. 26357c478bd9Sstevel@tonic-gate */ 26367c478bd9Sstevel@tonic-gate if ((error = ufs_iaccess(ip, IEXEC, cr)) != 0) { 26377c478bd9Sstevel@tonic-gate VN_RELE(vp); 26387c478bd9Sstevel@tonic-gate goto out; 26397c478bd9Sstevel@tonic-gate } 26407c478bd9Sstevel@tonic-gate if (vp == DNLC_NO_VNODE) { 26417c478bd9Sstevel@tonic-gate VN_RELE(vp); 26427c478bd9Sstevel@tonic-gate error = ENOENT; 26437c478bd9Sstevel@tonic-gate goto out; 26447c478bd9Sstevel@tonic-gate } 26457c478bd9Sstevel@tonic-gate xip = VTOI(vp); 26467c478bd9Sstevel@tonic-gate ulp = NULL; 26477c478bd9Sstevel@tonic-gate goto fastpath; 26487c478bd9Sstevel@tonic-gate } 26497c478bd9Sstevel@tonic-gate 26507c478bd9Sstevel@tonic-gate /* 26517c478bd9Sstevel@tonic-gate * Keep the idle queue from getting too long by 26527c478bd9Sstevel@tonic-gate * idling two inodes before attempting to allocate another. 26537c478bd9Sstevel@tonic-gate * This operation must be performed before entering 26547c478bd9Sstevel@tonic-gate * lockfs or a transaction. 26557c478bd9Sstevel@tonic-gate */ 26567c478bd9Sstevel@tonic-gate if (ufs_idle_q.uq_ne > ufs_idle_q.uq_hiwat) 26577c478bd9Sstevel@tonic-gate if ((curthread->t_flag & T_DONTBLOCK) == 0) { 26587c478bd9Sstevel@tonic-gate ins.in_lidles.value.ul += ufs_lookup_idle_count; 26597c478bd9Sstevel@tonic-gate ufs_idle_some(ufs_lookup_idle_count); 26607c478bd9Sstevel@tonic-gate } 26617c478bd9Sstevel@tonic-gate 26627c478bd9Sstevel@tonic-gate ufsvfsp = ip->i_ufsvfs; 26637c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_LOOKUP_MASK); 26647c478bd9Sstevel@tonic-gate if (error) 26657c478bd9Sstevel@tonic-gate goto out; 26667c478bd9Sstevel@tonic-gate 26677c478bd9Sstevel@tonic-gate error = ufs_dirlook(ip, nm, &xip, cr, 1); 26687c478bd9Sstevel@tonic-gate 26697c478bd9Sstevel@tonic-gate fastpath: 26707c478bd9Sstevel@tonic-gate if (error == 0) { 26717c478bd9Sstevel@tonic-gate ip = xip; 26727c478bd9Sstevel@tonic-gate *vpp = ITOV(ip); 26737c478bd9Sstevel@tonic-gate 26747c478bd9Sstevel@tonic-gate /* 26757c478bd9Sstevel@tonic-gate * If vnode is a device return special vnode instead. 26767c478bd9Sstevel@tonic-gate */ 26777c478bd9Sstevel@tonic-gate if (IS_DEVVP(*vpp)) { 26787c478bd9Sstevel@tonic-gate struct vnode *newvp; 26797c478bd9Sstevel@tonic-gate 26807c478bd9Sstevel@tonic-gate newvp = specvp(*vpp, (*vpp)->v_rdev, (*vpp)->v_type, 26817c478bd9Sstevel@tonic-gate cr); 26827c478bd9Sstevel@tonic-gate VN_RELE(*vpp); 26837c478bd9Sstevel@tonic-gate if (newvp == NULL) 26847c478bd9Sstevel@tonic-gate error = ENOSYS; 26857c478bd9Sstevel@tonic-gate else 26867c478bd9Sstevel@tonic-gate *vpp = newvp; 26877c478bd9Sstevel@tonic-gate } 26887c478bd9Sstevel@tonic-gate } 26897c478bd9Sstevel@tonic-gate if (ulp) { 26907c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 26917c478bd9Sstevel@tonic-gate } 26927c478bd9Sstevel@tonic-gate 26937c478bd9Sstevel@tonic-gate out: 26947c478bd9Sstevel@tonic-gate TRACE_3(TR_FAC_UFS, TR_UFS_LOOKUP_END, 26957c478bd9Sstevel@tonic-gate "ufs_lookup_end:dvp %p name %s error %d", vpp, nm, error); 26967c478bd9Sstevel@tonic-gate return (error); 26977c478bd9Sstevel@tonic-gate } 26987c478bd9Sstevel@tonic-gate 26997c478bd9Sstevel@tonic-gate static int 27007c478bd9Sstevel@tonic-gate ufs_create(struct vnode *dvp, char *name, struct vattr *vap, enum vcexcl excl, 27017c478bd9Sstevel@tonic-gate int mode, struct vnode **vpp, struct cred *cr, int flag) 27027c478bd9Sstevel@tonic-gate { 27037c478bd9Sstevel@tonic-gate struct inode *ip; 27047c478bd9Sstevel@tonic-gate struct inode *xip; 27057c478bd9Sstevel@tonic-gate struct inode *dip; 27067c478bd9Sstevel@tonic-gate struct vnode *xvp; 27077c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp; 27087c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 27097c478bd9Sstevel@tonic-gate int error; 27107c478bd9Sstevel@tonic-gate int issync; 27117c478bd9Sstevel@tonic-gate int truncflag; 27127c478bd9Sstevel@tonic-gate int trans_size; 27137c478bd9Sstevel@tonic-gate int noentry; 27147c478bd9Sstevel@tonic-gate int defer_dip_seq_update = 0; /* need to defer update of dip->i_seq */ 27157c478bd9Sstevel@tonic-gate int retry = 1; 27167c478bd9Sstevel@tonic-gate 27177c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_CREATE_START, 27187c478bd9Sstevel@tonic-gate "ufs_create_start:dvp %p", dvp); 27197c478bd9Sstevel@tonic-gate 27207c478bd9Sstevel@tonic-gate again: 27217c478bd9Sstevel@tonic-gate ip = VTOI(dvp); 27227c478bd9Sstevel@tonic-gate ufsvfsp = ip->i_ufsvfs; 27237c478bd9Sstevel@tonic-gate truncflag = 0; 27247c478bd9Sstevel@tonic-gate 27257c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_CREATE_MASK); 27267c478bd9Sstevel@tonic-gate if (error) 27277c478bd9Sstevel@tonic-gate goto out; 27287c478bd9Sstevel@tonic-gate 27297c478bd9Sstevel@tonic-gate if (ulp) { 27307c478bd9Sstevel@tonic-gate trans_size = (int)TOP_CREATE_SIZE(ip); 27317c478bd9Sstevel@tonic-gate TRANS_BEGIN_CSYNC(ufsvfsp, issync, TOP_CREATE, trans_size); 27327c478bd9Sstevel@tonic-gate } 27337c478bd9Sstevel@tonic-gate 27347c478bd9Sstevel@tonic-gate if ((vap->va_mode & VSVTX) && secpolicy_vnode_stky_modify(cr) != 0) 27357c478bd9Sstevel@tonic-gate vap->va_mode &= ~VSVTX; 27367c478bd9Sstevel@tonic-gate 27377c478bd9Sstevel@tonic-gate if (*name == '\0') { 27387c478bd9Sstevel@tonic-gate /* 27397c478bd9Sstevel@tonic-gate * Null component name refers to the directory itself. 27407c478bd9Sstevel@tonic-gate */ 27417c478bd9Sstevel@tonic-gate VN_HOLD(dvp); 27427c478bd9Sstevel@tonic-gate /* 27437c478bd9Sstevel@tonic-gate * Even though this is an error case, we need to grab the 27447c478bd9Sstevel@tonic-gate * quota lock since the error handling code below is common. 27457c478bd9Sstevel@tonic-gate */ 27467c478bd9Sstevel@tonic-gate rw_enter(&ufsvfsp->vfs_dqrwlock, RW_READER); 27477c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_WRITER); 27487c478bd9Sstevel@tonic-gate error = EEXIST; 27497c478bd9Sstevel@tonic-gate } else { 27507c478bd9Sstevel@tonic-gate xip = NULL; 27517c478bd9Sstevel@tonic-gate noentry = 0; 27527c478bd9Sstevel@tonic-gate rw_enter(&ip->i_rwlock, RW_WRITER); 27537c478bd9Sstevel@tonic-gate xvp = dnlc_lookup(dvp, name); 27547c478bd9Sstevel@tonic-gate if (xvp == DNLC_NO_VNODE) { 27557c478bd9Sstevel@tonic-gate noentry = 1; 27567c478bd9Sstevel@tonic-gate VN_RELE(xvp); 27577c478bd9Sstevel@tonic-gate xvp = NULL; 27587c478bd9Sstevel@tonic-gate } 27597c478bd9Sstevel@tonic-gate if (xvp) { 27607c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 27617c478bd9Sstevel@tonic-gate if (error = ufs_iaccess(ip, IEXEC, cr)) { 27627c478bd9Sstevel@tonic-gate VN_RELE(xvp); 27637c478bd9Sstevel@tonic-gate } else { 27647c478bd9Sstevel@tonic-gate error = EEXIST; 27657c478bd9Sstevel@tonic-gate xip = VTOI(xvp); 27667c478bd9Sstevel@tonic-gate } 27677c478bd9Sstevel@tonic-gate } else { 27687c478bd9Sstevel@tonic-gate /* 27697c478bd9Sstevel@tonic-gate * Suppress file system full message if we will retry 27707c478bd9Sstevel@tonic-gate */ 27717c478bd9Sstevel@tonic-gate error = ufs_direnter_cm(ip, name, DE_CREATE, 27727c478bd9Sstevel@tonic-gate vap, &xip, cr, 27737c478bd9Sstevel@tonic-gate (noentry | (retry ? IQUIET : 0))); 27747c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 27757c478bd9Sstevel@tonic-gate } 27767c478bd9Sstevel@tonic-gate ip = xip; 27777c478bd9Sstevel@tonic-gate if (ip != NULL) { 27787c478bd9Sstevel@tonic-gate rw_enter(&ufsvfsp->vfs_dqrwlock, RW_READER); 27797c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_WRITER); 27807c478bd9Sstevel@tonic-gate } 27817c478bd9Sstevel@tonic-gate } 27827c478bd9Sstevel@tonic-gate 27837c478bd9Sstevel@tonic-gate /* 27847c478bd9Sstevel@tonic-gate * If the file already exists and this is a non-exclusive create, 27857c478bd9Sstevel@tonic-gate * check permissions and allow access for non-directories. 27867c478bd9Sstevel@tonic-gate * Read-only create of an existing directory is also allowed. 27877c478bd9Sstevel@tonic-gate * We fail an exclusive create of anything which already exists. 27887c478bd9Sstevel@tonic-gate */ 27897c478bd9Sstevel@tonic-gate if (error == EEXIST) { 27907c478bd9Sstevel@tonic-gate dip = VTOI(dvp); 27917c478bd9Sstevel@tonic-gate if (excl == NONEXCL) { 27927c478bd9Sstevel@tonic-gate if ((((ip->i_mode & IFMT) == IFDIR) || 27937c478bd9Sstevel@tonic-gate ((ip->i_mode & IFMT) == IFATTRDIR)) && 27947c478bd9Sstevel@tonic-gate (mode & IWRITE)) 27957c478bd9Sstevel@tonic-gate error = EISDIR; 27967c478bd9Sstevel@tonic-gate else if (mode) 27977c478bd9Sstevel@tonic-gate error = ufs_iaccess(ip, mode, cr); 27987c478bd9Sstevel@tonic-gate else 27997c478bd9Sstevel@tonic-gate error = 0; 28007c478bd9Sstevel@tonic-gate } 28017c478bd9Sstevel@tonic-gate if (error) { 28027c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 28037c478bd9Sstevel@tonic-gate rw_exit(&ufsvfsp->vfs_dqrwlock); 28047c478bd9Sstevel@tonic-gate VN_RELE(ITOV(ip)); 28057c478bd9Sstevel@tonic-gate goto unlock; 28067c478bd9Sstevel@tonic-gate } 28077c478bd9Sstevel@tonic-gate /* 28087c478bd9Sstevel@tonic-gate * If the error EEXIST was set, then i_seq can not 28097c478bd9Sstevel@tonic-gate * have been updated. The sequence number interface 28107c478bd9Sstevel@tonic-gate * is defined such that a non-error VOP_CREATE must 28117c478bd9Sstevel@tonic-gate * increase the dir va_seq it by at least one. If we 28127c478bd9Sstevel@tonic-gate * have cleared the error, increase i_seq. Note that 28137c478bd9Sstevel@tonic-gate * we are increasing the dir i_seq and in rare cases 28147c478bd9Sstevel@tonic-gate * ip may actually be from the dvp, so we already have 28157c478bd9Sstevel@tonic-gate * the locks and it will not be subject to truncation. 28167c478bd9Sstevel@tonic-gate * In case we have to update i_seq of the parent 28177c478bd9Sstevel@tonic-gate * directory dip, we have to defer it till we have 28187c478bd9Sstevel@tonic-gate * released our locks on ip due to lock ordering requirements. 28197c478bd9Sstevel@tonic-gate */ 28207c478bd9Sstevel@tonic-gate if (ip != dip) 28217c478bd9Sstevel@tonic-gate defer_dip_seq_update = 1; 28227c478bd9Sstevel@tonic-gate else 28237c478bd9Sstevel@tonic-gate ip->i_seq++; 28247c478bd9Sstevel@tonic-gate 28257c478bd9Sstevel@tonic-gate if (((ip->i_mode & IFMT) == IFREG) && 28267c478bd9Sstevel@tonic-gate (vap->va_mask & AT_SIZE) && vap->va_size == 0) { 28277c478bd9Sstevel@tonic-gate /* 28287c478bd9Sstevel@tonic-gate * Truncate regular files, if requested by caller. 28297c478bd9Sstevel@tonic-gate * Grab i_rwlock to make sure no one else is 28307c478bd9Sstevel@tonic-gate * currently writing to the file (we promised 28317c478bd9Sstevel@tonic-gate * bmap we would do this). 28327c478bd9Sstevel@tonic-gate * Must get the locks in the correct order. 28337c478bd9Sstevel@tonic-gate */ 28347c478bd9Sstevel@tonic-gate if (ip->i_size == 0) { 28357c478bd9Sstevel@tonic-gate ip->i_flag |= ICHG | IUPD; 28367c478bd9Sstevel@tonic-gate ip->i_seq++; 28377c478bd9Sstevel@tonic-gate TRANS_INODE(ufsvfsp, ip); 28387c478bd9Sstevel@tonic-gate } else { 28397c478bd9Sstevel@tonic-gate /* 28407c478bd9Sstevel@tonic-gate * Large Files: Why this check here? 28417c478bd9Sstevel@tonic-gate * Though we do it in vn_create() we really 28427c478bd9Sstevel@tonic-gate * want to guarantee that we do not destroy 28437c478bd9Sstevel@tonic-gate * Large file data by atomically checking 28447c478bd9Sstevel@tonic-gate * the size while holding the contents 28457c478bd9Sstevel@tonic-gate * lock. 28467c478bd9Sstevel@tonic-gate */ 28477c478bd9Sstevel@tonic-gate if (flag && !(flag & FOFFMAX) && 28487c478bd9Sstevel@tonic-gate ((ip->i_mode & IFMT) == IFREG) && 28497c478bd9Sstevel@tonic-gate (ip->i_size > (offset_t)MAXOFF32_T)) { 28507c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 28517c478bd9Sstevel@tonic-gate rw_exit(&ufsvfsp->vfs_dqrwlock); 28527c478bd9Sstevel@tonic-gate error = EOVERFLOW; 28537c478bd9Sstevel@tonic-gate goto unlock; 28547c478bd9Sstevel@tonic-gate } 28557c478bd9Sstevel@tonic-gate if (TRANS_ISTRANS(ufsvfsp)) 28567c478bd9Sstevel@tonic-gate truncflag++; 28577c478bd9Sstevel@tonic-gate else { 28587c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 28597c478bd9Sstevel@tonic-gate rw_exit(&ufsvfsp->vfs_dqrwlock); 28607c478bd9Sstevel@tonic-gate rw_enter(&ip->i_rwlock, RW_WRITER); 28617c478bd9Sstevel@tonic-gate rw_enter(&ufsvfsp->vfs_dqrwlock, 28627c478bd9Sstevel@tonic-gate RW_READER); 28637c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_WRITER); 28647c478bd9Sstevel@tonic-gate (void) ufs_itrunc(ip, (u_offset_t)0, 0, 28657c478bd9Sstevel@tonic-gate cr); 28667c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 28677c478bd9Sstevel@tonic-gate } 28687c478bd9Sstevel@tonic-gate } 28697c478bd9Sstevel@tonic-gate } 28707c478bd9Sstevel@tonic-gate } 28717c478bd9Sstevel@tonic-gate 28727c478bd9Sstevel@tonic-gate if (error) { 28737c478bd9Sstevel@tonic-gate if (ip != NULL) { 28747c478bd9Sstevel@tonic-gate rw_exit(&ufsvfsp->vfs_dqrwlock); 28757c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 28767c478bd9Sstevel@tonic-gate } 28777c478bd9Sstevel@tonic-gate goto unlock; 28787c478bd9Sstevel@tonic-gate } 28797c478bd9Sstevel@tonic-gate 28807c478bd9Sstevel@tonic-gate *vpp = ITOV(ip); 28817c478bd9Sstevel@tonic-gate ITIMES(ip); 28827c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 28837c478bd9Sstevel@tonic-gate rw_exit(&ufsvfsp->vfs_dqrwlock); 28847c478bd9Sstevel@tonic-gate 28857c478bd9Sstevel@tonic-gate /* 28867c478bd9Sstevel@tonic-gate * If vnode is a device return special vnode instead. 28877c478bd9Sstevel@tonic-gate */ 28887c478bd9Sstevel@tonic-gate if (!error && IS_DEVVP(*vpp)) { 28897c478bd9Sstevel@tonic-gate struct vnode *newvp; 28907c478bd9Sstevel@tonic-gate 28917c478bd9Sstevel@tonic-gate newvp = specvp(*vpp, (*vpp)->v_rdev, (*vpp)->v_type, cr); 28927c478bd9Sstevel@tonic-gate VN_RELE(*vpp); 28937c478bd9Sstevel@tonic-gate if (newvp == NULL) { 28947c478bd9Sstevel@tonic-gate error = ENOSYS; 28957c478bd9Sstevel@tonic-gate goto unlock; 28967c478bd9Sstevel@tonic-gate } 28977c478bd9Sstevel@tonic-gate truncflag = 0; 28987c478bd9Sstevel@tonic-gate *vpp = newvp; 28997c478bd9Sstevel@tonic-gate } 29007c478bd9Sstevel@tonic-gate unlock: 29017c478bd9Sstevel@tonic-gate 29027c478bd9Sstevel@tonic-gate /* 29037c478bd9Sstevel@tonic-gate * Do the deferred update of the parent directory's sequence 29047c478bd9Sstevel@tonic-gate * number now. 29057c478bd9Sstevel@tonic-gate */ 29067c478bd9Sstevel@tonic-gate if (defer_dip_seq_update == 1) { 29077c478bd9Sstevel@tonic-gate rw_enter(&dip->i_contents, RW_READER); 29087c478bd9Sstevel@tonic-gate mutex_enter(&dip->i_tlock); 29097c478bd9Sstevel@tonic-gate dip->i_seq++; 29107c478bd9Sstevel@tonic-gate mutex_exit(&dip->i_tlock); 29117c478bd9Sstevel@tonic-gate rw_exit(&dip->i_contents); 29127c478bd9Sstevel@tonic-gate } 29137c478bd9Sstevel@tonic-gate 29147c478bd9Sstevel@tonic-gate if (ulp) { 29157c478bd9Sstevel@tonic-gate int terr = 0; 29167c478bd9Sstevel@tonic-gate 29177c478bd9Sstevel@tonic-gate TRANS_END_CSYNC(ufsvfsp, terr, issync, TOP_CREATE, 29187c478bd9Sstevel@tonic-gate trans_size); 29197c478bd9Sstevel@tonic-gate 29207c478bd9Sstevel@tonic-gate /* 29217c478bd9Sstevel@tonic-gate * If we haven't had a more interesting failure 29227c478bd9Sstevel@tonic-gate * already, then anything that might've happened 29237c478bd9Sstevel@tonic-gate * here should be reported. 29247c478bd9Sstevel@tonic-gate */ 29257c478bd9Sstevel@tonic-gate if (error == 0) 29267c478bd9Sstevel@tonic-gate error = terr; 29277c478bd9Sstevel@tonic-gate } 29287c478bd9Sstevel@tonic-gate 29297c478bd9Sstevel@tonic-gate if (!error && truncflag) { 29307c478bd9Sstevel@tonic-gate rw_enter(&ip->i_rwlock, RW_WRITER); 29317c478bd9Sstevel@tonic-gate (void) TRANS_ITRUNC(ip, (u_offset_t)0, 0, cr); 29327c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 29337c478bd9Sstevel@tonic-gate } 29347c478bd9Sstevel@tonic-gate 29357c478bd9Sstevel@tonic-gate if (ulp) 29367c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 29377c478bd9Sstevel@tonic-gate 29387c478bd9Sstevel@tonic-gate /* 29397c478bd9Sstevel@tonic-gate * If no inodes available, try to free one up out of the 29407c478bd9Sstevel@tonic-gate * pending delete queue. 29417c478bd9Sstevel@tonic-gate */ 29427c478bd9Sstevel@tonic-gate if ((error == ENOSPC) && retry && TRANS_ISTRANS(ufsvfsp)) { 29437c478bd9Sstevel@tonic-gate ufs_delete_drain_wait(ufsvfsp, 1); 29447c478bd9Sstevel@tonic-gate retry = 0; 29457c478bd9Sstevel@tonic-gate goto again; 29467c478bd9Sstevel@tonic-gate } 29477c478bd9Sstevel@tonic-gate 29487c478bd9Sstevel@tonic-gate out: 29497c478bd9Sstevel@tonic-gate TRACE_3(TR_FAC_UFS, TR_UFS_CREATE_END, 29507c478bd9Sstevel@tonic-gate "ufs_create_end:dvp %p name %s error %d", vpp, name, error); 29517c478bd9Sstevel@tonic-gate return (error); 29527c478bd9Sstevel@tonic-gate } 29537c478bd9Sstevel@tonic-gate 29547c478bd9Sstevel@tonic-gate extern int ufs_idle_max; 29557c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 29567c478bd9Sstevel@tonic-gate static int 29577c478bd9Sstevel@tonic-gate ufs_remove(struct vnode *vp, char *nm, struct cred *cr) 29587c478bd9Sstevel@tonic-gate { 29597c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 29607c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp = ip->i_ufsvfs; 29617c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 29627c478bd9Sstevel@tonic-gate vnode_t *rmvp = NULL; /* Vnode corresponding to name being removed */ 29637c478bd9Sstevel@tonic-gate int error; 29647c478bd9Sstevel@tonic-gate int issync; 29657c478bd9Sstevel@tonic-gate int trans_size; 29667c478bd9Sstevel@tonic-gate 29677c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_REMOVE_START, 29687c478bd9Sstevel@tonic-gate "ufs_remove_start:vp %p", vp); 29697c478bd9Sstevel@tonic-gate 29707c478bd9Sstevel@tonic-gate /* 29717c478bd9Sstevel@tonic-gate * don't let the delete queue get too long 29727c478bd9Sstevel@tonic-gate */ 29737c478bd9Sstevel@tonic-gate if (ufsvfsp == NULL) { 29747c478bd9Sstevel@tonic-gate error = EIO; 29757c478bd9Sstevel@tonic-gate goto out; 29767c478bd9Sstevel@tonic-gate } 29777c478bd9Sstevel@tonic-gate if (ufsvfsp->vfs_delete.uq_ne > ufs_idle_max) 29787c478bd9Sstevel@tonic-gate ufs_delete_drain(vp->v_vfsp, 1, 1); 29797c478bd9Sstevel@tonic-gate 29807c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_REMOVE_MASK); 29817c478bd9Sstevel@tonic-gate if (error) 29827c478bd9Sstevel@tonic-gate goto out; 29837c478bd9Sstevel@tonic-gate 29847c478bd9Sstevel@tonic-gate if (ulp) 29857c478bd9Sstevel@tonic-gate TRANS_BEGIN_CSYNC(ufsvfsp, issync, TOP_REMOVE, 29867c478bd9Sstevel@tonic-gate trans_size = (int)TOP_REMOVE_SIZE(VTOI(vp))); 29877c478bd9Sstevel@tonic-gate 29887c478bd9Sstevel@tonic-gate rw_enter(&ip->i_rwlock, RW_WRITER); 29897c478bd9Sstevel@tonic-gate error = ufs_dirremove(ip, nm, (struct inode *)0, (struct vnode *)0, 29907c478bd9Sstevel@tonic-gate DR_REMOVE, cr, &rmvp); 29917c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 29927c478bd9Sstevel@tonic-gate 29937c478bd9Sstevel@tonic-gate if (ulp) { 29947c478bd9Sstevel@tonic-gate TRANS_END_CSYNC(ufsvfsp, error, issync, TOP_REMOVE, trans_size); 29957c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 29967c478bd9Sstevel@tonic-gate } 29977c478bd9Sstevel@tonic-gate 29987c478bd9Sstevel@tonic-gate /* 29997c478bd9Sstevel@tonic-gate * This must be called after the remove transaction is closed. 30007c478bd9Sstevel@tonic-gate */ 30017c478bd9Sstevel@tonic-gate if (rmvp != NULL) { 30027c478bd9Sstevel@tonic-gate /* Only send the event if there were no errors */ 30037c478bd9Sstevel@tonic-gate if (error == 0) 30047c478bd9Sstevel@tonic-gate vnevent_remove(rmvp); 30057c478bd9Sstevel@tonic-gate VN_RELE(rmvp); 30067c478bd9Sstevel@tonic-gate } 30077c478bd9Sstevel@tonic-gate out: 30087c478bd9Sstevel@tonic-gate TRACE_3(TR_FAC_UFS, TR_UFS_REMOVE_END, 30097c478bd9Sstevel@tonic-gate "ufs_remove_end:vp %p name %s error %d", vp, nm, error); 30107c478bd9Sstevel@tonic-gate return (error); 30117c478bd9Sstevel@tonic-gate } 30127c478bd9Sstevel@tonic-gate 30137c478bd9Sstevel@tonic-gate /* 30147c478bd9Sstevel@tonic-gate * Link a file or a directory. Only privileged processes are allowed to 30157c478bd9Sstevel@tonic-gate * make links to directories. 30167c478bd9Sstevel@tonic-gate */ 30177c478bd9Sstevel@tonic-gate static int 30187c478bd9Sstevel@tonic-gate ufs_link(struct vnode *tdvp, struct vnode *svp, char *tnm, struct cred *cr) 30197c478bd9Sstevel@tonic-gate { 30207c478bd9Sstevel@tonic-gate struct inode *sip; 30217c478bd9Sstevel@tonic-gate struct inode *tdp = VTOI(tdvp); 30227c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp = tdp->i_ufsvfs; 30237c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 30247c478bd9Sstevel@tonic-gate struct vnode *realvp; 30257c478bd9Sstevel@tonic-gate int error; 30267c478bd9Sstevel@tonic-gate int issync; 30277c478bd9Sstevel@tonic-gate int trans_size; 30287c478bd9Sstevel@tonic-gate int isdev; 30297c478bd9Sstevel@tonic-gate 30307c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_LINK_START, 30317c478bd9Sstevel@tonic-gate "ufs_link_start:tdvp %p", tdvp); 30327c478bd9Sstevel@tonic-gate 30337c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_LINK_MASK); 30347c478bd9Sstevel@tonic-gate if (error) 30357c478bd9Sstevel@tonic-gate goto out; 30367c478bd9Sstevel@tonic-gate 30377c478bd9Sstevel@tonic-gate if (ulp) 30387c478bd9Sstevel@tonic-gate TRANS_BEGIN_CSYNC(ufsvfsp, issync, TOP_LINK, 30397c478bd9Sstevel@tonic-gate trans_size = (int)TOP_LINK_SIZE(VTOI(tdvp))); 30407c478bd9Sstevel@tonic-gate 30417c478bd9Sstevel@tonic-gate if (VOP_REALVP(svp, &realvp) == 0) 30427c478bd9Sstevel@tonic-gate svp = realvp; 30437c478bd9Sstevel@tonic-gate 30447c478bd9Sstevel@tonic-gate /* 30457c478bd9Sstevel@tonic-gate * Make sure link for extended attributes is valid 30467c478bd9Sstevel@tonic-gate * We only support hard linking of attr in ATTRDIR to ATTRDIR 30477c478bd9Sstevel@tonic-gate * 30487c478bd9Sstevel@tonic-gate * Make certain we don't attempt to look at a device node as 30497c478bd9Sstevel@tonic-gate * a ufs inode. 30507c478bd9Sstevel@tonic-gate */ 30517c478bd9Sstevel@tonic-gate 30527c478bd9Sstevel@tonic-gate isdev = IS_DEVVP(svp); 30537c478bd9Sstevel@tonic-gate if (((isdev == 0) && ((VTOI(svp)->i_cflags & IXATTR) == 0) && 30547c478bd9Sstevel@tonic-gate ((tdp->i_mode & IFMT) == IFATTRDIR)) || 30557c478bd9Sstevel@tonic-gate ((isdev == 0) && (VTOI(svp)->i_cflags & IXATTR) && 30567c478bd9Sstevel@tonic-gate ((tdp->i_mode & IFMT) == IFDIR))) { 30577c478bd9Sstevel@tonic-gate error = EINVAL; 30587c478bd9Sstevel@tonic-gate goto unlock; 30597c478bd9Sstevel@tonic-gate } 30607c478bd9Sstevel@tonic-gate 30617c478bd9Sstevel@tonic-gate sip = VTOI(svp); 30627c478bd9Sstevel@tonic-gate if ((svp->v_type == VDIR && 30637c478bd9Sstevel@tonic-gate secpolicy_fs_linkdir(cr, ufsvfsp->vfs_vfs) != 0) || 30647c478bd9Sstevel@tonic-gate (sip->i_uid != crgetuid(cr) && secpolicy_basic_link(cr) != 0)) { 30657c478bd9Sstevel@tonic-gate error = EPERM; 30667c478bd9Sstevel@tonic-gate goto unlock; 30677c478bd9Sstevel@tonic-gate } 30687c478bd9Sstevel@tonic-gate rw_enter(&tdp->i_rwlock, RW_WRITER); 30697c478bd9Sstevel@tonic-gate error = ufs_direnter_lr(tdp, tnm, DE_LINK, (struct inode *)0, 30707c478bd9Sstevel@tonic-gate sip, cr, NULL); 30717c478bd9Sstevel@tonic-gate rw_exit(&tdp->i_rwlock); 30727c478bd9Sstevel@tonic-gate 30737c478bd9Sstevel@tonic-gate unlock: 30747c478bd9Sstevel@tonic-gate if (ulp) { 30757c478bd9Sstevel@tonic-gate TRANS_END_CSYNC(ufsvfsp, error, issync, TOP_LINK, trans_size); 30767c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 30777c478bd9Sstevel@tonic-gate } 30787c478bd9Sstevel@tonic-gate out: 30797c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_LINK_END, 30807c478bd9Sstevel@tonic-gate "ufs_link_end:tdvp %p error %d", tdvp, error); 30817c478bd9Sstevel@tonic-gate return (error); 30827c478bd9Sstevel@tonic-gate } 30837c478bd9Sstevel@tonic-gate 30847c478bd9Sstevel@tonic-gate uint64_t ufs_rename_retry_cnt; 30857c478bd9Sstevel@tonic-gate uint64_t ufs_rename_upgrade_retry_cnt; 30867c478bd9Sstevel@tonic-gate uint64_t ufs_rename_dircheck_retry_cnt; 30877c478bd9Sstevel@tonic-gate clock_t ufs_rename_backoff_delay = 1; 30887c478bd9Sstevel@tonic-gate 30897c478bd9Sstevel@tonic-gate /* 30907c478bd9Sstevel@tonic-gate * Rename a file or directory. 30917c478bd9Sstevel@tonic-gate * We are given the vnode and entry string of the source and the 30927c478bd9Sstevel@tonic-gate * vnode and entry string of the place we want to move the source 30937c478bd9Sstevel@tonic-gate * to (the target). The essential operation is: 30947c478bd9Sstevel@tonic-gate * unlink(target); 30957c478bd9Sstevel@tonic-gate * link(source, target); 30967c478bd9Sstevel@tonic-gate * unlink(source); 30977c478bd9Sstevel@tonic-gate * but "atomically". Can't do full commit without saving state in 30987c478bd9Sstevel@tonic-gate * the inode on disk, which isn't feasible at this time. Best we 30997c478bd9Sstevel@tonic-gate * can do is always guarantee that the TARGET exists. 31007c478bd9Sstevel@tonic-gate */ 31017c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 31027c478bd9Sstevel@tonic-gate static int 31037c478bd9Sstevel@tonic-gate ufs_rename( 31047c478bd9Sstevel@tonic-gate struct vnode *sdvp, /* old (source) parent vnode */ 31057c478bd9Sstevel@tonic-gate char *snm, /* old (source) entry name */ 31067c478bd9Sstevel@tonic-gate struct vnode *tdvp, /* new (target) parent vnode */ 31077c478bd9Sstevel@tonic-gate char *tnm, /* new (target) entry name */ 31087c478bd9Sstevel@tonic-gate struct cred *cr) 31097c478bd9Sstevel@tonic-gate { 31107c478bd9Sstevel@tonic-gate struct inode *sip = NULL; /* source inode */ 31117c478bd9Sstevel@tonic-gate struct inode *sdp; /* old (source) parent inode */ 31127c478bd9Sstevel@tonic-gate struct inode *tdp; /* new (target) parent inode */ 31137c478bd9Sstevel@tonic-gate struct vnode *tvp = NULL; /* target vnode, if it exists */ 31147c478bd9Sstevel@tonic-gate struct vnode *realvp; 31157c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp; 31167c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 31177c478bd9Sstevel@tonic-gate int error; 31187c478bd9Sstevel@tonic-gate int issync; 31197c478bd9Sstevel@tonic-gate int trans_size; 31207c478bd9Sstevel@tonic-gate 31217c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_RENAME_START, 31227c478bd9Sstevel@tonic-gate "ufs_rename_start:sdvp %p", sdvp); 31237c478bd9Sstevel@tonic-gate 31247c478bd9Sstevel@tonic-gate 31257c478bd9Sstevel@tonic-gate sdp = VTOI(sdvp); 31267c478bd9Sstevel@tonic-gate ufsvfsp = sdp->i_ufsvfs; 31277c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_RENAME_MASK); 31287c478bd9Sstevel@tonic-gate if (error) 31297c478bd9Sstevel@tonic-gate goto out; 31307c478bd9Sstevel@tonic-gate 31317c478bd9Sstevel@tonic-gate if (ulp) 31327c478bd9Sstevel@tonic-gate TRANS_BEGIN_CSYNC(ufsvfsp, issync, TOP_RENAME, 31337c478bd9Sstevel@tonic-gate trans_size = (int)TOP_RENAME_SIZE(sdp)); 31347c478bd9Sstevel@tonic-gate 31357c478bd9Sstevel@tonic-gate if (VOP_REALVP(tdvp, &realvp) == 0) 31367c478bd9Sstevel@tonic-gate tdvp = realvp; 31377c478bd9Sstevel@tonic-gate 31387c478bd9Sstevel@tonic-gate tdp = VTOI(tdvp); 31397c478bd9Sstevel@tonic-gate 31407c478bd9Sstevel@tonic-gate /* 31417c478bd9Sstevel@tonic-gate * We only allow renaming of attributes from ATTRDIR to ATTRDIR. 31427c478bd9Sstevel@tonic-gate */ 31437c478bd9Sstevel@tonic-gate if ((tdp->i_mode & IFMT) != (sdp->i_mode & IFMT)) { 31447c478bd9Sstevel@tonic-gate error = EINVAL; 31457c478bd9Sstevel@tonic-gate goto unlock; 31467c478bd9Sstevel@tonic-gate } 31477c478bd9Sstevel@tonic-gate 31487c478bd9Sstevel@tonic-gate /* 31497c478bd9Sstevel@tonic-gate * Look up inode of file we're supposed to rename. 31507c478bd9Sstevel@tonic-gate */ 31517c478bd9Sstevel@tonic-gate if (error = ufs_dirlook(sdp, snm, &sip, cr, 0)) { 31527c478bd9Sstevel@tonic-gate goto unlock; 31537c478bd9Sstevel@tonic-gate } 31547c478bd9Sstevel@tonic-gate 31557c478bd9Sstevel@tonic-gate /* 31567c478bd9Sstevel@tonic-gate * Lock both the source and target directories (they may be 31577c478bd9Sstevel@tonic-gate * the same) to provide the atomicity semantics that was 31587c478bd9Sstevel@tonic-gate * previously provided by the per file system vfs_rename_lock 31597c478bd9Sstevel@tonic-gate * 31607c478bd9Sstevel@tonic-gate * with vfs_rename_lock removed to allow simultaneous renames 31617c478bd9Sstevel@tonic-gate * within a file system, ufs_dircheckpath can deadlock while 31627c478bd9Sstevel@tonic-gate * traversing back to ensure that source is not a parent directory 31637c478bd9Sstevel@tonic-gate * of target parent directory. This is because we get into 31647c478bd9Sstevel@tonic-gate * ufs_dircheckpath with the sdp and tdp locks held as RW_WRITER. 31657c478bd9Sstevel@tonic-gate * If the tdp and sdp of the simultaneous renames happen to be 31667c478bd9Sstevel@tonic-gate * in the path of each other, it can lead to a deadlock. This 31677c478bd9Sstevel@tonic-gate * can be avoided by getting the locks as RW_READER here and then 31687c478bd9Sstevel@tonic-gate * upgrading to RW_WRITER after completing the ufs_dircheckpath. 31697c478bd9Sstevel@tonic-gate */ 31707c478bd9Sstevel@tonic-gate retry: 31717c478bd9Sstevel@tonic-gate rw_enter(&tdp->i_rwlock, RW_READER); 31727c478bd9Sstevel@tonic-gate if (tdp != sdp) { 31737c478bd9Sstevel@tonic-gate /* 31747c478bd9Sstevel@tonic-gate * We're locking 2 peer level locks, so must use tryenter 31757c478bd9Sstevel@tonic-gate * on the 2nd to avoid deadlocks that would occur 31767c478bd9Sstevel@tonic-gate * if we renamed a->b and b->a concurrently. 31777c478bd9Sstevel@tonic-gate */ 31787c478bd9Sstevel@tonic-gate if (!rw_tryenter(&sdp->i_rwlock, RW_READER)) { 31797c478bd9Sstevel@tonic-gate /* 31807c478bd9Sstevel@tonic-gate * Reverse the lock grabs in case we have heavy 31817c478bd9Sstevel@tonic-gate * contention on the 2nd lock. 31827c478bd9Sstevel@tonic-gate */ 31837c478bd9Sstevel@tonic-gate rw_exit(&tdp->i_rwlock); 31847c478bd9Sstevel@tonic-gate rw_enter(&sdp->i_rwlock, RW_READER); 31857c478bd9Sstevel@tonic-gate if (!rw_tryenter(&tdp->i_rwlock, RW_READER)) { 31867c478bd9Sstevel@tonic-gate ufs_rename_retry_cnt++; 31877c478bd9Sstevel@tonic-gate rw_exit(&sdp->i_rwlock); 31887c478bd9Sstevel@tonic-gate goto retry; 31897c478bd9Sstevel@tonic-gate } 31907c478bd9Sstevel@tonic-gate } 31917c478bd9Sstevel@tonic-gate } 31927c478bd9Sstevel@tonic-gate 31937c478bd9Sstevel@tonic-gate if (sip == tdp) { 31947c478bd9Sstevel@tonic-gate error = EINVAL; 31957c478bd9Sstevel@tonic-gate goto errout; 31967c478bd9Sstevel@tonic-gate } 31977c478bd9Sstevel@tonic-gate /* 31987c478bd9Sstevel@tonic-gate * Make sure we can delete the source entry. This requires 31997c478bd9Sstevel@tonic-gate * write permission on the containing directory. 32007c478bd9Sstevel@tonic-gate * Check for sticky directories. 32017c478bd9Sstevel@tonic-gate */ 32027c478bd9Sstevel@tonic-gate rw_enter(&sdp->i_contents, RW_READER); 32037c478bd9Sstevel@tonic-gate rw_enter(&sip->i_contents, RW_READER); 32047c478bd9Sstevel@tonic-gate if ((error = ufs_iaccess(sdp, IWRITE, cr)) != 0 || 32057c478bd9Sstevel@tonic-gate (error = ufs_sticky_remove_access(sdp, sip, cr)) != 0) { 32067c478bd9Sstevel@tonic-gate rw_exit(&sip->i_contents); 32077c478bd9Sstevel@tonic-gate rw_exit(&sdp->i_contents); 32087c478bd9Sstevel@tonic-gate goto errout; 32097c478bd9Sstevel@tonic-gate } 32107c478bd9Sstevel@tonic-gate 32117c478bd9Sstevel@tonic-gate /* 32127c478bd9Sstevel@tonic-gate * If this is a rename of a directory and the parent is 32137c478bd9Sstevel@tonic-gate * different (".." must be changed), then the source 32147c478bd9Sstevel@tonic-gate * directory must not be in the directory hierarchy 32157c478bd9Sstevel@tonic-gate * above the target, as this would orphan everything 32167c478bd9Sstevel@tonic-gate * below the source directory. Also the user must have 32177c478bd9Sstevel@tonic-gate * write permission in the source so as to be able to 32187c478bd9Sstevel@tonic-gate * change "..". 32197c478bd9Sstevel@tonic-gate */ 32207c478bd9Sstevel@tonic-gate if ((((sip->i_mode & IFMT) == IFDIR) || 32217c478bd9Sstevel@tonic-gate ((sip->i_mode & IFMT) == IFATTRDIR)) && sdp != tdp) { 32227c478bd9Sstevel@tonic-gate ino_t inum; 32237c478bd9Sstevel@tonic-gate 32247c478bd9Sstevel@tonic-gate if ((error = ufs_iaccess(sip, IWRITE, cr))) { 32257c478bd9Sstevel@tonic-gate rw_exit(&sip->i_contents); 32267c478bd9Sstevel@tonic-gate rw_exit(&sdp->i_contents); 32277c478bd9Sstevel@tonic-gate goto errout; 32287c478bd9Sstevel@tonic-gate } 32297c478bd9Sstevel@tonic-gate inum = sip->i_number; 32307c478bd9Sstevel@tonic-gate rw_exit(&sip->i_contents); 32317c478bd9Sstevel@tonic-gate rw_exit(&sdp->i_contents); 32327c478bd9Sstevel@tonic-gate if ((error = ufs_dircheckpath(inum, tdp, sdp, cr))) { 32337c478bd9Sstevel@tonic-gate /* 32347c478bd9Sstevel@tonic-gate * If we got EAGAIN ufs_dircheckpath detected a 32357c478bd9Sstevel@tonic-gate * potential deadlock and backed out. We need 32367c478bd9Sstevel@tonic-gate * to retry the operation since sdp and tdp have 32377c478bd9Sstevel@tonic-gate * to be released to avoid the deadlock. 32387c478bd9Sstevel@tonic-gate */ 32397c478bd9Sstevel@tonic-gate if (error == EAGAIN) { 32407c478bd9Sstevel@tonic-gate rw_exit(&tdp->i_rwlock); 32417c478bd9Sstevel@tonic-gate if (tdp != sdp) 32427c478bd9Sstevel@tonic-gate rw_exit(&sdp->i_rwlock); 32437c478bd9Sstevel@tonic-gate delay(ufs_rename_backoff_delay); 32447c478bd9Sstevel@tonic-gate ufs_rename_dircheck_retry_cnt++; 32457c478bd9Sstevel@tonic-gate goto retry; 32467c478bd9Sstevel@tonic-gate } 32477c478bd9Sstevel@tonic-gate goto errout; 32487c478bd9Sstevel@tonic-gate } 32497c478bd9Sstevel@tonic-gate } else { 32507c478bd9Sstevel@tonic-gate rw_exit(&sip->i_contents); 32517c478bd9Sstevel@tonic-gate rw_exit(&sdp->i_contents); 32527c478bd9Sstevel@tonic-gate } 32537c478bd9Sstevel@tonic-gate 32547c478bd9Sstevel@tonic-gate 32557c478bd9Sstevel@tonic-gate /* 32567c478bd9Sstevel@tonic-gate * Check for renaming '.' or '..' or alias of '.' 32577c478bd9Sstevel@tonic-gate */ 32587c478bd9Sstevel@tonic-gate if (strcmp(snm, ".") == 0 || strcmp(snm, "..") == 0 || sdp == sip) { 32597c478bd9Sstevel@tonic-gate error = EINVAL; 32607c478bd9Sstevel@tonic-gate goto errout; 32617c478bd9Sstevel@tonic-gate } 32627c478bd9Sstevel@tonic-gate 32637c478bd9Sstevel@tonic-gate /* 32647c478bd9Sstevel@tonic-gate * Simultaneous renames can deadlock in ufs_dircheckpath since it 32657c478bd9Sstevel@tonic-gate * tries to traverse back the file tree with both tdp and sdp held 32667c478bd9Sstevel@tonic-gate * as RW_WRITER. To avoid that we have to hold the tdp and sdp locks 32677c478bd9Sstevel@tonic-gate * as RW_READERS till ufs_dircheckpath is done. 32687c478bd9Sstevel@tonic-gate * Now that ufs_dircheckpath is done with, we can upgrade the locks 32697c478bd9Sstevel@tonic-gate * to RW_WRITER. 32707c478bd9Sstevel@tonic-gate */ 32717c478bd9Sstevel@tonic-gate if (!rw_tryupgrade(&tdp->i_rwlock)) { 32727c478bd9Sstevel@tonic-gate /* 32737c478bd9Sstevel@tonic-gate * The upgrade failed. We got to give away the lock 32747c478bd9Sstevel@tonic-gate * as to avoid deadlocking with someone else who is 32757c478bd9Sstevel@tonic-gate * waiting for writer lock. With the lock gone, we 32767c478bd9Sstevel@tonic-gate * cannot be sure the checks done above will hold 32777c478bd9Sstevel@tonic-gate * good when we eventually get them back as writer. 32787c478bd9Sstevel@tonic-gate * So if we can't upgrade we drop the locks and retry 32797c478bd9Sstevel@tonic-gate * everything again. 32807c478bd9Sstevel@tonic-gate */ 32817c478bd9Sstevel@tonic-gate rw_exit(&tdp->i_rwlock); 32827c478bd9Sstevel@tonic-gate if (tdp != sdp) 32837c478bd9Sstevel@tonic-gate rw_exit(&sdp->i_rwlock); 32847c478bd9Sstevel@tonic-gate delay(ufs_rename_backoff_delay); 32857c478bd9Sstevel@tonic-gate ufs_rename_upgrade_retry_cnt++; 32867c478bd9Sstevel@tonic-gate goto retry; 32877c478bd9Sstevel@tonic-gate } 32887c478bd9Sstevel@tonic-gate if (tdp != sdp) { 32897c478bd9Sstevel@tonic-gate if (!rw_tryupgrade(&sdp->i_rwlock)) { 32907c478bd9Sstevel@tonic-gate /* 32917c478bd9Sstevel@tonic-gate * The upgrade failed. We got to give away the lock 32927c478bd9Sstevel@tonic-gate * as to avoid deadlocking with someone else who is 32937c478bd9Sstevel@tonic-gate * waiting for writer lock. With the lock gone, we 32947c478bd9Sstevel@tonic-gate * cannot be sure the checks done above will hold 32957c478bd9Sstevel@tonic-gate * good when we eventually get them back as writer. 32967c478bd9Sstevel@tonic-gate * So if we can't upgrade we drop the locks and retry 32977c478bd9Sstevel@tonic-gate * everything again. 32987c478bd9Sstevel@tonic-gate */ 32997c478bd9Sstevel@tonic-gate rw_exit(&tdp->i_rwlock); 33007c478bd9Sstevel@tonic-gate rw_exit(&sdp->i_rwlock); 33017c478bd9Sstevel@tonic-gate delay(ufs_rename_backoff_delay); 33027c478bd9Sstevel@tonic-gate ufs_rename_upgrade_retry_cnt++; 33037c478bd9Sstevel@tonic-gate goto retry; 33047c478bd9Sstevel@tonic-gate } 33057c478bd9Sstevel@tonic-gate } 33067c478bd9Sstevel@tonic-gate /* 33077c478bd9Sstevel@tonic-gate * Link source to the target. If a target exists, return its 33087c478bd9Sstevel@tonic-gate * vnode pointer in tvp. We'll release it after sending the 33097c478bd9Sstevel@tonic-gate * vnevent. 33107c478bd9Sstevel@tonic-gate */ 33117c478bd9Sstevel@tonic-gate if (error = ufs_direnter_lr(tdp, tnm, DE_RENAME, sdp, sip, cr, &tvp)) { 33127c478bd9Sstevel@tonic-gate /* 33137c478bd9Sstevel@tonic-gate * ESAME isn't really an error; it indicates that the 33147c478bd9Sstevel@tonic-gate * operation should not be done because the source and target 33157c478bd9Sstevel@tonic-gate * are the same file, but that no error should be reported. 33167c478bd9Sstevel@tonic-gate */ 33177c478bd9Sstevel@tonic-gate if (error == ESAME) 33187c478bd9Sstevel@tonic-gate error = 0; 33197c478bd9Sstevel@tonic-gate goto errout; 33207c478bd9Sstevel@tonic-gate } 33217c478bd9Sstevel@tonic-gate 33227c478bd9Sstevel@tonic-gate /* 33237c478bd9Sstevel@tonic-gate * Unlink the source. 33247c478bd9Sstevel@tonic-gate * Remove the source entry. ufs_dirremove() checks that the entry 33257c478bd9Sstevel@tonic-gate * still reflects sip, and returns an error if it doesn't. 33267c478bd9Sstevel@tonic-gate * If the entry has changed just forget about it. Release 33277c478bd9Sstevel@tonic-gate * the source inode. 33287c478bd9Sstevel@tonic-gate */ 33297c478bd9Sstevel@tonic-gate if ((error = ufs_dirremove(sdp, snm, sip, (struct vnode *)0, 33307c478bd9Sstevel@tonic-gate DR_RENAME, cr, NULL)) == ENOENT) 33317c478bd9Sstevel@tonic-gate error = 0; 33327c478bd9Sstevel@tonic-gate 33337c478bd9Sstevel@tonic-gate errout: 33347c478bd9Sstevel@tonic-gate rw_exit(&tdp->i_rwlock); 33357c478bd9Sstevel@tonic-gate if (sdp != tdp) { 33367c478bd9Sstevel@tonic-gate rw_exit(&sdp->i_rwlock); 33377c478bd9Sstevel@tonic-gate } 33387c478bd9Sstevel@tonic-gate 33397c478bd9Sstevel@tonic-gate unlock: 33407c478bd9Sstevel@tonic-gate if (ulp) { 33417c478bd9Sstevel@tonic-gate TRANS_END_CSYNC(ufsvfsp, error, issync, TOP_RENAME, trans_size); 33427c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 33437c478bd9Sstevel@tonic-gate } 33447c478bd9Sstevel@tonic-gate 33457c478bd9Sstevel@tonic-gate /* 33467c478bd9Sstevel@tonic-gate * If no errors, send the appropriate events on the source 33477c478bd9Sstevel@tonic-gate * and destination (a.k.a, target) vnodes, if they exist. 33487c478bd9Sstevel@tonic-gate * This has to be done after the rename transaction has closed. 33497c478bd9Sstevel@tonic-gate */ 33507c478bd9Sstevel@tonic-gate if (error == 0) { 33517c478bd9Sstevel@tonic-gate if (tvp != NULL) 33527c478bd9Sstevel@tonic-gate vnevent_rename_dest(tvp); 33537c478bd9Sstevel@tonic-gate /* 33547c478bd9Sstevel@tonic-gate * Note that if ufs_direnter_lr() returned ESAME then 33557c478bd9Sstevel@tonic-gate * this event will still be sent. This isn't expected 33567c478bd9Sstevel@tonic-gate * to be a problem for anticipated usage by consumers. 33577c478bd9Sstevel@tonic-gate */ 33587c478bd9Sstevel@tonic-gate if (sip != NULL) 33597c478bd9Sstevel@tonic-gate vnevent_rename_src(ITOV(sip)); 33607c478bd9Sstevel@tonic-gate } 33617c478bd9Sstevel@tonic-gate 33627c478bd9Sstevel@tonic-gate if (tvp != NULL) 33637c478bd9Sstevel@tonic-gate VN_RELE(tvp); 33647c478bd9Sstevel@tonic-gate 33657c478bd9Sstevel@tonic-gate if (sip != NULL) 33667c478bd9Sstevel@tonic-gate VN_RELE(ITOV(sip)); 33677c478bd9Sstevel@tonic-gate 33687c478bd9Sstevel@tonic-gate out: 33697c478bd9Sstevel@tonic-gate TRACE_5(TR_FAC_UFS, TR_UFS_RENAME_END, 33707c478bd9Sstevel@tonic-gate "ufs_rename_end:sdvp %p snm %s tdvp %p tnm %s error %d", 33717c478bd9Sstevel@tonic-gate sdvp, snm, tdvp, tnm, error); 33727c478bd9Sstevel@tonic-gate return (error); 33737c478bd9Sstevel@tonic-gate } 33747c478bd9Sstevel@tonic-gate 33757c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 33767c478bd9Sstevel@tonic-gate static int 33777c478bd9Sstevel@tonic-gate ufs_mkdir(struct vnode *dvp, char *dirname, struct vattr *vap, 33787c478bd9Sstevel@tonic-gate struct vnode **vpp, struct cred *cr) 33797c478bd9Sstevel@tonic-gate { 33807c478bd9Sstevel@tonic-gate struct inode *ip; 33817c478bd9Sstevel@tonic-gate struct inode *xip; 33827c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp; 33837c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 33847c478bd9Sstevel@tonic-gate int error; 33857c478bd9Sstevel@tonic-gate int issync; 33867c478bd9Sstevel@tonic-gate int trans_size; 33877c478bd9Sstevel@tonic-gate int retry = 1; 33887c478bd9Sstevel@tonic-gate 33897c478bd9Sstevel@tonic-gate ASSERT((vap->va_mask & (AT_TYPE|AT_MODE)) == (AT_TYPE|AT_MODE)); 33907c478bd9Sstevel@tonic-gate 33917c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_MKDIR_START, 33927c478bd9Sstevel@tonic-gate "ufs_mkdir_start:dvp %p", dvp); 33937c478bd9Sstevel@tonic-gate 33947c478bd9Sstevel@tonic-gate /* 33957c478bd9Sstevel@tonic-gate * Can't make directory in attr hidden dir 33967c478bd9Sstevel@tonic-gate */ 33977c478bd9Sstevel@tonic-gate if ((VTOI(dvp)->i_mode & IFMT) == IFATTRDIR) 33987c478bd9Sstevel@tonic-gate return (EINVAL); 33997c478bd9Sstevel@tonic-gate 34007c478bd9Sstevel@tonic-gate again: 34017c478bd9Sstevel@tonic-gate ip = VTOI(dvp); 34027c478bd9Sstevel@tonic-gate ufsvfsp = ip->i_ufsvfs; 34037c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_MKDIR_MASK); 34047c478bd9Sstevel@tonic-gate if (error) 34057c478bd9Sstevel@tonic-gate goto out; 34067c478bd9Sstevel@tonic-gate if (ulp) 34077c478bd9Sstevel@tonic-gate TRANS_BEGIN_CSYNC(ufsvfsp, issync, TOP_MKDIR, 34087c478bd9Sstevel@tonic-gate trans_size = (int)TOP_MKDIR_SIZE(ip)); 34097c478bd9Sstevel@tonic-gate 34107c478bd9Sstevel@tonic-gate rw_enter(&ip->i_rwlock, RW_WRITER); 34117c478bd9Sstevel@tonic-gate 34127c478bd9Sstevel@tonic-gate error = ufs_direnter_cm(ip, dirname, DE_MKDIR, vap, &xip, cr, 34137c478bd9Sstevel@tonic-gate (retry ? IQUIET : 0)); 34147c478bd9Sstevel@tonic-gate 34157c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 34167c478bd9Sstevel@tonic-gate if (error == 0) { 34177c478bd9Sstevel@tonic-gate ip = xip; 34187c478bd9Sstevel@tonic-gate *vpp = ITOV(ip); 34197c478bd9Sstevel@tonic-gate } else if (error == EEXIST) 34207c478bd9Sstevel@tonic-gate VN_RELE(ITOV(xip)); 34217c478bd9Sstevel@tonic-gate 34227c478bd9Sstevel@tonic-gate if (ulp) { 34237c478bd9Sstevel@tonic-gate int terr = 0; 34247c478bd9Sstevel@tonic-gate TRANS_END_CSYNC(ufsvfsp, terr, issync, TOP_MKDIR, trans_size); 34257c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 34267c478bd9Sstevel@tonic-gate if (error == 0) 34277c478bd9Sstevel@tonic-gate error = terr; 34287c478bd9Sstevel@tonic-gate } 34297c478bd9Sstevel@tonic-gate out: 34307c478bd9Sstevel@tonic-gate if ((error == ENOSPC) && retry && TRANS_ISTRANS(ufsvfsp)) { 34317c478bd9Sstevel@tonic-gate ufs_delete_drain_wait(ufsvfsp, 1); 34327c478bd9Sstevel@tonic-gate retry = 0; 34337c478bd9Sstevel@tonic-gate goto again; 34347c478bd9Sstevel@tonic-gate } 34357c478bd9Sstevel@tonic-gate 34367c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_MKDIR_END, 34377c478bd9Sstevel@tonic-gate "ufs_mkdir_end:dvp %p error %d", dvp, error); 34387c478bd9Sstevel@tonic-gate return (error); 34397c478bd9Sstevel@tonic-gate } 34407c478bd9Sstevel@tonic-gate 34417c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 34427c478bd9Sstevel@tonic-gate static int 34437c478bd9Sstevel@tonic-gate ufs_rmdir(struct vnode *vp, char *nm, struct vnode *cdir, struct cred *cr) 34447c478bd9Sstevel@tonic-gate { 34457c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 34467c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp = ip->i_ufsvfs; 34477c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 34487c478bd9Sstevel@tonic-gate vnode_t *rmvp = NULL; /* Vnode of removed directory */ 34497c478bd9Sstevel@tonic-gate int error; 34507c478bd9Sstevel@tonic-gate int issync; 34517c478bd9Sstevel@tonic-gate 34527c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_RMDIR_START, 34537c478bd9Sstevel@tonic-gate "ufs_rmdir_start:vp %p", vp); 34547c478bd9Sstevel@tonic-gate 34557c478bd9Sstevel@tonic-gate /* 34567c478bd9Sstevel@tonic-gate * don't let the delete queue get too long 34577c478bd9Sstevel@tonic-gate */ 34587c478bd9Sstevel@tonic-gate if (ufsvfsp == NULL) { 34597c478bd9Sstevel@tonic-gate error = EIO; 34607c478bd9Sstevel@tonic-gate goto out; 34617c478bd9Sstevel@tonic-gate } 34627c478bd9Sstevel@tonic-gate if (ufsvfsp->vfs_delete.uq_ne > ufs_idle_max) 34637c478bd9Sstevel@tonic-gate ufs_delete_drain(vp->v_vfsp, 1, 1); 34647c478bd9Sstevel@tonic-gate 34657c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_RMDIR_MASK); 34667c478bd9Sstevel@tonic-gate if (error) 34677c478bd9Sstevel@tonic-gate goto out; 34687c478bd9Sstevel@tonic-gate 34697c478bd9Sstevel@tonic-gate if (ulp) 34707c478bd9Sstevel@tonic-gate TRANS_BEGIN_CSYNC(ufsvfsp, issync, TOP_RMDIR, TOP_RMDIR_SIZE); 34717c478bd9Sstevel@tonic-gate 34727c478bd9Sstevel@tonic-gate rw_enter(&ip->i_rwlock, RW_WRITER); 34737c478bd9Sstevel@tonic-gate error = ufs_dirremove(ip, nm, (struct inode *)0, cdir, DR_RMDIR, cr, 34747c478bd9Sstevel@tonic-gate &rmvp); 34757c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 34767c478bd9Sstevel@tonic-gate 34777c478bd9Sstevel@tonic-gate if (ulp) { 34787c478bd9Sstevel@tonic-gate TRANS_END_CSYNC(ufsvfsp, error, issync, TOP_RMDIR, 34797c478bd9Sstevel@tonic-gate TOP_RMDIR_SIZE); 34807c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 34817c478bd9Sstevel@tonic-gate } 34827c478bd9Sstevel@tonic-gate 34837c478bd9Sstevel@tonic-gate /* 34847c478bd9Sstevel@tonic-gate * This must be done AFTER the rmdir transaction has closed. 34857c478bd9Sstevel@tonic-gate */ 34867c478bd9Sstevel@tonic-gate if (rmvp != NULL) { 34877c478bd9Sstevel@tonic-gate /* Only send the event if there were no errors */ 34887c478bd9Sstevel@tonic-gate if (error == 0) 34897c478bd9Sstevel@tonic-gate vnevent_rmdir(rmvp); 34907c478bd9Sstevel@tonic-gate VN_RELE(rmvp); 34917c478bd9Sstevel@tonic-gate } 34927c478bd9Sstevel@tonic-gate out: 34937c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_RMDIR_END, 34947c478bd9Sstevel@tonic-gate "ufs_rmdir_end:vp %p error %d", vp, error); 34957c478bd9Sstevel@tonic-gate 34967c478bd9Sstevel@tonic-gate return (error); 34977c478bd9Sstevel@tonic-gate } 34987c478bd9Sstevel@tonic-gate 34997c478bd9Sstevel@tonic-gate /* ARGSUSED */ 35007c478bd9Sstevel@tonic-gate static int 35017c478bd9Sstevel@tonic-gate ufs_readdir( 35027c478bd9Sstevel@tonic-gate struct vnode *vp, 35037c478bd9Sstevel@tonic-gate struct uio *uiop, 35047c478bd9Sstevel@tonic-gate struct cred *cr, 35057c478bd9Sstevel@tonic-gate int *eofp) 35067c478bd9Sstevel@tonic-gate { 35077c478bd9Sstevel@tonic-gate struct iovec *iovp; 35087c478bd9Sstevel@tonic-gate struct inode *ip; 35097c478bd9Sstevel@tonic-gate struct direct *idp; 35107c478bd9Sstevel@tonic-gate struct dirent64 *odp; 35117c478bd9Sstevel@tonic-gate struct fbuf *fbp; 35127c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp; 35137c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 35147c478bd9Sstevel@tonic-gate caddr_t outbuf; 35157c478bd9Sstevel@tonic-gate size_t bufsize; 35167c478bd9Sstevel@tonic-gate uint_t offset; 35177c478bd9Sstevel@tonic-gate uint_t bytes_wanted, total_bytes_wanted; 35187c478bd9Sstevel@tonic-gate int incount = 0; 35197c478bd9Sstevel@tonic-gate int outcount = 0; 35207c478bd9Sstevel@tonic-gate int error; 35217c478bd9Sstevel@tonic-gate 35227c478bd9Sstevel@tonic-gate ip = VTOI(vp); 35237c478bd9Sstevel@tonic-gate ASSERT(RW_READ_HELD(&ip->i_rwlock)); 35247c478bd9Sstevel@tonic-gate 35257c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_READDIR_START, 35267c478bd9Sstevel@tonic-gate "ufs_readdir_start:vp %p uiop %p", vp, uiop); 35277c478bd9Sstevel@tonic-gate 35287c478bd9Sstevel@tonic-gate if (uiop->uio_loffset >= MAXOFF32_T) { 35297c478bd9Sstevel@tonic-gate if (eofp) 35307c478bd9Sstevel@tonic-gate *eofp = 1; 35317c478bd9Sstevel@tonic-gate return (0); 35327c478bd9Sstevel@tonic-gate } 35337c478bd9Sstevel@tonic-gate 35347c478bd9Sstevel@tonic-gate /* 35357c478bd9Sstevel@tonic-gate * Check if we have been called with a valid iov_len 35367c478bd9Sstevel@tonic-gate * and bail out if not, otherwise we may potentially loop 35377c478bd9Sstevel@tonic-gate * forever further down. 35387c478bd9Sstevel@tonic-gate */ 35397c478bd9Sstevel@tonic-gate if (uiop->uio_iov->iov_len <= 0) { 35407c478bd9Sstevel@tonic-gate error = EINVAL; 35417c478bd9Sstevel@tonic-gate goto out; 35427c478bd9Sstevel@tonic-gate } 35437c478bd9Sstevel@tonic-gate 35447c478bd9Sstevel@tonic-gate /* 35457c478bd9Sstevel@tonic-gate * Large Files: When we come here we are guaranteed that 35467c478bd9Sstevel@tonic-gate * uio_offset can be used safely. The high word is zero. 35477c478bd9Sstevel@tonic-gate */ 35487c478bd9Sstevel@tonic-gate 35497c478bd9Sstevel@tonic-gate ufsvfsp = ip->i_ufsvfs; 35507c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_READDIR_MASK); 35517c478bd9Sstevel@tonic-gate if (error) 35527c478bd9Sstevel@tonic-gate goto out; 35537c478bd9Sstevel@tonic-gate 35547c478bd9Sstevel@tonic-gate iovp = uiop->uio_iov; 35557c478bd9Sstevel@tonic-gate total_bytes_wanted = iovp->iov_len; 35567c478bd9Sstevel@tonic-gate 35577c478bd9Sstevel@tonic-gate /* Large Files: directory files should not be "large" */ 35587c478bd9Sstevel@tonic-gate 35597c478bd9Sstevel@tonic-gate ASSERT(ip->i_size <= MAXOFF32_T); 35607c478bd9Sstevel@tonic-gate 35617c478bd9Sstevel@tonic-gate /* Force offset to be valid (to guard against bogus lseek() values) */ 35627c478bd9Sstevel@tonic-gate offset = (uint_t)uiop->uio_offset & ~(DIRBLKSIZ - 1); 35637c478bd9Sstevel@tonic-gate 35647c478bd9Sstevel@tonic-gate /* Quit if at end of file or link count of zero (posix) */ 35657c478bd9Sstevel@tonic-gate if (offset >= (uint_t)ip->i_size || ip->i_nlink <= 0) { 35667c478bd9Sstevel@tonic-gate if (eofp) 35677c478bd9Sstevel@tonic-gate *eofp = 1; 35687c478bd9Sstevel@tonic-gate error = 0; 35697c478bd9Sstevel@tonic-gate goto unlock; 35707c478bd9Sstevel@tonic-gate } 35717c478bd9Sstevel@tonic-gate 35727c478bd9Sstevel@tonic-gate /* 35737c478bd9Sstevel@tonic-gate * Get space to change directory entries into fs independent format. 35747c478bd9Sstevel@tonic-gate * Do fast alloc for the most commonly used-request size (filesystem 35757c478bd9Sstevel@tonic-gate * block size). 35767c478bd9Sstevel@tonic-gate */ 35777c478bd9Sstevel@tonic-gate if (uiop->uio_segflg != UIO_SYSSPACE || uiop->uio_iovcnt != 1) { 35787c478bd9Sstevel@tonic-gate bufsize = total_bytes_wanted; 35797c478bd9Sstevel@tonic-gate outbuf = kmem_alloc(bufsize, KM_SLEEP); 35807c478bd9Sstevel@tonic-gate odp = (struct dirent64 *)outbuf; 35817c478bd9Sstevel@tonic-gate } else { 35827c478bd9Sstevel@tonic-gate bufsize = total_bytes_wanted; 35837c478bd9Sstevel@tonic-gate odp = (struct dirent64 *)iovp->iov_base; 35847c478bd9Sstevel@tonic-gate } 35857c478bd9Sstevel@tonic-gate 35867c478bd9Sstevel@tonic-gate nextblk: 35877c478bd9Sstevel@tonic-gate bytes_wanted = total_bytes_wanted; 35887c478bd9Sstevel@tonic-gate 35897c478bd9Sstevel@tonic-gate /* Truncate request to file size */ 35907c478bd9Sstevel@tonic-gate if (offset + bytes_wanted > (int)ip->i_size) 35917c478bd9Sstevel@tonic-gate bytes_wanted = (int)(ip->i_size - offset); 35927c478bd9Sstevel@tonic-gate 35937c478bd9Sstevel@tonic-gate /* Comply with MAXBSIZE boundary restrictions of fbread() */ 35947c478bd9Sstevel@tonic-gate if ((offset & MAXBOFFSET) + bytes_wanted > MAXBSIZE) 35957c478bd9Sstevel@tonic-gate bytes_wanted = MAXBSIZE - (offset & MAXBOFFSET); 35967c478bd9Sstevel@tonic-gate 35977c478bd9Sstevel@tonic-gate /* 35987c478bd9Sstevel@tonic-gate * Read in the next chunk. 35997c478bd9Sstevel@tonic-gate * We are still holding the i_rwlock. 36007c478bd9Sstevel@tonic-gate */ 36017c478bd9Sstevel@tonic-gate error = fbread(vp, (offset_t)offset, bytes_wanted, S_OTHER, &fbp); 36027c478bd9Sstevel@tonic-gate 36037c478bd9Sstevel@tonic-gate if (error) 36047c478bd9Sstevel@tonic-gate goto update_inode; 36057c478bd9Sstevel@tonic-gate if (!ULOCKFS_IS_NOIACC(ITOUL(ip)) && (ip->i_fs->fs_ronly == 0) && 36067c478bd9Sstevel@tonic-gate (!ufsvfsp->vfs_noatime)) { 36077c478bd9Sstevel@tonic-gate ip->i_flag |= IACC; 36087c478bd9Sstevel@tonic-gate } 36097c478bd9Sstevel@tonic-gate incount = 0; 36107c478bd9Sstevel@tonic-gate idp = (struct direct *)fbp->fb_addr; 36117c478bd9Sstevel@tonic-gate if (idp->d_ino == 0 && idp->d_reclen == 0 && 36127c478bd9Sstevel@tonic-gate idp->d_namlen == 0) { 36137c478bd9Sstevel@tonic-gate cmn_err(CE_WARN, "ufs_readdir: bad dir, inumber = %llu, " 36147c478bd9Sstevel@tonic-gate "fs = %s\n", 36157c478bd9Sstevel@tonic-gate (u_longlong_t)ip->i_number, ufsvfsp->vfs_fs->fs_fsmnt); 36167c478bd9Sstevel@tonic-gate fbrelse(fbp, S_OTHER); 36177c478bd9Sstevel@tonic-gate error = ENXIO; 36187c478bd9Sstevel@tonic-gate goto update_inode; 36197c478bd9Sstevel@tonic-gate } 36207c478bd9Sstevel@tonic-gate /* Transform to file-system independent format */ 36217c478bd9Sstevel@tonic-gate while (incount < bytes_wanted) { 36227c478bd9Sstevel@tonic-gate /* 36237c478bd9Sstevel@tonic-gate * If the current directory entry is mangled, then skip 36247c478bd9Sstevel@tonic-gate * to the next block. It would be nice to set the FSBAD 36257c478bd9Sstevel@tonic-gate * flag in the super-block so that a fsck is forced on 36267c478bd9Sstevel@tonic-gate * next reboot, but locking is a problem. 36277c478bd9Sstevel@tonic-gate */ 36287c478bd9Sstevel@tonic-gate if (idp->d_reclen & 0x3) { 36297c478bd9Sstevel@tonic-gate offset = (offset + DIRBLKSIZ) & ~(DIRBLKSIZ-1); 36307c478bd9Sstevel@tonic-gate break; 36317c478bd9Sstevel@tonic-gate } 36327c478bd9Sstevel@tonic-gate 36337c478bd9Sstevel@tonic-gate /* Skip to requested offset and skip empty entries */ 36347c478bd9Sstevel@tonic-gate if (idp->d_ino != 0 && offset >= (uint_t)uiop->uio_offset) { 36357c478bd9Sstevel@tonic-gate ushort_t this_reclen = 36367c478bd9Sstevel@tonic-gate DIRENT64_RECLEN(idp->d_namlen); 36377c478bd9Sstevel@tonic-gate /* Buffer too small for any entries */ 36387c478bd9Sstevel@tonic-gate if (!outcount && this_reclen > bufsize) { 36397c478bd9Sstevel@tonic-gate fbrelse(fbp, S_OTHER); 36407c478bd9Sstevel@tonic-gate error = EINVAL; 36417c478bd9Sstevel@tonic-gate goto update_inode; 36427c478bd9Sstevel@tonic-gate } 36437c478bd9Sstevel@tonic-gate /* If would overrun the buffer, quit */ 36447c478bd9Sstevel@tonic-gate if (outcount + this_reclen > bufsize) { 36457c478bd9Sstevel@tonic-gate break; 36467c478bd9Sstevel@tonic-gate } 36477c478bd9Sstevel@tonic-gate /* Take this entry */ 36487c478bd9Sstevel@tonic-gate odp->d_ino = (ino64_t)idp->d_ino; 36497c478bd9Sstevel@tonic-gate odp->d_reclen = (ushort_t)this_reclen; 36507c478bd9Sstevel@tonic-gate odp->d_off = (offset_t)(offset + idp->d_reclen); 36517c478bd9Sstevel@tonic-gate 36527c478bd9Sstevel@tonic-gate /* use strncpy(9f) to zero out uninitialized bytes */ 36537c478bd9Sstevel@tonic-gate 36547c478bd9Sstevel@tonic-gate ASSERT(strlen(idp->d_name) + 1 <= 36557c478bd9Sstevel@tonic-gate DIRENT64_NAMELEN(this_reclen)); 36567c478bd9Sstevel@tonic-gate (void) strncpy(odp->d_name, idp->d_name, 36577c478bd9Sstevel@tonic-gate DIRENT64_NAMELEN(this_reclen)); 36587c478bd9Sstevel@tonic-gate outcount += odp->d_reclen; 36597c478bd9Sstevel@tonic-gate odp = (struct dirent64 *)((intptr_t)odp + 36607c478bd9Sstevel@tonic-gate odp->d_reclen); 36617c478bd9Sstevel@tonic-gate ASSERT(outcount <= bufsize); 36627c478bd9Sstevel@tonic-gate } 36637c478bd9Sstevel@tonic-gate if (idp->d_reclen) { 36647c478bd9Sstevel@tonic-gate incount += idp->d_reclen; 36657c478bd9Sstevel@tonic-gate offset += idp->d_reclen; 36667c478bd9Sstevel@tonic-gate idp = (struct direct *)((intptr_t)idp + idp->d_reclen); 36677c478bd9Sstevel@tonic-gate } else { 36687c478bd9Sstevel@tonic-gate offset = (offset + DIRBLKSIZ) & ~(DIRBLKSIZ-1); 36697c478bd9Sstevel@tonic-gate break; 36707c478bd9Sstevel@tonic-gate } 36717c478bd9Sstevel@tonic-gate } 36727c478bd9Sstevel@tonic-gate /* Release the chunk */ 36737c478bd9Sstevel@tonic-gate fbrelse(fbp, S_OTHER); 36747c478bd9Sstevel@tonic-gate 36757c478bd9Sstevel@tonic-gate /* Read whole block, but got no entries, read another if not eof */ 36767c478bd9Sstevel@tonic-gate 36777c478bd9Sstevel@tonic-gate /* 36787c478bd9Sstevel@tonic-gate * Large Files: casting i_size to int here is not a problem 36797c478bd9Sstevel@tonic-gate * because directory sizes are always less than MAXOFF32_T. 36807c478bd9Sstevel@tonic-gate * See assertion above. 36817c478bd9Sstevel@tonic-gate */ 36827c478bd9Sstevel@tonic-gate 36837c478bd9Sstevel@tonic-gate if (offset < (int)ip->i_size && !outcount) 36847c478bd9Sstevel@tonic-gate goto nextblk; 36857c478bd9Sstevel@tonic-gate 36867c478bd9Sstevel@tonic-gate /* Copy out the entry data */ 36877c478bd9Sstevel@tonic-gate if (uiop->uio_segflg == UIO_SYSSPACE && uiop->uio_iovcnt == 1) { 36887c478bd9Sstevel@tonic-gate iovp->iov_base += outcount; 36897c478bd9Sstevel@tonic-gate iovp->iov_len -= outcount; 36907c478bd9Sstevel@tonic-gate uiop->uio_resid -= outcount; 36917c478bd9Sstevel@tonic-gate uiop->uio_offset = offset; 36927c478bd9Sstevel@tonic-gate } else if ((error = uiomove(outbuf, (long)outcount, UIO_READ, 36937c478bd9Sstevel@tonic-gate uiop)) == 0) 36947c478bd9Sstevel@tonic-gate uiop->uio_offset = offset; 36957c478bd9Sstevel@tonic-gate update_inode: 36967c478bd9Sstevel@tonic-gate ITIMES(ip); 36977c478bd9Sstevel@tonic-gate if (uiop->uio_segflg != UIO_SYSSPACE || uiop->uio_iovcnt != 1) 36987c478bd9Sstevel@tonic-gate kmem_free(outbuf, bufsize); 36997c478bd9Sstevel@tonic-gate 37007c478bd9Sstevel@tonic-gate if (eofp && error == 0) 37017c478bd9Sstevel@tonic-gate *eofp = (uiop->uio_offset >= (int)ip->i_size); 37027c478bd9Sstevel@tonic-gate unlock: 37037c478bd9Sstevel@tonic-gate if (ulp) { 37047c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 37057c478bd9Sstevel@tonic-gate } 37067c478bd9Sstevel@tonic-gate out: 37077c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_READDIR_END, 37087c478bd9Sstevel@tonic-gate "ufs_readdir_end:vp %p error %d", vp, error); 37097c478bd9Sstevel@tonic-gate return (error); 37107c478bd9Sstevel@tonic-gate } 37117c478bd9Sstevel@tonic-gate 37127c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 37137c478bd9Sstevel@tonic-gate static int 37147c478bd9Sstevel@tonic-gate ufs_symlink( 37157c478bd9Sstevel@tonic-gate struct vnode *dvp, /* ptr to parent dir vnode */ 37167c478bd9Sstevel@tonic-gate char *linkname, /* name of symbolic link */ 37177c478bd9Sstevel@tonic-gate struct vattr *vap, /* attributes */ 37187c478bd9Sstevel@tonic-gate char *target, /* target path */ 37197c478bd9Sstevel@tonic-gate struct cred *cr) /* user credentials */ 37207c478bd9Sstevel@tonic-gate { 37217c478bd9Sstevel@tonic-gate struct inode *ip, *dip = VTOI(dvp); 37227c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp = dip->i_ufsvfs; 37237c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 37247c478bd9Sstevel@tonic-gate int error; 37257c478bd9Sstevel@tonic-gate int issync; 37267c478bd9Sstevel@tonic-gate int trans_size; 37277c478bd9Sstevel@tonic-gate int residual; 37287c478bd9Sstevel@tonic-gate int ioflag; 37297c478bd9Sstevel@tonic-gate int retry = 1; 37307c478bd9Sstevel@tonic-gate 37317c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_SYMLINK_START, 37327c478bd9Sstevel@tonic-gate "ufs_symlink_start:dvp %p", dvp); 37337c478bd9Sstevel@tonic-gate 37347c478bd9Sstevel@tonic-gate /* 37357c478bd9Sstevel@tonic-gate * No symlinks in attrdirs at this time 37367c478bd9Sstevel@tonic-gate */ 37377c478bd9Sstevel@tonic-gate if ((VTOI(dvp)->i_mode & IFMT) == IFATTRDIR) 37387c478bd9Sstevel@tonic-gate return (EINVAL); 37397c478bd9Sstevel@tonic-gate 37407c478bd9Sstevel@tonic-gate again: 37417c478bd9Sstevel@tonic-gate ip = (struct inode *)NULL; 37427c478bd9Sstevel@tonic-gate vap->va_type = VLNK; 37437c478bd9Sstevel@tonic-gate vap->va_rdev = 0; 37447c478bd9Sstevel@tonic-gate 37457c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_SYMLINK_MASK); 37467c478bd9Sstevel@tonic-gate if (error) 37477c478bd9Sstevel@tonic-gate goto out; 37487c478bd9Sstevel@tonic-gate 37497c478bd9Sstevel@tonic-gate if (ulp) 37507c478bd9Sstevel@tonic-gate TRANS_BEGIN_CSYNC(ufsvfsp, issync, TOP_SYMLINK, 37517c478bd9Sstevel@tonic-gate trans_size = (int)TOP_SYMLINK_SIZE(dip)); 37527c478bd9Sstevel@tonic-gate 37537c478bd9Sstevel@tonic-gate /* 37547c478bd9Sstevel@tonic-gate * We must create the inode before the directory entry, to avoid 37557c478bd9Sstevel@tonic-gate * racing with readlink(). ufs_dirmakeinode requires that we 37567c478bd9Sstevel@tonic-gate * hold the quota lock as reader, and directory locks as writer. 37577c478bd9Sstevel@tonic-gate */ 37587c478bd9Sstevel@tonic-gate 37597c478bd9Sstevel@tonic-gate rw_enter(&dip->i_rwlock, RW_WRITER); 37607c478bd9Sstevel@tonic-gate rw_enter(&ufsvfsp->vfs_dqrwlock, RW_READER); 37617c478bd9Sstevel@tonic-gate rw_enter(&dip->i_contents, RW_WRITER); 37627c478bd9Sstevel@tonic-gate 37637c478bd9Sstevel@tonic-gate /* 37647c478bd9Sstevel@tonic-gate * Suppress any out of inodes messages if we will retry on 37657c478bd9Sstevel@tonic-gate * ENOSP 37667c478bd9Sstevel@tonic-gate */ 37677c478bd9Sstevel@tonic-gate if (retry) 37687c478bd9Sstevel@tonic-gate dip->i_flag |= IQUIET; 37697c478bd9Sstevel@tonic-gate 37707c478bd9Sstevel@tonic-gate error = ufs_dirmakeinode(dip, &ip, vap, DE_SYMLINK, cr); 37717c478bd9Sstevel@tonic-gate 37727c478bd9Sstevel@tonic-gate dip->i_flag &= ~IQUIET; 37737c478bd9Sstevel@tonic-gate 37747c478bd9Sstevel@tonic-gate rw_exit(&dip->i_contents); 37757c478bd9Sstevel@tonic-gate rw_exit(&ufsvfsp->vfs_dqrwlock); 37767c478bd9Sstevel@tonic-gate rw_exit(&dip->i_rwlock); 37777c478bd9Sstevel@tonic-gate 37787c478bd9Sstevel@tonic-gate if (error) 37797c478bd9Sstevel@tonic-gate goto unlock; 37807c478bd9Sstevel@tonic-gate 37817c478bd9Sstevel@tonic-gate /* 37827c478bd9Sstevel@tonic-gate * OK. The inode has been created. Write out the data of the 37837c478bd9Sstevel@tonic-gate * symbolic link. Since symbolic links are metadata, and should 37847c478bd9Sstevel@tonic-gate * remain consistent across a system crash, we need to force the 37857c478bd9Sstevel@tonic-gate * data out synchronously. 37867c478bd9Sstevel@tonic-gate * 37877c478bd9Sstevel@tonic-gate * (This is a change from the semantics in earlier releases, which 37887c478bd9Sstevel@tonic-gate * only created symbolic links synchronously if the semi-documented 37897c478bd9Sstevel@tonic-gate * 'syncdir' option was set, or if we were being invoked by the NFS 37907c478bd9Sstevel@tonic-gate * server, which requires symbolic links to be created synchronously.) 37917c478bd9Sstevel@tonic-gate * 37927c478bd9Sstevel@tonic-gate * We need to pass in a pointer for the residual length; otherwise 37937c478bd9Sstevel@tonic-gate * ufs_rdwri() will always return EIO if it can't write the data, 37947c478bd9Sstevel@tonic-gate * even if the error was really ENOSPC or EDQUOT. 37957c478bd9Sstevel@tonic-gate */ 37967c478bd9Sstevel@tonic-gate 37977c478bd9Sstevel@tonic-gate ioflag = FWRITE | FDSYNC; 37987c478bd9Sstevel@tonic-gate residual = 0; 37997c478bd9Sstevel@tonic-gate 38007c478bd9Sstevel@tonic-gate rw_enter(&ufsvfsp->vfs_dqrwlock, RW_READER); 38017c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_WRITER); 38027c478bd9Sstevel@tonic-gate 38037c478bd9Sstevel@tonic-gate /* 38047c478bd9Sstevel@tonic-gate * Suppress file system full messages if we will retry 38057c478bd9Sstevel@tonic-gate */ 38067c478bd9Sstevel@tonic-gate if (retry) 38077c478bd9Sstevel@tonic-gate ip->i_flag |= IQUIET; 38087c478bd9Sstevel@tonic-gate 38097c478bd9Sstevel@tonic-gate error = ufs_rdwri(UIO_WRITE, ioflag, ip, target, strlen(target), 38107c478bd9Sstevel@tonic-gate (offset_t)0, UIO_SYSSPACE, &residual, cr); 38117c478bd9Sstevel@tonic-gate 38127c478bd9Sstevel@tonic-gate ip->i_flag &= ~IQUIET; 38137c478bd9Sstevel@tonic-gate 38147c478bd9Sstevel@tonic-gate if (error) { 38157c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 38167c478bd9Sstevel@tonic-gate rw_exit(&ufsvfsp->vfs_dqrwlock); 38177c478bd9Sstevel@tonic-gate goto remove; 38187c478bd9Sstevel@tonic-gate } 38197c478bd9Sstevel@tonic-gate 38207c478bd9Sstevel@tonic-gate /* 38217c478bd9Sstevel@tonic-gate * If the link's data is small enough, we can cache it in the inode. 38227c478bd9Sstevel@tonic-gate * This is a "fast symbolic link". We don't use the first direct 38237c478bd9Sstevel@tonic-gate * block because that's actually used to point at the symbolic link's 38247c478bd9Sstevel@tonic-gate * contents on disk; but we know that none of the other direct or 38257c478bd9Sstevel@tonic-gate * indirect blocks can be used because symbolic links are restricted 38267c478bd9Sstevel@tonic-gate * to be smaller than a file system block. 38277c478bd9Sstevel@tonic-gate */ 38287c478bd9Sstevel@tonic-gate 38297c478bd9Sstevel@tonic-gate ASSERT(MAXPATHLEN <= VBSIZE(ITOV(ip))); 38307c478bd9Sstevel@tonic-gate 38317c478bd9Sstevel@tonic-gate if (ip->i_size > 0 && ip->i_size <= FSL_SIZE) { 38327c478bd9Sstevel@tonic-gate if (kcopy(target, &ip->i_db[1], ip->i_size) == 0) { 38337c478bd9Sstevel@tonic-gate ip->i_flag |= IFASTSYMLNK; 38347c478bd9Sstevel@tonic-gate } else { 38357c478bd9Sstevel@tonic-gate int i; 38367c478bd9Sstevel@tonic-gate /* error, clear garbage left behind */ 38377c478bd9Sstevel@tonic-gate for (i = 1; i < NDADDR; i++) 38387c478bd9Sstevel@tonic-gate ip->i_db[i] = 0; 38397c478bd9Sstevel@tonic-gate for (i = 0; i < NIADDR; i++) 38407c478bd9Sstevel@tonic-gate ip->i_ib[i] = 0; 38417c478bd9Sstevel@tonic-gate } 38427c478bd9Sstevel@tonic-gate } 38437c478bd9Sstevel@tonic-gate 38447c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 38457c478bd9Sstevel@tonic-gate rw_exit(&ufsvfsp->vfs_dqrwlock); 38467c478bd9Sstevel@tonic-gate 38477c478bd9Sstevel@tonic-gate /* 38487c478bd9Sstevel@tonic-gate * OK. We've successfully created the symbolic link. All that 38497c478bd9Sstevel@tonic-gate * remains is to insert it into the appropriate directory. 38507c478bd9Sstevel@tonic-gate */ 38517c478bd9Sstevel@tonic-gate 38527c478bd9Sstevel@tonic-gate rw_enter(&dip->i_rwlock, RW_WRITER); 38537c478bd9Sstevel@tonic-gate error = ufs_direnter_lr(dip, linkname, DE_SYMLINK, NULL, ip, cr, NULL); 38547c478bd9Sstevel@tonic-gate rw_exit(&dip->i_rwlock); 38557c478bd9Sstevel@tonic-gate 38567c478bd9Sstevel@tonic-gate /* 38577c478bd9Sstevel@tonic-gate * Fall through into remove-on-error code. We're either done, or we 38587c478bd9Sstevel@tonic-gate * need to remove the inode (if we couldn't insert it). 38597c478bd9Sstevel@tonic-gate */ 38607c478bd9Sstevel@tonic-gate 38617c478bd9Sstevel@tonic-gate remove: 38627c478bd9Sstevel@tonic-gate if (error && (ip != NULL)) { 38637c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_WRITER); 38647c478bd9Sstevel@tonic-gate ip->i_nlink--; 38657c478bd9Sstevel@tonic-gate ip->i_flag |= ICHG; 38667c478bd9Sstevel@tonic-gate ip->i_seq++; 38677c478bd9Sstevel@tonic-gate ufs_setreclaim(ip); 38687c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 38697c478bd9Sstevel@tonic-gate } 38707c478bd9Sstevel@tonic-gate 38717c478bd9Sstevel@tonic-gate unlock: 38727c478bd9Sstevel@tonic-gate if (ip != NULL) 38737c478bd9Sstevel@tonic-gate VN_RELE(ITOV(ip)); 38747c478bd9Sstevel@tonic-gate 38757c478bd9Sstevel@tonic-gate if (ulp) { 38767c478bd9Sstevel@tonic-gate int terr = 0; 38777c478bd9Sstevel@tonic-gate 38787c478bd9Sstevel@tonic-gate TRANS_END_CSYNC(ufsvfsp, terr, issync, TOP_SYMLINK, 38797c478bd9Sstevel@tonic-gate trans_size); 38807c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 38817c478bd9Sstevel@tonic-gate if (error == 0) 38827c478bd9Sstevel@tonic-gate error = terr; 38837c478bd9Sstevel@tonic-gate } 38847c478bd9Sstevel@tonic-gate 38857c478bd9Sstevel@tonic-gate /* 38867c478bd9Sstevel@tonic-gate * We may have failed due to lack of an inode or of a block to 38877c478bd9Sstevel@tonic-gate * store the target in. Try flushing the delete queue to free 38887c478bd9Sstevel@tonic-gate * logically-available things up and try again. 38897c478bd9Sstevel@tonic-gate */ 38907c478bd9Sstevel@tonic-gate if ((error == ENOSPC) && retry && TRANS_ISTRANS(ufsvfsp)) { 38917c478bd9Sstevel@tonic-gate ufs_delete_drain_wait(ufsvfsp, 1); 38927c478bd9Sstevel@tonic-gate retry = 0; 38937c478bd9Sstevel@tonic-gate goto again; 38947c478bd9Sstevel@tonic-gate } 38957c478bd9Sstevel@tonic-gate 38967c478bd9Sstevel@tonic-gate out: 38977c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_SYMLINK_END, 38987c478bd9Sstevel@tonic-gate "ufs_symlink_end:dvp %p error %d", dvp, error); 38997c478bd9Sstevel@tonic-gate return (error); 39007c478bd9Sstevel@tonic-gate } 39017c478bd9Sstevel@tonic-gate 39027c478bd9Sstevel@tonic-gate /* 39037c478bd9Sstevel@tonic-gate * Ufs specific routine used to do ufs io. 39047c478bd9Sstevel@tonic-gate */ 39057c478bd9Sstevel@tonic-gate int 39067c478bd9Sstevel@tonic-gate ufs_rdwri(enum uio_rw rw, int ioflag, struct inode *ip, caddr_t base, 39077c478bd9Sstevel@tonic-gate ssize_t len, offset_t offset, enum uio_seg seg, int *aresid, 39087c478bd9Sstevel@tonic-gate struct cred *cr) 39097c478bd9Sstevel@tonic-gate { 39107c478bd9Sstevel@tonic-gate struct uio auio; 39117c478bd9Sstevel@tonic-gate struct iovec aiov; 39127c478bd9Sstevel@tonic-gate int error; 39137c478bd9Sstevel@tonic-gate 39147c478bd9Sstevel@tonic-gate ASSERT(RW_LOCK_HELD(&ip->i_contents)); 39157c478bd9Sstevel@tonic-gate 39167c478bd9Sstevel@tonic-gate bzero((caddr_t)&auio, sizeof (uio_t)); 39177c478bd9Sstevel@tonic-gate bzero((caddr_t)&aiov, sizeof (iovec_t)); 39187c478bd9Sstevel@tonic-gate 39197c478bd9Sstevel@tonic-gate aiov.iov_base = base; 39207c478bd9Sstevel@tonic-gate aiov.iov_len = len; 39217c478bd9Sstevel@tonic-gate auio.uio_iov = &aiov; 39227c478bd9Sstevel@tonic-gate auio.uio_iovcnt = 1; 39237c478bd9Sstevel@tonic-gate auio.uio_loffset = offset; 39247c478bd9Sstevel@tonic-gate auio.uio_segflg = (short)seg; 39257c478bd9Sstevel@tonic-gate auio.uio_resid = len; 39267c478bd9Sstevel@tonic-gate 39277c478bd9Sstevel@tonic-gate if (rw == UIO_WRITE) { 39287c478bd9Sstevel@tonic-gate auio.uio_fmode = FWRITE; 39297c478bd9Sstevel@tonic-gate auio.uio_extflg = UIO_COPY_DEFAULT; 39307c478bd9Sstevel@tonic-gate auio.uio_llimit = curproc->p_fsz_ctl; 39317c478bd9Sstevel@tonic-gate error = wrip(ip, &auio, ioflag, cr); 39327c478bd9Sstevel@tonic-gate } else { 39337c478bd9Sstevel@tonic-gate auio.uio_fmode = FREAD; 39347c478bd9Sstevel@tonic-gate auio.uio_extflg = UIO_COPY_CACHED; 39357c478bd9Sstevel@tonic-gate auio.uio_llimit = MAXOFFSET_T; 39367c478bd9Sstevel@tonic-gate error = rdip(ip, &auio, ioflag, cr); 39377c478bd9Sstevel@tonic-gate } 39387c478bd9Sstevel@tonic-gate 39397c478bd9Sstevel@tonic-gate if (aresid) { 39407c478bd9Sstevel@tonic-gate *aresid = auio.uio_resid; 39417c478bd9Sstevel@tonic-gate } else if (auio.uio_resid) { 39427c478bd9Sstevel@tonic-gate error = EIO; 39437c478bd9Sstevel@tonic-gate } 39447c478bd9Sstevel@tonic-gate return (error); 39457c478bd9Sstevel@tonic-gate } 39467c478bd9Sstevel@tonic-gate 39477c478bd9Sstevel@tonic-gate static int 39487c478bd9Sstevel@tonic-gate ufs_fid(vp, fidp) 39497c478bd9Sstevel@tonic-gate struct vnode *vp; 39507c478bd9Sstevel@tonic-gate struct fid *fidp; 39517c478bd9Sstevel@tonic-gate { 39527c478bd9Sstevel@tonic-gate struct ufid *ufid; 39537c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 39547c478bd9Sstevel@tonic-gate 39557c478bd9Sstevel@tonic-gate if (ip->i_ufsvfs == NULL) 39567c478bd9Sstevel@tonic-gate return (EIO); 39577c478bd9Sstevel@tonic-gate 39587c478bd9Sstevel@tonic-gate if (fidp->fid_len < (sizeof (struct ufid) - sizeof (ushort_t))) { 39597c478bd9Sstevel@tonic-gate fidp->fid_len = sizeof (struct ufid) - sizeof (ushort_t); 39607c478bd9Sstevel@tonic-gate return (ENOSPC); 39617c478bd9Sstevel@tonic-gate } 39627c478bd9Sstevel@tonic-gate 39637c478bd9Sstevel@tonic-gate ufid = (struct ufid *)fidp; 39647c478bd9Sstevel@tonic-gate bzero((char *)ufid, sizeof (struct ufid)); 39657c478bd9Sstevel@tonic-gate ufid->ufid_len = sizeof (struct ufid) - sizeof (ushort_t); 39667c478bd9Sstevel@tonic-gate ufid->ufid_ino = ip->i_number; 39677c478bd9Sstevel@tonic-gate ufid->ufid_gen = ip->i_gen; 39687c478bd9Sstevel@tonic-gate 39697c478bd9Sstevel@tonic-gate return (0); 39707c478bd9Sstevel@tonic-gate } 39717c478bd9Sstevel@tonic-gate 39727c478bd9Sstevel@tonic-gate /* ARGSUSED2 */ 39737c478bd9Sstevel@tonic-gate static int 39747c478bd9Sstevel@tonic-gate ufs_rwlock(struct vnode *vp, int write_lock, caller_context_t *ctp) 39757c478bd9Sstevel@tonic-gate { 39767c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 39777c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp; 39787c478bd9Sstevel@tonic-gate int forcedirectio; 39797c478bd9Sstevel@tonic-gate 39807c478bd9Sstevel@tonic-gate /* 39817c478bd9Sstevel@tonic-gate * Read case is easy. 39827c478bd9Sstevel@tonic-gate */ 39837c478bd9Sstevel@tonic-gate if (!write_lock) { 39847c478bd9Sstevel@tonic-gate rw_enter(&ip->i_rwlock, RW_READER); 39857c478bd9Sstevel@tonic-gate return (V_WRITELOCK_FALSE); 39867c478bd9Sstevel@tonic-gate } 39877c478bd9Sstevel@tonic-gate 39887c478bd9Sstevel@tonic-gate /* 39897c478bd9Sstevel@tonic-gate * Caller has requested a writer lock, but that inhibits any 39907c478bd9Sstevel@tonic-gate * concurrency in the VOPs that follow. Acquire the lock shared 39917c478bd9Sstevel@tonic-gate * and defer exclusive access until it is known to be needed in 39927c478bd9Sstevel@tonic-gate * other VOP handlers. Some cases can be determined here. 39937c478bd9Sstevel@tonic-gate */ 39947c478bd9Sstevel@tonic-gate 39957c478bd9Sstevel@tonic-gate /* 39967c478bd9Sstevel@tonic-gate * If directio is not set, there is no chance of concurrency, 39977c478bd9Sstevel@tonic-gate * so just acquire the lock exclusive. Beware of a forced 39987c478bd9Sstevel@tonic-gate * unmount before looking at the mount option. 39997c478bd9Sstevel@tonic-gate */ 40007c478bd9Sstevel@tonic-gate ufsvfsp = ip->i_ufsvfs; 40017c478bd9Sstevel@tonic-gate forcedirectio = ufsvfsp ? ufsvfsp->vfs_forcedirectio : 0; 40027c478bd9Sstevel@tonic-gate if (!(ip->i_flag & IDIRECTIO || forcedirectio) || 40037c478bd9Sstevel@tonic-gate !ufs_allow_shared_writes) { 40047c478bd9Sstevel@tonic-gate rw_enter(&ip->i_rwlock, RW_WRITER); 40057c478bd9Sstevel@tonic-gate return (V_WRITELOCK_TRUE); 40067c478bd9Sstevel@tonic-gate } 40077c478bd9Sstevel@tonic-gate 40087c478bd9Sstevel@tonic-gate /* 40097c478bd9Sstevel@tonic-gate * Mandatory locking forces acquiring i_rwlock exclusive. 40107c478bd9Sstevel@tonic-gate */ 40117c478bd9Sstevel@tonic-gate if (MANDLOCK(vp, ip->i_mode)) { 40127c478bd9Sstevel@tonic-gate rw_enter(&ip->i_rwlock, RW_WRITER); 40137c478bd9Sstevel@tonic-gate return (V_WRITELOCK_TRUE); 40147c478bd9Sstevel@tonic-gate } 40157c478bd9Sstevel@tonic-gate 40167c478bd9Sstevel@tonic-gate /* 40177c478bd9Sstevel@tonic-gate * Acquire the lock shared in case a concurrent write follows. 40187c478bd9Sstevel@tonic-gate * Mandatory locking could have become enabled before the lock 40197c478bd9Sstevel@tonic-gate * was acquired. Re-check and upgrade if needed. 40207c478bd9Sstevel@tonic-gate */ 40217c478bd9Sstevel@tonic-gate rw_enter(&ip->i_rwlock, RW_READER); 40227c478bd9Sstevel@tonic-gate if (MANDLOCK(vp, ip->i_mode)) { 40237c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 40247c478bd9Sstevel@tonic-gate rw_enter(&ip->i_rwlock, RW_WRITER); 40257c478bd9Sstevel@tonic-gate return (V_WRITELOCK_TRUE); 40267c478bd9Sstevel@tonic-gate } 40277c478bd9Sstevel@tonic-gate return (V_WRITELOCK_FALSE); 40287c478bd9Sstevel@tonic-gate } 40297c478bd9Sstevel@tonic-gate 40307c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 40317c478bd9Sstevel@tonic-gate static void 40327c478bd9Sstevel@tonic-gate ufs_rwunlock(struct vnode *vp, int write_lock, caller_context_t *ctp) 40337c478bd9Sstevel@tonic-gate { 40347c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 40357c478bd9Sstevel@tonic-gate 40367c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 40377c478bd9Sstevel@tonic-gate } 40387c478bd9Sstevel@tonic-gate 40397c478bd9Sstevel@tonic-gate /* ARGSUSED */ 40407c478bd9Sstevel@tonic-gate static int 40417c478bd9Sstevel@tonic-gate ufs_seek(struct vnode *vp, offset_t ooff, offset_t *noffp) 40427c478bd9Sstevel@tonic-gate { 40437c478bd9Sstevel@tonic-gate return ((*noffp < 0 || *noffp > MAXOFFSET_T) ? EINVAL : 0); 40447c478bd9Sstevel@tonic-gate } 40457c478bd9Sstevel@tonic-gate 40467c478bd9Sstevel@tonic-gate /* ARGSUSED */ 40477c478bd9Sstevel@tonic-gate static int 40487c478bd9Sstevel@tonic-gate ufs_frlock(struct vnode *vp, int cmd, struct flock64 *bfp, int flag, 40497c478bd9Sstevel@tonic-gate offset_t offset, struct flk_callback *flk_cbp, struct cred *cr) 40507c478bd9Sstevel@tonic-gate { 40517c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 40527c478bd9Sstevel@tonic-gate 40537c478bd9Sstevel@tonic-gate if (ip->i_ufsvfs == NULL) 40547c478bd9Sstevel@tonic-gate return (EIO); 40557c478bd9Sstevel@tonic-gate 40567c478bd9Sstevel@tonic-gate /* 40577c478bd9Sstevel@tonic-gate * If file is being mapped, disallow frlock. 40587c478bd9Sstevel@tonic-gate * XXX I am not holding tlock while checking i_mapcnt because the 40597c478bd9Sstevel@tonic-gate * current locking strategy drops all locks before calling fs_frlock. 40607c478bd9Sstevel@tonic-gate * So, mapcnt could change before we enter fs_frlock making is 40617c478bd9Sstevel@tonic-gate * meaningless to have held tlock in the first place. 40627c478bd9Sstevel@tonic-gate */ 40637c478bd9Sstevel@tonic-gate if (ip->i_mapcnt > 0 && MANDLOCK(vp, ip->i_mode)) 40647c478bd9Sstevel@tonic-gate return (EAGAIN); 40657c478bd9Sstevel@tonic-gate return (fs_frlock(vp, cmd, bfp, flag, offset, flk_cbp, cr)); 40667c478bd9Sstevel@tonic-gate } 40677c478bd9Sstevel@tonic-gate 40687c478bd9Sstevel@tonic-gate /* ARGSUSED */ 40697c478bd9Sstevel@tonic-gate static int 40707c478bd9Sstevel@tonic-gate ufs_space( 40717c478bd9Sstevel@tonic-gate struct vnode *vp, 40727c478bd9Sstevel@tonic-gate int cmd, 40737c478bd9Sstevel@tonic-gate struct flock64 *bfp, 40747c478bd9Sstevel@tonic-gate int flag, 40757c478bd9Sstevel@tonic-gate offset_t offset, 40767c478bd9Sstevel@tonic-gate cred_t *cr, 40777c478bd9Sstevel@tonic-gate caller_context_t *ct) 40787c478bd9Sstevel@tonic-gate { 40797c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp = VTOI(vp)->i_ufsvfs; 40807c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 40817c478bd9Sstevel@tonic-gate int error; 40827c478bd9Sstevel@tonic-gate 40837c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_SPACE_MASK); 40847c478bd9Sstevel@tonic-gate if (error) 40857c478bd9Sstevel@tonic-gate return (error); 40867c478bd9Sstevel@tonic-gate 40877c478bd9Sstevel@tonic-gate 40887c478bd9Sstevel@tonic-gate if (cmd != F_FREESP) 40897c478bd9Sstevel@tonic-gate error = EINVAL; 40907c478bd9Sstevel@tonic-gate else if ((error = convoff(vp, bfp, 0, offset)) == 0) 40917c478bd9Sstevel@tonic-gate error = ufs_freesp(vp, bfp, flag, cr); 40927c478bd9Sstevel@tonic-gate 40937c478bd9Sstevel@tonic-gate if (ulp) 40947c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 40957c478bd9Sstevel@tonic-gate return (error); 40967c478bd9Sstevel@tonic-gate } 40977c478bd9Sstevel@tonic-gate 40987c478bd9Sstevel@tonic-gate /* 40997c478bd9Sstevel@tonic-gate * Used to determine if read ahead should be done. Also used to 41007c478bd9Sstevel@tonic-gate * to determine when write back occurs. 41017c478bd9Sstevel@tonic-gate */ 41027c478bd9Sstevel@tonic-gate #define CLUSTSZ(ip) ((ip)->i_ufsvfs->vfs_ioclustsz) 41037c478bd9Sstevel@tonic-gate 41047c478bd9Sstevel@tonic-gate /* 41057c478bd9Sstevel@tonic-gate * A faster version of ufs_getpage. 41067c478bd9Sstevel@tonic-gate * 41077c478bd9Sstevel@tonic-gate * We optimize by inlining the pvn_getpages iterator, eliminating 41087c478bd9Sstevel@tonic-gate * calls to bmap_read if file doesn't have UFS holes, and avoiding 41097c478bd9Sstevel@tonic-gate * the overhead of page_exists(). 41107c478bd9Sstevel@tonic-gate * 41117c478bd9Sstevel@tonic-gate * When files has UFS_HOLES and ufs_getpage is called with S_READ, 41127c478bd9Sstevel@tonic-gate * we set *protp to PROT_READ to avoid calling bmap_read. This approach 41137c478bd9Sstevel@tonic-gate * victimizes performance when a file with UFS holes is faulted 41147c478bd9Sstevel@tonic-gate * first in the S_READ mode, and then in the S_WRITE mode. We will get 41157c478bd9Sstevel@tonic-gate * two MMU faults in this case. 41167c478bd9Sstevel@tonic-gate * 41177c478bd9Sstevel@tonic-gate * XXX - the inode fields which control the sequential mode are not 41187c478bd9Sstevel@tonic-gate * protected by any mutex. The read ahead will act wild if 41197c478bd9Sstevel@tonic-gate * multiple processes will access the file concurrently and 41207c478bd9Sstevel@tonic-gate * some of them in sequential mode. One particulary bad case 41217c478bd9Sstevel@tonic-gate * is if another thread will change the value of i_nextrio between 41227c478bd9Sstevel@tonic-gate * the time this thread tests the i_nextrio value and then reads it 41237c478bd9Sstevel@tonic-gate * again to use it as the offset for the read ahead. 41247c478bd9Sstevel@tonic-gate */ 41257c478bd9Sstevel@tonic-gate static int 41267c478bd9Sstevel@tonic-gate ufs_getpage(struct vnode *vp, offset_t off, size_t len, uint_t *protp, 41277c478bd9Sstevel@tonic-gate page_t *plarr[], size_t plsz, struct seg *seg, caddr_t addr, 41287c478bd9Sstevel@tonic-gate enum seg_rw rw, struct cred *cr) 41297c478bd9Sstevel@tonic-gate { 41307c478bd9Sstevel@tonic-gate u_offset_t uoff = (u_offset_t)off; /* type conversion */ 41317c478bd9Sstevel@tonic-gate u_offset_t pgoff; 41327c478bd9Sstevel@tonic-gate u_offset_t eoff; 41337c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 41347c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp = ip->i_ufsvfs; 41357c478bd9Sstevel@tonic-gate struct fs *fs; 41367c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 41377c478bd9Sstevel@tonic-gate page_t **pl; 41387c478bd9Sstevel@tonic-gate caddr_t pgaddr; 41397c478bd9Sstevel@tonic-gate krw_t rwtype; 41407c478bd9Sstevel@tonic-gate int err; 41417c478bd9Sstevel@tonic-gate int has_holes; 41427c478bd9Sstevel@tonic-gate int beyond_eof; 41437c478bd9Sstevel@tonic-gate int seqmode; 41447c478bd9Sstevel@tonic-gate int pgsize = PAGESIZE; 41457c478bd9Sstevel@tonic-gate int dolock; 41467c478bd9Sstevel@tonic-gate int do_qlock; 41477c478bd9Sstevel@tonic-gate int trans_size; 41487c478bd9Sstevel@tonic-gate 41497c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_GETPAGE_START, 41507c478bd9Sstevel@tonic-gate "ufs_getpage_start:vp %p", vp); 41517c478bd9Sstevel@tonic-gate 41527c478bd9Sstevel@tonic-gate ASSERT((uoff & PAGEOFFSET) == 0); 41537c478bd9Sstevel@tonic-gate 41547c478bd9Sstevel@tonic-gate if (protp) 41557c478bd9Sstevel@tonic-gate *protp = PROT_ALL; 41567c478bd9Sstevel@tonic-gate 41577c478bd9Sstevel@tonic-gate /* 41587c478bd9Sstevel@tonic-gate * Obey the lockfs protocol 41597c478bd9Sstevel@tonic-gate */ 41607c478bd9Sstevel@tonic-gate err = ufs_lockfs_begin_getpage(ufsvfsp, &ulp, seg, 41617c478bd9Sstevel@tonic-gate rw == S_READ || rw == S_EXEC, protp); 41627c478bd9Sstevel@tonic-gate if (err) 41637c478bd9Sstevel@tonic-gate goto out; 41647c478bd9Sstevel@tonic-gate 41657c478bd9Sstevel@tonic-gate fs = ufsvfsp->vfs_fs; 41667c478bd9Sstevel@tonic-gate 41677c478bd9Sstevel@tonic-gate if (ulp && (rw == S_CREATE || rw == S_WRITE) && 41687c478bd9Sstevel@tonic-gate !(vp->v_flag & VISSWAP)) { 41697c478bd9Sstevel@tonic-gate /* 41707c478bd9Sstevel@tonic-gate * Try to start a transaction, will return if blocking is 41717c478bd9Sstevel@tonic-gate * expected to occur and the address space is not the 41727c478bd9Sstevel@tonic-gate * kernel address space. 41737c478bd9Sstevel@tonic-gate */ 41747c478bd9Sstevel@tonic-gate trans_size = TOP_GETPAGE_SIZE(ip); 41757c478bd9Sstevel@tonic-gate if (seg->s_as != &kas) { 41767c478bd9Sstevel@tonic-gate TRANS_TRY_BEGIN_ASYNC(ufsvfsp, TOP_GETPAGE, 41777c478bd9Sstevel@tonic-gate trans_size, err) 41787c478bd9Sstevel@tonic-gate if (err == EWOULDBLOCK) { 41797c478bd9Sstevel@tonic-gate /* 41807c478bd9Sstevel@tonic-gate * Use EDEADLK here because the VM code 41817c478bd9Sstevel@tonic-gate * can normally never see this error. 41827c478bd9Sstevel@tonic-gate */ 41837c478bd9Sstevel@tonic-gate err = EDEADLK; 41847c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 41857c478bd9Sstevel@tonic-gate goto out; 41867c478bd9Sstevel@tonic-gate } 41877c478bd9Sstevel@tonic-gate } else { 41887c478bd9Sstevel@tonic-gate TRANS_BEGIN_ASYNC(ufsvfsp, TOP_GETPAGE, trans_size); 41897c478bd9Sstevel@tonic-gate } 41907c478bd9Sstevel@tonic-gate } 41917c478bd9Sstevel@tonic-gate 41927c478bd9Sstevel@tonic-gate if (vp->v_flag & VNOMAP) { 41937c478bd9Sstevel@tonic-gate err = ENOSYS; 41947c478bd9Sstevel@tonic-gate goto unlock; 41957c478bd9Sstevel@tonic-gate } 41967c478bd9Sstevel@tonic-gate 41977c478bd9Sstevel@tonic-gate seqmode = ip->i_nextr == uoff && rw != S_CREATE; 41987c478bd9Sstevel@tonic-gate 41997c478bd9Sstevel@tonic-gate rwtype = RW_READER; /* start as a reader */ 42007c478bd9Sstevel@tonic-gate dolock = (rw_owner(&ip->i_contents) != curthread); 42017c478bd9Sstevel@tonic-gate /* 42027c478bd9Sstevel@tonic-gate * If this thread owns the lock, i.e., this thread grabbed it 42037c478bd9Sstevel@tonic-gate * as writer somewhere above, then we don't need to grab the 42047c478bd9Sstevel@tonic-gate * lock as reader in this routine. 42057c478bd9Sstevel@tonic-gate */ 42067c478bd9Sstevel@tonic-gate do_qlock = (rw_owner(&ufsvfsp->vfs_dqrwlock) != curthread); 42077c478bd9Sstevel@tonic-gate 42087c478bd9Sstevel@tonic-gate retrylock: 42097c478bd9Sstevel@tonic-gate if (dolock) { 42107c478bd9Sstevel@tonic-gate /* 42117c478bd9Sstevel@tonic-gate * Grab the quota lock if we need to call 42127c478bd9Sstevel@tonic-gate * bmap_write() below (with i_contents as writer). 42137c478bd9Sstevel@tonic-gate */ 42147c478bd9Sstevel@tonic-gate if (do_qlock && rwtype == RW_WRITER) 42157c478bd9Sstevel@tonic-gate rw_enter(&ufsvfsp->vfs_dqrwlock, RW_READER); 42167c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, rwtype); 42177c478bd9Sstevel@tonic-gate } 42187c478bd9Sstevel@tonic-gate 42197c478bd9Sstevel@tonic-gate /* 42207c478bd9Sstevel@tonic-gate * We may be getting called as a side effect of a bmap using 42217c478bd9Sstevel@tonic-gate * fbread() when the blocks might be being allocated and the 42227c478bd9Sstevel@tonic-gate * size has not yet been up'ed. In this case we want to be 42237c478bd9Sstevel@tonic-gate * able to return zero pages if we get back UFS_HOLE from 42247c478bd9Sstevel@tonic-gate * calling bmap for a non write case here. We also might have 42257c478bd9Sstevel@tonic-gate * to read some frags from the disk into a page if we are 42267c478bd9Sstevel@tonic-gate * extending the number of frags for a given lbn in bmap(). 42277c478bd9Sstevel@tonic-gate * Large Files: The read of i_size here is atomic because 42287c478bd9Sstevel@tonic-gate * i_contents is held here. If dolock is zero, the lock 42297c478bd9Sstevel@tonic-gate * is held in bmap routines. 42307c478bd9Sstevel@tonic-gate */ 42317c478bd9Sstevel@tonic-gate beyond_eof = uoff + len > ip->i_size + PAGEOFFSET; 42327c478bd9Sstevel@tonic-gate if (beyond_eof && seg != segkmap) { 42337c478bd9Sstevel@tonic-gate if (dolock) { 42347c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 42357c478bd9Sstevel@tonic-gate if (do_qlock && rwtype == RW_WRITER) 42367c478bd9Sstevel@tonic-gate rw_exit(&ufsvfsp->vfs_dqrwlock); 42377c478bd9Sstevel@tonic-gate } 42387c478bd9Sstevel@tonic-gate err = EFAULT; 42397c478bd9Sstevel@tonic-gate goto unlock; 42407c478bd9Sstevel@tonic-gate } 42417c478bd9Sstevel@tonic-gate 42427c478bd9Sstevel@tonic-gate /* 42437c478bd9Sstevel@tonic-gate * Must hold i_contents lock throughout the call to pvn_getpages 42447c478bd9Sstevel@tonic-gate * since locked pages are returned from each call to ufs_getapage. 42457c478bd9Sstevel@tonic-gate * Must *not* return locked pages and then try for contents lock 42467c478bd9Sstevel@tonic-gate * due to lock ordering requirements (inode > page) 42477c478bd9Sstevel@tonic-gate */ 42487c478bd9Sstevel@tonic-gate 42497c478bd9Sstevel@tonic-gate has_holes = bmap_has_holes(ip); 42507c478bd9Sstevel@tonic-gate 42517c478bd9Sstevel@tonic-gate if ((rw == S_WRITE || rw == S_CREATE) && has_holes && !beyond_eof) { 42527c478bd9Sstevel@tonic-gate int blk_size; 42537c478bd9Sstevel@tonic-gate u_offset_t offset; 42547c478bd9Sstevel@tonic-gate 42557c478bd9Sstevel@tonic-gate /* 42567c478bd9Sstevel@tonic-gate * We must acquire the RW_WRITER lock in order to 42577c478bd9Sstevel@tonic-gate * call bmap_write(). 42587c478bd9Sstevel@tonic-gate */ 42597c478bd9Sstevel@tonic-gate if (dolock && rwtype == RW_READER) { 42607c478bd9Sstevel@tonic-gate rwtype = RW_WRITER; 42617c478bd9Sstevel@tonic-gate 42627c478bd9Sstevel@tonic-gate /* 42637c478bd9Sstevel@tonic-gate * Grab the quota lock before 42647c478bd9Sstevel@tonic-gate * upgrading i_contents, but if we can't grab it 42657c478bd9Sstevel@tonic-gate * don't wait here due to lock order: 42667c478bd9Sstevel@tonic-gate * vfs_dqrwlock > i_contents. 42677c478bd9Sstevel@tonic-gate */ 42687c478bd9Sstevel@tonic-gate if (do_qlock && rw_tryenter(&ufsvfsp->vfs_dqrwlock, 42697c478bd9Sstevel@tonic-gate RW_READER) == 0) { 42707c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 42717c478bd9Sstevel@tonic-gate goto retrylock; 42727c478bd9Sstevel@tonic-gate } 42737c478bd9Sstevel@tonic-gate if (!rw_tryupgrade(&ip->i_contents)) { 42747c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 42757c478bd9Sstevel@tonic-gate if (do_qlock) 42767c478bd9Sstevel@tonic-gate rw_exit(&ufsvfsp->vfs_dqrwlock); 42777c478bd9Sstevel@tonic-gate goto retrylock; 42787c478bd9Sstevel@tonic-gate } 42797c478bd9Sstevel@tonic-gate } 42807c478bd9Sstevel@tonic-gate 42817c478bd9Sstevel@tonic-gate /* 42827c478bd9Sstevel@tonic-gate * May be allocating disk blocks for holes here as 42837c478bd9Sstevel@tonic-gate * a result of mmap faults. write(2) does the bmap_write 42847c478bd9Sstevel@tonic-gate * in rdip/wrip, not here. We are not dealing with frags 42857c478bd9Sstevel@tonic-gate * in this case. 42867c478bd9Sstevel@tonic-gate */ 42877c478bd9Sstevel@tonic-gate /* 42887c478bd9Sstevel@tonic-gate * Large Files: We cast fs_bmask field to offset_t 42897c478bd9Sstevel@tonic-gate * just as we do for MAXBMASK because uoff is a 64-bit 42907c478bd9Sstevel@tonic-gate * data type. fs_bmask will still be a 32-bit type 42917c478bd9Sstevel@tonic-gate * as we cannot change any ondisk data structures. 42927c478bd9Sstevel@tonic-gate */ 42937c478bd9Sstevel@tonic-gate 42947c478bd9Sstevel@tonic-gate offset = uoff & (offset_t)fs->fs_bmask; 42957c478bd9Sstevel@tonic-gate while (offset < uoff + len) { 42967c478bd9Sstevel@tonic-gate blk_size = (int)blksize(fs, ip, lblkno(fs, offset)); 42977c478bd9Sstevel@tonic-gate err = bmap_write(ip, offset, blk_size, 0, cr); 42987c478bd9Sstevel@tonic-gate if (ip->i_flag & (ICHG|IUPD)) 42997c478bd9Sstevel@tonic-gate ip->i_seq++; 43007c478bd9Sstevel@tonic-gate if (err) 43017c478bd9Sstevel@tonic-gate goto update_inode; 43027c478bd9Sstevel@tonic-gate offset += blk_size; /* XXX - make this contig */ 43037c478bd9Sstevel@tonic-gate } 43047c478bd9Sstevel@tonic-gate } 43057c478bd9Sstevel@tonic-gate 43067c478bd9Sstevel@tonic-gate /* 43077c478bd9Sstevel@tonic-gate * Can be a reader from now on. 43087c478bd9Sstevel@tonic-gate */ 43097c478bd9Sstevel@tonic-gate if (dolock && rwtype == RW_WRITER) { 43107c478bd9Sstevel@tonic-gate rw_downgrade(&ip->i_contents); 43117c478bd9Sstevel@tonic-gate /* 43127c478bd9Sstevel@tonic-gate * We can release vfs_dqrwlock early so do it, but make 43137c478bd9Sstevel@tonic-gate * sure we don't try to release it again at the bottom. 43147c478bd9Sstevel@tonic-gate */ 43157c478bd9Sstevel@tonic-gate if (do_qlock) { 43167c478bd9Sstevel@tonic-gate rw_exit(&ufsvfsp->vfs_dqrwlock); 43177c478bd9Sstevel@tonic-gate do_qlock = 0; 43187c478bd9Sstevel@tonic-gate } 43197c478bd9Sstevel@tonic-gate } 43207c478bd9Sstevel@tonic-gate 43217c478bd9Sstevel@tonic-gate /* 43227c478bd9Sstevel@tonic-gate * We remove PROT_WRITE in cases when the file has UFS holes 43237c478bd9Sstevel@tonic-gate * because we don't want to call bmap_read() to check each 43247c478bd9Sstevel@tonic-gate * page if it is backed with a disk block. 43257c478bd9Sstevel@tonic-gate */ 43267c478bd9Sstevel@tonic-gate if (protp && has_holes && rw != S_WRITE && rw != S_CREATE) 43277c478bd9Sstevel@tonic-gate *protp &= ~PROT_WRITE; 43287c478bd9Sstevel@tonic-gate 43297c478bd9Sstevel@tonic-gate err = 0; 43307c478bd9Sstevel@tonic-gate 43317c478bd9Sstevel@tonic-gate /* 43327c478bd9Sstevel@tonic-gate * The loop looks up pages in the range [off, off + len). 43337c478bd9Sstevel@tonic-gate * For each page, we first check if we should initiate an asynchronous 43347c478bd9Sstevel@tonic-gate * read ahead before we call page_lookup (we may sleep in page_lookup 43357c478bd9Sstevel@tonic-gate * for a previously initiated disk read). 43367c478bd9Sstevel@tonic-gate */ 43377c478bd9Sstevel@tonic-gate eoff = (uoff + len); 43387c478bd9Sstevel@tonic-gate for (pgoff = uoff, pgaddr = addr, pl = plarr; 43397c478bd9Sstevel@tonic-gate pgoff < eoff; /* empty */) { 43407c478bd9Sstevel@tonic-gate page_t *pp; 43417c478bd9Sstevel@tonic-gate u_offset_t nextrio; 43427c478bd9Sstevel@tonic-gate se_t se; 43437c478bd9Sstevel@tonic-gate int retval; 43447c478bd9Sstevel@tonic-gate 43457c478bd9Sstevel@tonic-gate se = ((rw == S_CREATE || rw == S_OTHER) ? SE_EXCL : SE_SHARED); 43467c478bd9Sstevel@tonic-gate 43477c478bd9Sstevel@tonic-gate /* Handle async getpage (faultahead) */ 43487c478bd9Sstevel@tonic-gate if (plarr == NULL) { 43497c478bd9Sstevel@tonic-gate ip->i_nextrio = pgoff; 43507c478bd9Sstevel@tonic-gate (void) ufs_getpage_ra(vp, pgoff, seg, pgaddr); 43517c478bd9Sstevel@tonic-gate pgoff += pgsize; 43527c478bd9Sstevel@tonic-gate pgaddr += pgsize; 43537c478bd9Sstevel@tonic-gate continue; 43547c478bd9Sstevel@tonic-gate } 43557c478bd9Sstevel@tonic-gate /* 43567c478bd9Sstevel@tonic-gate * Check if we should initiate read ahead of next cluster. 43577c478bd9Sstevel@tonic-gate * We call page_exists only when we need to confirm that 43587c478bd9Sstevel@tonic-gate * we have the current page before we initiate the read ahead. 43597c478bd9Sstevel@tonic-gate */ 43607c478bd9Sstevel@tonic-gate nextrio = ip->i_nextrio; 43617c478bd9Sstevel@tonic-gate if (seqmode && 43627c478bd9Sstevel@tonic-gate pgoff + CLUSTSZ(ip) >= nextrio && pgoff <= nextrio && 43637c478bd9Sstevel@tonic-gate nextrio < ip->i_size && page_exists(vp, pgoff)) { 43647c478bd9Sstevel@tonic-gate retval = ufs_getpage_ra(vp, pgoff, seg, pgaddr); 43657c478bd9Sstevel@tonic-gate /* 43667c478bd9Sstevel@tonic-gate * We always read ahead the next cluster of data 43677c478bd9Sstevel@tonic-gate * starting from i_nextrio. If the page (vp,nextrio) 43687c478bd9Sstevel@tonic-gate * is actually in core at this point, the routine 43697c478bd9Sstevel@tonic-gate * ufs_getpage_ra() will stop pre-fetching data 43707c478bd9Sstevel@tonic-gate * until we read that page in a synchronized manner 43717c478bd9Sstevel@tonic-gate * through ufs_getpage_miss(). So, we should increase 43727c478bd9Sstevel@tonic-gate * i_nextrio if the page (vp, nextrio) exists. 43737c478bd9Sstevel@tonic-gate */ 43747c478bd9Sstevel@tonic-gate if ((retval == 0) && page_exists(vp, nextrio)) { 43757c478bd9Sstevel@tonic-gate ip->i_nextrio = nextrio + pgsize; 43767c478bd9Sstevel@tonic-gate } 43777c478bd9Sstevel@tonic-gate } 43787c478bd9Sstevel@tonic-gate 43797c478bd9Sstevel@tonic-gate if ((pp = page_lookup(vp, pgoff, se)) != NULL) { 43807c478bd9Sstevel@tonic-gate /* 43817c478bd9Sstevel@tonic-gate * We found the page in the page cache. 43827c478bd9Sstevel@tonic-gate */ 43837c478bd9Sstevel@tonic-gate *pl++ = pp; 43847c478bd9Sstevel@tonic-gate pgoff += pgsize; 43857c478bd9Sstevel@tonic-gate pgaddr += pgsize; 43867c478bd9Sstevel@tonic-gate len -= pgsize; 43877c478bd9Sstevel@tonic-gate plsz -= pgsize; 43887c478bd9Sstevel@tonic-gate } else { 43897c478bd9Sstevel@tonic-gate /* 43907c478bd9Sstevel@tonic-gate * We have to create the page, or read it from disk. 43917c478bd9Sstevel@tonic-gate */ 43927c478bd9Sstevel@tonic-gate if (err = ufs_getpage_miss(vp, pgoff, len, seg, pgaddr, 43937c478bd9Sstevel@tonic-gate pl, plsz, rw, seqmode)) 43947c478bd9Sstevel@tonic-gate goto error; 43957c478bd9Sstevel@tonic-gate 43967c478bd9Sstevel@tonic-gate while (*pl != NULL) { 43977c478bd9Sstevel@tonic-gate pl++; 43987c478bd9Sstevel@tonic-gate pgoff += pgsize; 43997c478bd9Sstevel@tonic-gate pgaddr += pgsize; 44007c478bd9Sstevel@tonic-gate len -= pgsize; 44017c478bd9Sstevel@tonic-gate plsz -= pgsize; 44027c478bd9Sstevel@tonic-gate } 44037c478bd9Sstevel@tonic-gate } 44047c478bd9Sstevel@tonic-gate } 44057c478bd9Sstevel@tonic-gate 44067c478bd9Sstevel@tonic-gate /* 44077c478bd9Sstevel@tonic-gate * Return pages up to plsz if they are in the page cache. 44087c478bd9Sstevel@tonic-gate * We cannot return pages if there is a chance that they are 44097c478bd9Sstevel@tonic-gate * backed with a UFS hole and rw is S_WRITE or S_CREATE. 44107c478bd9Sstevel@tonic-gate */ 44117c478bd9Sstevel@tonic-gate if (plarr && !(has_holes && (rw == S_WRITE || rw == S_CREATE))) { 44127c478bd9Sstevel@tonic-gate 44137c478bd9Sstevel@tonic-gate ASSERT((protp == NULL) || 44147c478bd9Sstevel@tonic-gate !(has_holes && (*protp & PROT_WRITE))); 44157c478bd9Sstevel@tonic-gate 44167c478bd9Sstevel@tonic-gate eoff = pgoff + plsz; 44177c478bd9Sstevel@tonic-gate while (pgoff < eoff) { 44187c478bd9Sstevel@tonic-gate page_t *pp; 44197c478bd9Sstevel@tonic-gate 44207c478bd9Sstevel@tonic-gate if ((pp = page_lookup_nowait(vp, pgoff, 44217c478bd9Sstevel@tonic-gate SE_SHARED)) == NULL) 44227c478bd9Sstevel@tonic-gate break; 44237c478bd9Sstevel@tonic-gate 44247c478bd9Sstevel@tonic-gate *pl++ = pp; 44257c478bd9Sstevel@tonic-gate pgoff += pgsize; 44267c478bd9Sstevel@tonic-gate plsz -= pgsize; 44277c478bd9Sstevel@tonic-gate } 44287c478bd9Sstevel@tonic-gate } 44297c478bd9Sstevel@tonic-gate 44307c478bd9Sstevel@tonic-gate if (plarr) 44317c478bd9Sstevel@tonic-gate *pl = NULL; /* Terminate page list */ 44327c478bd9Sstevel@tonic-gate ip->i_nextr = pgoff; 44337c478bd9Sstevel@tonic-gate 44347c478bd9Sstevel@tonic-gate error: 44357c478bd9Sstevel@tonic-gate if (err && plarr) { 44367c478bd9Sstevel@tonic-gate /* 44377c478bd9Sstevel@tonic-gate * Release any pages we have locked. 44387c478bd9Sstevel@tonic-gate */ 44397c478bd9Sstevel@tonic-gate while (pl > &plarr[0]) 44407c478bd9Sstevel@tonic-gate page_unlock(*--pl); 44417c478bd9Sstevel@tonic-gate 44427c478bd9Sstevel@tonic-gate plarr[0] = NULL; 44437c478bd9Sstevel@tonic-gate } 44447c478bd9Sstevel@tonic-gate 44457c478bd9Sstevel@tonic-gate update_inode: 44467c478bd9Sstevel@tonic-gate /* 44477c478bd9Sstevel@tonic-gate * If the inode is not already marked for IACC (in rdip() for read) 44487c478bd9Sstevel@tonic-gate * and the inode is not marked for no access time update (in wrip() 44497c478bd9Sstevel@tonic-gate * for write) then update the inode access time and mod time now. 44507c478bd9Sstevel@tonic-gate */ 44517c478bd9Sstevel@tonic-gate if ((ip->i_flag & (IACC | INOACC)) == 0) { 44527c478bd9Sstevel@tonic-gate if ((rw != S_OTHER) && (ip->i_mode & IFMT) != IFDIR) { 44537c478bd9Sstevel@tonic-gate if (!ULOCKFS_IS_NOIACC(ITOUL(ip)) && 44547c478bd9Sstevel@tonic-gate (fs->fs_ronly == 0) && 44557c478bd9Sstevel@tonic-gate (!ufsvfsp->vfs_noatime)) { 44567c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 44577c478bd9Sstevel@tonic-gate ip->i_flag |= IACC; 44587c478bd9Sstevel@tonic-gate ITIMES_NOLOCK(ip); 44597c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 44607c478bd9Sstevel@tonic-gate } 44617c478bd9Sstevel@tonic-gate } 44627c478bd9Sstevel@tonic-gate } 44637c478bd9Sstevel@tonic-gate 44647c478bd9Sstevel@tonic-gate if (dolock) { 44657c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 44667c478bd9Sstevel@tonic-gate if (do_qlock && rwtype == RW_WRITER) 44677c478bd9Sstevel@tonic-gate rw_exit(&ufsvfsp->vfs_dqrwlock); 44687c478bd9Sstevel@tonic-gate } 44697c478bd9Sstevel@tonic-gate 44707c478bd9Sstevel@tonic-gate unlock: 44717c478bd9Sstevel@tonic-gate if (ulp) { 44727c478bd9Sstevel@tonic-gate if ((rw == S_CREATE || rw == S_WRITE) && 44737c478bd9Sstevel@tonic-gate !(vp->v_flag & VISSWAP)) { 44747c478bd9Sstevel@tonic-gate TRANS_END_ASYNC(ufsvfsp, TOP_GETPAGE, trans_size); 44757c478bd9Sstevel@tonic-gate } 44767c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 44777c478bd9Sstevel@tonic-gate } 44787c478bd9Sstevel@tonic-gate out: 44797c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_GETPAGE_END, 44807c478bd9Sstevel@tonic-gate "ufs_getpage_end:vp %p error %d", vp, err); 44817c478bd9Sstevel@tonic-gate return (err); 44827c478bd9Sstevel@tonic-gate } 44837c478bd9Sstevel@tonic-gate 44847c478bd9Sstevel@tonic-gate /* 44857c478bd9Sstevel@tonic-gate * ufs_getpage_miss is called when ufs_getpage missed the page in the page 44867c478bd9Sstevel@tonic-gate * cache. The page is either read from the disk, or it's created. 44877c478bd9Sstevel@tonic-gate * A page is created (without disk read) if rw == S_CREATE, or if 44887c478bd9Sstevel@tonic-gate * the page is not backed with a real disk block (UFS hole). 44897c478bd9Sstevel@tonic-gate */ 44907c478bd9Sstevel@tonic-gate /* ARGSUSED */ 44917c478bd9Sstevel@tonic-gate static int 44927c478bd9Sstevel@tonic-gate ufs_getpage_miss(struct vnode *vp, u_offset_t off, size_t len, struct seg *seg, 44937c478bd9Sstevel@tonic-gate caddr_t addr, page_t *pl[], size_t plsz, enum seg_rw rw, int seq) 44947c478bd9Sstevel@tonic-gate { 44957c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 44967c478bd9Sstevel@tonic-gate page_t *pp; 44977c478bd9Sstevel@tonic-gate daddr_t bn; 44987c478bd9Sstevel@tonic-gate size_t io_len; 44997c478bd9Sstevel@tonic-gate int crpage; 45007c478bd9Sstevel@tonic-gate int err; 45017c478bd9Sstevel@tonic-gate int contig; 45027c478bd9Sstevel@tonic-gate int bsize = ip->i_fs->fs_bsize; 45037c478bd9Sstevel@tonic-gate 45047c478bd9Sstevel@tonic-gate /* 45057c478bd9Sstevel@tonic-gate * Figure out whether the page can be created, or must be 45067c478bd9Sstevel@tonic-gate * must be read from the disk. 45077c478bd9Sstevel@tonic-gate */ 45087c478bd9Sstevel@tonic-gate if (rw == S_CREATE) 45097c478bd9Sstevel@tonic-gate crpage = 1; 45107c478bd9Sstevel@tonic-gate else { 45117c478bd9Sstevel@tonic-gate contig = 0; 45127c478bd9Sstevel@tonic-gate if (err = bmap_read(ip, off, &bn, &contig)) 45137c478bd9Sstevel@tonic-gate return (err); 45147c478bd9Sstevel@tonic-gate crpage = (bn == UFS_HOLE); 45157c478bd9Sstevel@tonic-gate } 45167c478bd9Sstevel@tonic-gate 45177c478bd9Sstevel@tonic-gate if (crpage) { 45187c478bd9Sstevel@tonic-gate if ((pp = page_create_va(vp, off, PAGESIZE, PG_WAIT, seg, 45197c478bd9Sstevel@tonic-gate addr)) == NULL) { 45207c478bd9Sstevel@tonic-gate return (ufs_fault(vp, 45217c478bd9Sstevel@tonic-gate "ufs_getpage_miss: page_create == NULL")); 45227c478bd9Sstevel@tonic-gate } 45237c478bd9Sstevel@tonic-gate 45247c478bd9Sstevel@tonic-gate if (rw != S_CREATE) 45257c478bd9Sstevel@tonic-gate pagezero(pp, 0, PAGESIZE); 45267c478bd9Sstevel@tonic-gate io_len = PAGESIZE; 45277c478bd9Sstevel@tonic-gate } else { 45287c478bd9Sstevel@tonic-gate u_offset_t io_off; 45297c478bd9Sstevel@tonic-gate uint_t xlen; 45307c478bd9Sstevel@tonic-gate struct buf *bp; 45317c478bd9Sstevel@tonic-gate ufsvfs_t *ufsvfsp = ip->i_ufsvfs; 45327c478bd9Sstevel@tonic-gate 45337c478bd9Sstevel@tonic-gate /* 45347c478bd9Sstevel@tonic-gate * If access is not in sequential order, we read from disk 45357c478bd9Sstevel@tonic-gate * in bsize units. 45367c478bd9Sstevel@tonic-gate * 45377c478bd9Sstevel@tonic-gate * We limit the size of the transfer to bsize if we are reading 45387c478bd9Sstevel@tonic-gate * from the beginning of the file. Note in this situation we 45397c478bd9Sstevel@tonic-gate * will hedge our bets and initiate an async read ahead of 45407c478bd9Sstevel@tonic-gate * the second block. 45417c478bd9Sstevel@tonic-gate */ 45427c478bd9Sstevel@tonic-gate if (!seq || off == 0) 45437c478bd9Sstevel@tonic-gate contig = MIN(contig, bsize); 45447c478bd9Sstevel@tonic-gate 45457c478bd9Sstevel@tonic-gate pp = pvn_read_kluster(vp, off, seg, addr, &io_off, 45467c478bd9Sstevel@tonic-gate &io_len, off, contig, 0); 45477c478bd9Sstevel@tonic-gate 45487c478bd9Sstevel@tonic-gate /* 45497c478bd9Sstevel@tonic-gate * Some other thread has entered the page. 45507c478bd9Sstevel@tonic-gate * ufs_getpage will retry page_lookup. 45517c478bd9Sstevel@tonic-gate */ 45527c478bd9Sstevel@tonic-gate if (pp == NULL) { 45537c478bd9Sstevel@tonic-gate pl[0] = NULL; 45547c478bd9Sstevel@tonic-gate return (0); 45557c478bd9Sstevel@tonic-gate } 45567c478bd9Sstevel@tonic-gate 45577c478bd9Sstevel@tonic-gate /* 45587c478bd9Sstevel@tonic-gate * Zero part of the page which we are not 45597c478bd9Sstevel@tonic-gate * going to read from the disk. 45607c478bd9Sstevel@tonic-gate */ 45617c478bd9Sstevel@tonic-gate xlen = io_len & PAGEOFFSET; 45627c478bd9Sstevel@tonic-gate if (xlen != 0) 45637c478bd9Sstevel@tonic-gate pagezero(pp->p_prev, xlen, PAGESIZE - xlen); 45647c478bd9Sstevel@tonic-gate 45657c478bd9Sstevel@tonic-gate bp = pageio_setup(pp, io_len, ip->i_devvp, B_READ); 45667c478bd9Sstevel@tonic-gate bp->b_edev = ip->i_dev; 45677c478bd9Sstevel@tonic-gate bp->b_dev = cmpdev(ip->i_dev); 45687c478bd9Sstevel@tonic-gate bp->b_blkno = bn; 45697c478bd9Sstevel@tonic-gate bp->b_un.b_addr = (caddr_t)0; 45707c478bd9Sstevel@tonic-gate bp->b_file = ip->i_vnode; 45717c478bd9Sstevel@tonic-gate bp->b_offset = off; 45727c478bd9Sstevel@tonic-gate 45737c478bd9Sstevel@tonic-gate if (ufsvfsp->vfs_log) { 45747c478bd9Sstevel@tonic-gate lufs_read_strategy(ufsvfsp->vfs_log, bp); 45757c478bd9Sstevel@tonic-gate } else if (ufsvfsp->vfs_snapshot) { 45767c478bd9Sstevel@tonic-gate fssnap_strategy(&ufsvfsp->vfs_snapshot, bp); 45777c478bd9Sstevel@tonic-gate } else { 45787c478bd9Sstevel@tonic-gate ufsvfsp->vfs_iotstamp = lbolt; 45797c478bd9Sstevel@tonic-gate ub.ub_getpages.value.ul++; 45807c478bd9Sstevel@tonic-gate (void) bdev_strategy(bp); 45817c478bd9Sstevel@tonic-gate lwp_stat_update(LWP_STAT_INBLK, 1); 45827c478bd9Sstevel@tonic-gate } 45837c478bd9Sstevel@tonic-gate 45847c478bd9Sstevel@tonic-gate ip->i_nextrio = off + ((io_len + PAGESIZE - 1) & PAGEMASK); 45857c478bd9Sstevel@tonic-gate 45867c478bd9Sstevel@tonic-gate /* 45877c478bd9Sstevel@tonic-gate * If the file access is sequential, initiate read ahead 45887c478bd9Sstevel@tonic-gate * of the next cluster. 45897c478bd9Sstevel@tonic-gate */ 45907c478bd9Sstevel@tonic-gate if (seq && ip->i_nextrio < ip->i_size) 45917c478bd9Sstevel@tonic-gate (void) ufs_getpage_ra(vp, off, seg, addr); 45927c478bd9Sstevel@tonic-gate err = biowait(bp); 45937c478bd9Sstevel@tonic-gate pageio_done(bp); 45947c478bd9Sstevel@tonic-gate 45957c478bd9Sstevel@tonic-gate if (err) { 45967c478bd9Sstevel@tonic-gate pvn_read_done(pp, B_ERROR); 45977c478bd9Sstevel@tonic-gate return (err); 45987c478bd9Sstevel@tonic-gate } 45997c478bd9Sstevel@tonic-gate } 46007c478bd9Sstevel@tonic-gate 46017c478bd9Sstevel@tonic-gate pvn_plist_init(pp, pl, plsz, off, io_len, rw); 46027c478bd9Sstevel@tonic-gate return (0); 46037c478bd9Sstevel@tonic-gate } 46047c478bd9Sstevel@tonic-gate 46057c478bd9Sstevel@tonic-gate /* 46067c478bd9Sstevel@tonic-gate * Read ahead a cluster from the disk. Returns the length in bytes. 46077c478bd9Sstevel@tonic-gate */ 46087c478bd9Sstevel@tonic-gate static int 46097c478bd9Sstevel@tonic-gate ufs_getpage_ra(struct vnode *vp, u_offset_t off, struct seg *seg, caddr_t addr) 46107c478bd9Sstevel@tonic-gate { 46117c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 46127c478bd9Sstevel@tonic-gate page_t *pp; 46137c478bd9Sstevel@tonic-gate u_offset_t io_off = ip->i_nextrio; 46147c478bd9Sstevel@tonic-gate ufsvfs_t *ufsvfsp; 46157c478bd9Sstevel@tonic-gate caddr_t addr2 = addr + (io_off - off); 46167c478bd9Sstevel@tonic-gate struct buf *bp; 46177c478bd9Sstevel@tonic-gate daddr_t bn; 46187c478bd9Sstevel@tonic-gate size_t io_len; 46197c478bd9Sstevel@tonic-gate int contig; 46207c478bd9Sstevel@tonic-gate int xlen; 46217c478bd9Sstevel@tonic-gate int bsize = ip->i_fs->fs_bsize; 46227c478bd9Sstevel@tonic-gate 46237c478bd9Sstevel@tonic-gate /* 46247c478bd9Sstevel@tonic-gate * If the directio advisory is in effect on this file, 46257c478bd9Sstevel@tonic-gate * then do not do buffered read ahead. Read ahead makes 46267c478bd9Sstevel@tonic-gate * it more difficult on threads using directio as they 46277c478bd9Sstevel@tonic-gate * will be forced to flush the pages from this vnode. 46287c478bd9Sstevel@tonic-gate */ 46297c478bd9Sstevel@tonic-gate if ((ufsvfsp = ip->i_ufsvfs) == NULL) 46307c478bd9Sstevel@tonic-gate return (0); 46317c478bd9Sstevel@tonic-gate if (ip->i_flag & IDIRECTIO || ufsvfsp->vfs_forcedirectio) 46327c478bd9Sstevel@tonic-gate return (0); 46337c478bd9Sstevel@tonic-gate 46347c478bd9Sstevel@tonic-gate /* 46357c478bd9Sstevel@tonic-gate * Is this test needed? 46367c478bd9Sstevel@tonic-gate */ 46377c478bd9Sstevel@tonic-gate if (addr2 >= seg->s_base + seg->s_size) 46387c478bd9Sstevel@tonic-gate return (0); 46397c478bd9Sstevel@tonic-gate 46407c478bd9Sstevel@tonic-gate contig = 0; 46417c478bd9Sstevel@tonic-gate if (bmap_read(ip, io_off, &bn, &contig) != 0 || bn == UFS_HOLE) 46427c478bd9Sstevel@tonic-gate return (0); 46437c478bd9Sstevel@tonic-gate 46447c478bd9Sstevel@tonic-gate /* 46457c478bd9Sstevel@tonic-gate * Limit the transfer size to bsize if this is the 2nd block. 46467c478bd9Sstevel@tonic-gate */ 46477c478bd9Sstevel@tonic-gate if (io_off == (u_offset_t)bsize) 46487c478bd9Sstevel@tonic-gate contig = MIN(contig, bsize); 46497c478bd9Sstevel@tonic-gate 46507c478bd9Sstevel@tonic-gate if ((pp = pvn_read_kluster(vp, io_off, seg, addr2, &io_off, 46517c478bd9Sstevel@tonic-gate &io_len, io_off, contig, 1)) == NULL) 46527c478bd9Sstevel@tonic-gate return (0); 46537c478bd9Sstevel@tonic-gate 46547c478bd9Sstevel@tonic-gate /* 46557c478bd9Sstevel@tonic-gate * Zero part of page which we are not going to read from disk 46567c478bd9Sstevel@tonic-gate */ 46577c478bd9Sstevel@tonic-gate if ((xlen = (io_len & PAGEOFFSET)) > 0) 46587c478bd9Sstevel@tonic-gate pagezero(pp->p_prev, xlen, PAGESIZE - xlen); 46597c478bd9Sstevel@tonic-gate 46607c478bd9Sstevel@tonic-gate ip->i_nextrio = (io_off + io_len + PAGESIZE - 1) & PAGEMASK; 46617c478bd9Sstevel@tonic-gate 46627c478bd9Sstevel@tonic-gate bp = pageio_setup(pp, io_len, ip->i_devvp, B_READ | B_ASYNC); 46637c478bd9Sstevel@tonic-gate bp->b_edev = ip->i_dev; 46647c478bd9Sstevel@tonic-gate bp->b_dev = cmpdev(ip->i_dev); 46657c478bd9Sstevel@tonic-gate bp->b_blkno = bn; 46667c478bd9Sstevel@tonic-gate bp->b_un.b_addr = (caddr_t)0; 46677c478bd9Sstevel@tonic-gate bp->b_file = ip->i_vnode; 46687c478bd9Sstevel@tonic-gate bp->b_offset = off; 46697c478bd9Sstevel@tonic-gate 46707c478bd9Sstevel@tonic-gate if (ufsvfsp->vfs_log) { 46717c478bd9Sstevel@tonic-gate lufs_read_strategy(ufsvfsp->vfs_log, bp); 46727c478bd9Sstevel@tonic-gate } else if (ufsvfsp->vfs_snapshot) { 46737c478bd9Sstevel@tonic-gate fssnap_strategy(&ufsvfsp->vfs_snapshot, bp); 46747c478bd9Sstevel@tonic-gate } else { 46757c478bd9Sstevel@tonic-gate ufsvfsp->vfs_iotstamp = lbolt; 46767c478bd9Sstevel@tonic-gate ub.ub_getras.value.ul++; 46777c478bd9Sstevel@tonic-gate (void) bdev_strategy(bp); 46787c478bd9Sstevel@tonic-gate lwp_stat_update(LWP_STAT_INBLK, 1); 46797c478bd9Sstevel@tonic-gate } 46807c478bd9Sstevel@tonic-gate 46817c478bd9Sstevel@tonic-gate return (io_len); 46827c478bd9Sstevel@tonic-gate } 46837c478bd9Sstevel@tonic-gate 46847c478bd9Sstevel@tonic-gate int ufs_delay = 1; 46857c478bd9Sstevel@tonic-gate /* 46867c478bd9Sstevel@tonic-gate * Flags are composed of {B_INVAL, B_FREE, B_DONTNEED, B_FORCE, B_ASYNC} 46877c478bd9Sstevel@tonic-gate * 46887c478bd9Sstevel@tonic-gate * LMXXX - the inode really ought to contain a pointer to one of these 46897c478bd9Sstevel@tonic-gate * async args. Stuff gunk in there and just hand the whole mess off. 46907c478bd9Sstevel@tonic-gate * This would replace i_delaylen, i_delayoff. 46917c478bd9Sstevel@tonic-gate */ 46927c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 46937c478bd9Sstevel@tonic-gate static int 46947c478bd9Sstevel@tonic-gate ufs_putpage(struct vnode *vp, offset_t off, size_t len, int flags, 46957c478bd9Sstevel@tonic-gate struct cred *cr) 46967c478bd9Sstevel@tonic-gate { 46977c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 46987c478bd9Sstevel@tonic-gate int err = 0; 46997c478bd9Sstevel@tonic-gate 47007c478bd9Sstevel@tonic-gate if (vp->v_count == 0) { 47017c478bd9Sstevel@tonic-gate return (ufs_fault(vp, "ufs_putpage: bad v_count == 0")); 47027c478bd9Sstevel@tonic-gate } 47037c478bd9Sstevel@tonic-gate 47047c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_PUTPAGE_START, 47057c478bd9Sstevel@tonic-gate "ufs_putpage_start:vp %p", vp); 47067c478bd9Sstevel@tonic-gate 47077c478bd9Sstevel@tonic-gate /* 47087c478bd9Sstevel@tonic-gate * XXX - Why should this check be made here? 47097c478bd9Sstevel@tonic-gate */ 47107c478bd9Sstevel@tonic-gate if (vp->v_flag & VNOMAP) { 47117c478bd9Sstevel@tonic-gate err = ENOSYS; 47127c478bd9Sstevel@tonic-gate goto errout; 47137c478bd9Sstevel@tonic-gate } 47147c478bd9Sstevel@tonic-gate 47157c478bd9Sstevel@tonic-gate if (ip->i_ufsvfs == NULL) { 47167c478bd9Sstevel@tonic-gate err = EIO; 47177c478bd9Sstevel@tonic-gate goto errout; 47187c478bd9Sstevel@tonic-gate } 47197c478bd9Sstevel@tonic-gate 47207c478bd9Sstevel@tonic-gate if (flags & B_ASYNC) { 47217c478bd9Sstevel@tonic-gate if (ufs_delay && len && 47227c478bd9Sstevel@tonic-gate (flags & ~(B_ASYNC|B_DONTNEED|B_FREE)) == 0) { 47237c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 47247c478bd9Sstevel@tonic-gate /* 47257c478bd9Sstevel@tonic-gate * If nobody stalled, start a new cluster. 47267c478bd9Sstevel@tonic-gate */ 47277c478bd9Sstevel@tonic-gate if (ip->i_delaylen == 0) { 47287c478bd9Sstevel@tonic-gate ip->i_delayoff = off; 47297c478bd9Sstevel@tonic-gate ip->i_delaylen = len; 47307c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 47317c478bd9Sstevel@tonic-gate goto errout; 47327c478bd9Sstevel@tonic-gate } 47337c478bd9Sstevel@tonic-gate /* 47347c478bd9Sstevel@tonic-gate * If we have a full cluster or they are not contig, 47357c478bd9Sstevel@tonic-gate * then push last cluster and start over. 47367c478bd9Sstevel@tonic-gate */ 47377c478bd9Sstevel@tonic-gate if (ip->i_delaylen >= CLUSTSZ(ip) || 47387c478bd9Sstevel@tonic-gate ip->i_delayoff + ip->i_delaylen != off) { 47397c478bd9Sstevel@tonic-gate u_offset_t doff; 47407c478bd9Sstevel@tonic-gate size_t dlen; 47417c478bd9Sstevel@tonic-gate 47427c478bd9Sstevel@tonic-gate doff = ip->i_delayoff; 47437c478bd9Sstevel@tonic-gate dlen = ip->i_delaylen; 47447c478bd9Sstevel@tonic-gate ip->i_delayoff = off; 47457c478bd9Sstevel@tonic-gate ip->i_delaylen = len; 47467c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 47477c478bd9Sstevel@tonic-gate err = ufs_putpages(vp, doff, dlen, 47487c478bd9Sstevel@tonic-gate flags, cr); 47497c478bd9Sstevel@tonic-gate /* LMXXX - flags are new val, not old */ 47507c478bd9Sstevel@tonic-gate goto errout; 47517c478bd9Sstevel@tonic-gate } 47527c478bd9Sstevel@tonic-gate /* 47537c478bd9Sstevel@tonic-gate * There is something there, it's not full, and 47547c478bd9Sstevel@tonic-gate * it is contig. 47557c478bd9Sstevel@tonic-gate */ 47567c478bd9Sstevel@tonic-gate ip->i_delaylen += len; 47577c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 47587c478bd9Sstevel@tonic-gate goto errout; 47597c478bd9Sstevel@tonic-gate } 47607c478bd9Sstevel@tonic-gate /* 47617c478bd9Sstevel@tonic-gate * Must have weird flags or we are not clustering. 47627c478bd9Sstevel@tonic-gate */ 47637c478bd9Sstevel@tonic-gate } 47647c478bd9Sstevel@tonic-gate 47657c478bd9Sstevel@tonic-gate err = ufs_putpages(vp, off, len, flags, cr); 47667c478bd9Sstevel@tonic-gate 47677c478bd9Sstevel@tonic-gate errout: 47687c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_PUTPAGE_END, 47697c478bd9Sstevel@tonic-gate "ufs_putpage_end:vp %p error %d", vp, err); 47707c478bd9Sstevel@tonic-gate return (err); 47717c478bd9Sstevel@tonic-gate } 47727c478bd9Sstevel@tonic-gate 47737c478bd9Sstevel@tonic-gate /* 47747c478bd9Sstevel@tonic-gate * If len == 0, do from off to EOF. 47757c478bd9Sstevel@tonic-gate * 47767c478bd9Sstevel@tonic-gate * The normal cases should be len == 0 & off == 0 (entire vp list), 47777c478bd9Sstevel@tonic-gate * len == MAXBSIZE (from segmap_release actions), and len == PAGESIZE 47787c478bd9Sstevel@tonic-gate * (from pageout). 47797c478bd9Sstevel@tonic-gate */ 47807c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 47817c478bd9Sstevel@tonic-gate static int 47827c478bd9Sstevel@tonic-gate ufs_putpages( 47837c478bd9Sstevel@tonic-gate struct vnode *vp, 47847c478bd9Sstevel@tonic-gate offset_t off, 47857c478bd9Sstevel@tonic-gate size_t len, 47867c478bd9Sstevel@tonic-gate int flags, 47877c478bd9Sstevel@tonic-gate struct cred *cr) 47887c478bd9Sstevel@tonic-gate { 47897c478bd9Sstevel@tonic-gate u_offset_t io_off; 47907c478bd9Sstevel@tonic-gate u_offset_t eoff; 47917c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 47927c478bd9Sstevel@tonic-gate page_t *pp; 47937c478bd9Sstevel@tonic-gate size_t io_len; 47947c478bd9Sstevel@tonic-gate int err = 0; 47957c478bd9Sstevel@tonic-gate int dolock; 47967c478bd9Sstevel@tonic-gate 47977c478bd9Sstevel@tonic-gate if (vp->v_count == 0) 47987c478bd9Sstevel@tonic-gate return (ufs_fault(vp, "ufs_putpages: v_count == 0")); 47997c478bd9Sstevel@tonic-gate /* 48007c478bd9Sstevel@tonic-gate * Acquire the readers/write inode lock before locking 48017c478bd9Sstevel@tonic-gate * any pages in this inode. 48027c478bd9Sstevel@tonic-gate * The inode lock is held during i/o. 48037c478bd9Sstevel@tonic-gate */ 48047c478bd9Sstevel@tonic-gate if (len == 0) { 48057c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 48067c478bd9Sstevel@tonic-gate ip->i_delayoff = ip->i_delaylen = 0; 48077c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 48087c478bd9Sstevel@tonic-gate } 48097c478bd9Sstevel@tonic-gate dolock = (rw_owner(&ip->i_contents) != curthread); 48107c478bd9Sstevel@tonic-gate if (dolock) { 48117c478bd9Sstevel@tonic-gate /* 48127c478bd9Sstevel@tonic-gate * Must synchronize this thread and any possible thread 48137c478bd9Sstevel@tonic-gate * operating in the window of vulnerability in wrip(). 48147c478bd9Sstevel@tonic-gate * It is dangerous to allow both a thread doing a putpage 48157c478bd9Sstevel@tonic-gate * and a thread writing, so serialize them. The exception 48167c478bd9Sstevel@tonic-gate * is when the thread in wrip() does something which causes 48177c478bd9Sstevel@tonic-gate * a putpage operation. Then, the thread must be allowed 48187c478bd9Sstevel@tonic-gate * to continue. It may encounter a bmap_read problem in 48197c478bd9Sstevel@tonic-gate * ufs_putapage, but that is handled in ufs_putapage. 48207c478bd9Sstevel@tonic-gate * Allow async writers to proceed, we don't want to block 48217c478bd9Sstevel@tonic-gate * the pageout daemon. 48227c478bd9Sstevel@tonic-gate */ 48237c478bd9Sstevel@tonic-gate if (ip->i_writer == curthread) 48247c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_READER); 48257c478bd9Sstevel@tonic-gate else { 48267c478bd9Sstevel@tonic-gate for (;;) { 48277c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_READER); 48287c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 48297c478bd9Sstevel@tonic-gate /* 48307c478bd9Sstevel@tonic-gate * If there is no thread in the critical 48317c478bd9Sstevel@tonic-gate * section of wrip(), then proceed. 48327c478bd9Sstevel@tonic-gate * Otherwise, wait until there isn't one. 48337c478bd9Sstevel@tonic-gate */ 48347c478bd9Sstevel@tonic-gate if (ip->i_writer == NULL) { 48357c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 48367c478bd9Sstevel@tonic-gate break; 48377c478bd9Sstevel@tonic-gate } 48387c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 48397c478bd9Sstevel@tonic-gate /* 48407c478bd9Sstevel@tonic-gate * Bounce async writers when we have a writer 48417c478bd9Sstevel@tonic-gate * working on this file so we don't deadlock 48427c478bd9Sstevel@tonic-gate * the pageout daemon. 48437c478bd9Sstevel@tonic-gate */ 48447c478bd9Sstevel@tonic-gate if (flags & B_ASYNC) { 48457c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 48467c478bd9Sstevel@tonic-gate return (0); 48477c478bd9Sstevel@tonic-gate } 48487c478bd9Sstevel@tonic-gate cv_wait(&ip->i_wrcv, &ip->i_tlock); 48497c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 48507c478bd9Sstevel@tonic-gate } 48517c478bd9Sstevel@tonic-gate } 48527c478bd9Sstevel@tonic-gate } 48537c478bd9Sstevel@tonic-gate 48547c478bd9Sstevel@tonic-gate if (!vn_has_cached_data(vp)) { 48557c478bd9Sstevel@tonic-gate if (dolock) 48567c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 48577c478bd9Sstevel@tonic-gate return (0); 48587c478bd9Sstevel@tonic-gate } 48597c478bd9Sstevel@tonic-gate 48607c478bd9Sstevel@tonic-gate if (len == 0) { 48617c478bd9Sstevel@tonic-gate /* 48627c478bd9Sstevel@tonic-gate * Search the entire vp list for pages >= off. 48637c478bd9Sstevel@tonic-gate */ 48647c478bd9Sstevel@tonic-gate err = pvn_vplist_dirty(vp, (u_offset_t)off, ufs_putapage, 48657c478bd9Sstevel@tonic-gate flags, cr); 48667c478bd9Sstevel@tonic-gate } else { 48677c478bd9Sstevel@tonic-gate /* 48687c478bd9Sstevel@tonic-gate * Loop over all offsets in the range looking for 48697c478bd9Sstevel@tonic-gate * pages to deal with. 48707c478bd9Sstevel@tonic-gate */ 48717c478bd9Sstevel@tonic-gate if ((eoff = blkroundup(ip->i_fs, ip->i_size)) != 0) 48727c478bd9Sstevel@tonic-gate eoff = MIN(off + len, eoff); 48737c478bd9Sstevel@tonic-gate else 48747c478bd9Sstevel@tonic-gate eoff = off + len; 48757c478bd9Sstevel@tonic-gate 48767c478bd9Sstevel@tonic-gate for (io_off = off; io_off < eoff; io_off += io_len) { 48777c478bd9Sstevel@tonic-gate /* 48787c478bd9Sstevel@tonic-gate * If we are not invalidating, synchronously 48797c478bd9Sstevel@tonic-gate * freeing or writing pages, use the routine 48807c478bd9Sstevel@tonic-gate * page_lookup_nowait() to prevent reclaiming 48817c478bd9Sstevel@tonic-gate * them from the free list. 48827c478bd9Sstevel@tonic-gate */ 48837c478bd9Sstevel@tonic-gate if ((flags & B_INVAL) || ((flags & B_ASYNC) == 0)) { 48847c478bd9Sstevel@tonic-gate pp = page_lookup(vp, io_off, 48857c478bd9Sstevel@tonic-gate (flags & (B_INVAL | B_FREE)) ? 48867c478bd9Sstevel@tonic-gate SE_EXCL : SE_SHARED); 48877c478bd9Sstevel@tonic-gate } else { 48887c478bd9Sstevel@tonic-gate pp = page_lookup_nowait(vp, io_off, 48897c478bd9Sstevel@tonic-gate (flags & B_FREE) ? SE_EXCL : SE_SHARED); 48907c478bd9Sstevel@tonic-gate } 48917c478bd9Sstevel@tonic-gate 48927c478bd9Sstevel@tonic-gate if (pp == NULL || pvn_getdirty(pp, flags) == 0) 48937c478bd9Sstevel@tonic-gate io_len = PAGESIZE; 48947c478bd9Sstevel@tonic-gate else { 48957c478bd9Sstevel@tonic-gate u_offset_t *io_offp = &io_off; 48967c478bd9Sstevel@tonic-gate 48977c478bd9Sstevel@tonic-gate err = ufs_putapage(vp, pp, io_offp, &io_len, 48987c478bd9Sstevel@tonic-gate flags, cr); 48997c478bd9Sstevel@tonic-gate if (err != 0) 49007c478bd9Sstevel@tonic-gate break; 49017c478bd9Sstevel@tonic-gate /* 49027c478bd9Sstevel@tonic-gate * "io_off" and "io_len" are returned as 49037c478bd9Sstevel@tonic-gate * the range of pages we actually wrote. 49047c478bd9Sstevel@tonic-gate * This allows us to skip ahead more quickly 49057c478bd9Sstevel@tonic-gate * since several pages may've been dealt 49067c478bd9Sstevel@tonic-gate * with by this iteration of the loop. 49077c478bd9Sstevel@tonic-gate */ 49087c478bd9Sstevel@tonic-gate } 49097c478bd9Sstevel@tonic-gate } 49107c478bd9Sstevel@tonic-gate } 49117c478bd9Sstevel@tonic-gate if (err == 0 && off == 0 && (len == 0 || len >= ip->i_size)) { 49127c478bd9Sstevel@tonic-gate /* 49137c478bd9Sstevel@tonic-gate * We have just sync'ed back all the pages on 49147c478bd9Sstevel@tonic-gate * the inode, turn off the IMODTIME flag. 49157c478bd9Sstevel@tonic-gate */ 49167c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 49177c478bd9Sstevel@tonic-gate ip->i_flag &= ~IMODTIME; 49187c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 49197c478bd9Sstevel@tonic-gate } 49207c478bd9Sstevel@tonic-gate if (dolock) 49217c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 49227c478bd9Sstevel@tonic-gate return (err); 49237c478bd9Sstevel@tonic-gate } 49247c478bd9Sstevel@tonic-gate 49257c478bd9Sstevel@tonic-gate static void 49267c478bd9Sstevel@tonic-gate ufs_iodone(buf_t *bp) 49277c478bd9Sstevel@tonic-gate { 49287c478bd9Sstevel@tonic-gate struct inode *ip; 49297c478bd9Sstevel@tonic-gate 49307c478bd9Sstevel@tonic-gate ASSERT((bp->b_pages->p_vnode != NULL) && !(bp->b_flags & B_READ)); 49317c478bd9Sstevel@tonic-gate 49327c478bd9Sstevel@tonic-gate bp->b_iodone = NULL; 49337c478bd9Sstevel@tonic-gate 49347c478bd9Sstevel@tonic-gate ip = VTOI(bp->b_pages->p_vnode); 49357c478bd9Sstevel@tonic-gate 49367c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 49377c478bd9Sstevel@tonic-gate if (ip->i_writes >= ufs_LW) { 49387c478bd9Sstevel@tonic-gate if ((ip->i_writes -= bp->b_bcount) <= ufs_LW) 49397c478bd9Sstevel@tonic-gate if (ufs_WRITES) 49407c478bd9Sstevel@tonic-gate cv_broadcast(&ip->i_wrcv); /* wake all up */ 49417c478bd9Sstevel@tonic-gate } else { 49427c478bd9Sstevel@tonic-gate ip->i_writes -= bp->b_bcount; 49437c478bd9Sstevel@tonic-gate } 49447c478bd9Sstevel@tonic-gate 49457c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 49467c478bd9Sstevel@tonic-gate iodone(bp); 49477c478bd9Sstevel@tonic-gate } 49487c478bd9Sstevel@tonic-gate 49497c478bd9Sstevel@tonic-gate /* 49507c478bd9Sstevel@tonic-gate * Write out a single page, possibly klustering adjacent 49517c478bd9Sstevel@tonic-gate * dirty pages. The inode lock must be held. 49527c478bd9Sstevel@tonic-gate * 49537c478bd9Sstevel@tonic-gate * LMXXX - bsize < pagesize not done. 49547c478bd9Sstevel@tonic-gate */ 49557c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 49567c478bd9Sstevel@tonic-gate int 49577c478bd9Sstevel@tonic-gate ufs_putapage( 49587c478bd9Sstevel@tonic-gate struct vnode *vp, 49597c478bd9Sstevel@tonic-gate page_t *pp, 49607c478bd9Sstevel@tonic-gate u_offset_t *offp, 49617c478bd9Sstevel@tonic-gate size_t *lenp, /* return values */ 49627c478bd9Sstevel@tonic-gate int flags, 49637c478bd9Sstevel@tonic-gate struct cred *cr) 49647c478bd9Sstevel@tonic-gate { 49657c478bd9Sstevel@tonic-gate u_offset_t io_off; 49667c478bd9Sstevel@tonic-gate u_offset_t off; 49677c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 49687c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp = ip->i_ufsvfs; 49697c478bd9Sstevel@tonic-gate struct fs *fs; 49707c478bd9Sstevel@tonic-gate struct buf *bp; 49717c478bd9Sstevel@tonic-gate size_t io_len; 49727c478bd9Sstevel@tonic-gate daddr_t bn; 49737c478bd9Sstevel@tonic-gate int err; 49747c478bd9Sstevel@tonic-gate int contig; 49757c478bd9Sstevel@tonic-gate 49767c478bd9Sstevel@tonic-gate ASSERT(RW_LOCK_HELD(&ip->i_contents)); 49777c478bd9Sstevel@tonic-gate 49787c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_PUTAPAGE_START, 49797c478bd9Sstevel@tonic-gate "ufs_putapage_start:vp %p", vp); 49807c478bd9Sstevel@tonic-gate 49817c478bd9Sstevel@tonic-gate if (ufsvfsp == NULL) { 49827c478bd9Sstevel@tonic-gate err = EIO; 49837c478bd9Sstevel@tonic-gate goto out_trace; 49847c478bd9Sstevel@tonic-gate } 49857c478bd9Sstevel@tonic-gate 49867c478bd9Sstevel@tonic-gate fs = ip->i_fs; 49877c478bd9Sstevel@tonic-gate ASSERT(fs->fs_ronly == 0); 49887c478bd9Sstevel@tonic-gate 49897c478bd9Sstevel@tonic-gate /* 49907c478bd9Sstevel@tonic-gate * If the modified time on the inode has not already been 49917c478bd9Sstevel@tonic-gate * set elsewhere (e.g. for write/setattr) we set the time now. 49927c478bd9Sstevel@tonic-gate * This gives us approximate modified times for mmap'ed files 49937c478bd9Sstevel@tonic-gate * which are modified via stores in the user address space. 49947c478bd9Sstevel@tonic-gate */ 49957c478bd9Sstevel@tonic-gate if ((ip->i_flag & IMODTIME) == 0) { 49967c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 49977c478bd9Sstevel@tonic-gate ip->i_flag |= IUPD; 49987c478bd9Sstevel@tonic-gate ip->i_seq++; 49997c478bd9Sstevel@tonic-gate ITIMES_NOLOCK(ip); 50007c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 50017c478bd9Sstevel@tonic-gate } 50027c478bd9Sstevel@tonic-gate 50037c478bd9Sstevel@tonic-gate /* 50047c478bd9Sstevel@tonic-gate * Align the request to a block boundry (for old file systems), 50057c478bd9Sstevel@tonic-gate * and go ask bmap() how contiguous things are for this file. 50067c478bd9Sstevel@tonic-gate */ 50077c478bd9Sstevel@tonic-gate off = pp->p_offset & (offset_t)fs->fs_bmask; /* block align it */ 50087c478bd9Sstevel@tonic-gate contig = 0; 50097c478bd9Sstevel@tonic-gate err = bmap_read(ip, off, &bn, &contig); 50107c478bd9Sstevel@tonic-gate if (err) 50117c478bd9Sstevel@tonic-gate goto out; 50127c478bd9Sstevel@tonic-gate if (bn == UFS_HOLE) { /* putpage never allocates */ 50137c478bd9Sstevel@tonic-gate /* 50147c478bd9Sstevel@tonic-gate * logging device is in error mode; simply return EIO 50157c478bd9Sstevel@tonic-gate */ 50167c478bd9Sstevel@tonic-gate if (TRANS_ISERROR(ufsvfsp)) { 50177c478bd9Sstevel@tonic-gate err = EIO; 50187c478bd9Sstevel@tonic-gate goto out; 50197c478bd9Sstevel@tonic-gate } 50207c478bd9Sstevel@tonic-gate /* 50217c478bd9Sstevel@tonic-gate * Oops, the thread in the window in wrip() did some 50227c478bd9Sstevel@tonic-gate * sort of operation which caused a putpage in the bad 50237c478bd9Sstevel@tonic-gate * range. In this case, just return an error which will 50247c478bd9Sstevel@tonic-gate * cause the software modified bit on the page to set 50257c478bd9Sstevel@tonic-gate * and the page will get written out again later. 50267c478bd9Sstevel@tonic-gate */ 50277c478bd9Sstevel@tonic-gate if (ip->i_writer == curthread) { 50287c478bd9Sstevel@tonic-gate err = EIO; 50297c478bd9Sstevel@tonic-gate goto out; 50307c478bd9Sstevel@tonic-gate } 50317c478bd9Sstevel@tonic-gate /* 50327c478bd9Sstevel@tonic-gate * If the pager is trying to push a page in the bad range 50337c478bd9Sstevel@tonic-gate * just tell him to try again later when things are better. 50347c478bd9Sstevel@tonic-gate */ 50357c478bd9Sstevel@tonic-gate if (flags & B_ASYNC) { 50367c478bd9Sstevel@tonic-gate err = EAGAIN; 50377c478bd9Sstevel@tonic-gate goto out; 50387c478bd9Sstevel@tonic-gate } 50397c478bd9Sstevel@tonic-gate err = ufs_fault(ITOV(ip), "ufs_putapage: bn == UFS_HOLE"); 50407c478bd9Sstevel@tonic-gate goto out; 50417c478bd9Sstevel@tonic-gate } 50427c478bd9Sstevel@tonic-gate 50437c478bd9Sstevel@tonic-gate /* 50447c478bd9Sstevel@tonic-gate * Take the length (of contiguous bytes) passed back from bmap() 50457c478bd9Sstevel@tonic-gate * and _try_ and get a set of pages covering that extent. 50467c478bd9Sstevel@tonic-gate */ 50477c478bd9Sstevel@tonic-gate pp = pvn_write_kluster(vp, pp, &io_off, &io_len, off, contig, flags); 50487c478bd9Sstevel@tonic-gate 50497c478bd9Sstevel@tonic-gate /* 50507c478bd9Sstevel@tonic-gate * May have run out of memory and not clustered backwards. 50517c478bd9Sstevel@tonic-gate * off p_offset 50527c478bd9Sstevel@tonic-gate * [ pp - 1 ][ pp ] 50537c478bd9Sstevel@tonic-gate * [ block ] 50547c478bd9Sstevel@tonic-gate * We told bmap off, so we have to adjust the bn accordingly. 50557c478bd9Sstevel@tonic-gate */ 50567c478bd9Sstevel@tonic-gate if (io_off > off) { 50577c478bd9Sstevel@tonic-gate bn += btod(io_off - off); 50587c478bd9Sstevel@tonic-gate contig -= (io_off - off); 50597c478bd9Sstevel@tonic-gate } 50607c478bd9Sstevel@tonic-gate 50617c478bd9Sstevel@tonic-gate /* 50627c478bd9Sstevel@tonic-gate * bmap was carefull to tell us the right size so use that. 50637c478bd9Sstevel@tonic-gate * There might be unallocated frags at the end. 50647c478bd9Sstevel@tonic-gate * LMXXX - bzero the end of the page? We must be writing after EOF. 50657c478bd9Sstevel@tonic-gate */ 50667c478bd9Sstevel@tonic-gate if (io_len > contig) { 50677c478bd9Sstevel@tonic-gate ASSERT(io_len - contig < fs->fs_bsize); 50687c478bd9Sstevel@tonic-gate io_len -= (io_len - contig); 50697c478bd9Sstevel@tonic-gate } 50707c478bd9Sstevel@tonic-gate 50717c478bd9Sstevel@tonic-gate /* 50727c478bd9Sstevel@tonic-gate * Handle the case where we are writing the last page after EOF. 50737c478bd9Sstevel@tonic-gate * 50747c478bd9Sstevel@tonic-gate * XXX - just a patch for i-mt3. 50757c478bd9Sstevel@tonic-gate */ 50767c478bd9Sstevel@tonic-gate if (io_len == 0) { 50777c478bd9Sstevel@tonic-gate ASSERT(pp->p_offset >= (u_offset_t)(roundup(ip->i_size, 50787c478bd9Sstevel@tonic-gate PAGESIZE))); 50797c478bd9Sstevel@tonic-gate io_len = PAGESIZE; 50807c478bd9Sstevel@tonic-gate } 50817c478bd9Sstevel@tonic-gate 50827c478bd9Sstevel@tonic-gate bp = pageio_setup(pp, io_len, ip->i_devvp, B_WRITE | flags); 50837c478bd9Sstevel@tonic-gate 50847c478bd9Sstevel@tonic-gate ULOCKFS_SET_MOD(ITOUL(ip)); 50857c478bd9Sstevel@tonic-gate 50867c478bd9Sstevel@tonic-gate bp->b_edev = ip->i_dev; 50877c478bd9Sstevel@tonic-gate bp->b_dev = cmpdev(ip->i_dev); 50887c478bd9Sstevel@tonic-gate bp->b_blkno = bn; 50897c478bd9Sstevel@tonic-gate bp->b_un.b_addr = (caddr_t)0; 50907c478bd9Sstevel@tonic-gate bp->b_file = ip->i_vnode; 50917c478bd9Sstevel@tonic-gate 50927c478bd9Sstevel@tonic-gate if (TRANS_ISTRANS(ufsvfsp)) { 50937c478bd9Sstevel@tonic-gate if ((ip->i_mode & IFMT) == IFSHAD) { 50947c478bd9Sstevel@tonic-gate TRANS_BUF(ufsvfsp, 0, io_len, bp, DT_SHAD); 50957c478bd9Sstevel@tonic-gate } else if (ufsvfsp->vfs_qinod == ip) { 50967c478bd9Sstevel@tonic-gate TRANS_DELTA(ufsvfsp, ldbtob(bn), bp->b_bcount, DT_QR, 50977c478bd9Sstevel@tonic-gate 0, 0); 50987c478bd9Sstevel@tonic-gate } 50997c478bd9Sstevel@tonic-gate } 51007c478bd9Sstevel@tonic-gate 51017c478bd9Sstevel@tonic-gate /* write throttle */ 51027c478bd9Sstevel@tonic-gate 51037c478bd9Sstevel@tonic-gate ASSERT(bp->b_iodone == NULL); 51047c478bd9Sstevel@tonic-gate bp->b_iodone = (int (*)())ufs_iodone; 51057c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 51067c478bd9Sstevel@tonic-gate ip->i_writes += bp->b_bcount; 51077c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 51087c478bd9Sstevel@tonic-gate 51097c478bd9Sstevel@tonic-gate if (bp->b_flags & B_ASYNC) { 51107c478bd9Sstevel@tonic-gate if (ufsvfsp->vfs_log) { 51117c478bd9Sstevel@tonic-gate lufs_write_strategy(ufsvfsp->vfs_log, bp); 51127c478bd9Sstevel@tonic-gate } else if (ufsvfsp->vfs_snapshot) { 51137c478bd9Sstevel@tonic-gate fssnap_strategy(&ufsvfsp->vfs_snapshot, bp); 51147c478bd9Sstevel@tonic-gate } else { 51157c478bd9Sstevel@tonic-gate ufsvfsp->vfs_iotstamp = lbolt; 51167c478bd9Sstevel@tonic-gate ub.ub_putasyncs.value.ul++; 51177c478bd9Sstevel@tonic-gate (void) bdev_strategy(bp); 51187c478bd9Sstevel@tonic-gate lwp_stat_update(LWP_STAT_OUBLK, 1); 51197c478bd9Sstevel@tonic-gate } 51207c478bd9Sstevel@tonic-gate } else { 51217c478bd9Sstevel@tonic-gate if (ufsvfsp->vfs_log) { 51227c478bd9Sstevel@tonic-gate lufs_write_strategy(ufsvfsp->vfs_log, bp); 51237c478bd9Sstevel@tonic-gate } else if (ufsvfsp->vfs_snapshot) { 51247c478bd9Sstevel@tonic-gate fssnap_strategy(&ufsvfsp->vfs_snapshot, bp); 51257c478bd9Sstevel@tonic-gate } else { 51267c478bd9Sstevel@tonic-gate ufsvfsp->vfs_iotstamp = lbolt; 51277c478bd9Sstevel@tonic-gate ub.ub_putsyncs.value.ul++; 51287c478bd9Sstevel@tonic-gate (void) bdev_strategy(bp); 51297c478bd9Sstevel@tonic-gate lwp_stat_update(LWP_STAT_OUBLK, 1); 51307c478bd9Sstevel@tonic-gate } 51317c478bd9Sstevel@tonic-gate err = biowait(bp); 51327c478bd9Sstevel@tonic-gate pageio_done(bp); 51337c478bd9Sstevel@tonic-gate pvn_write_done(pp, ((err) ? B_ERROR : 0) | B_WRITE | flags); 51347c478bd9Sstevel@tonic-gate } 51357c478bd9Sstevel@tonic-gate 51367c478bd9Sstevel@tonic-gate pp = NULL; 51377c478bd9Sstevel@tonic-gate 51387c478bd9Sstevel@tonic-gate out: 51397c478bd9Sstevel@tonic-gate if (err != 0 && pp != NULL) 51407c478bd9Sstevel@tonic-gate pvn_write_done(pp, B_ERROR | B_WRITE | flags); 51417c478bd9Sstevel@tonic-gate 51427c478bd9Sstevel@tonic-gate if (offp) 51437c478bd9Sstevel@tonic-gate *offp = io_off; 51447c478bd9Sstevel@tonic-gate if (lenp) 51457c478bd9Sstevel@tonic-gate *lenp = io_len; 51467c478bd9Sstevel@tonic-gate out_trace: 51477c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_PUTAPAGE_END, 51487c478bd9Sstevel@tonic-gate "ufs_putapage_end:vp %p error %d", vp, err); 51497c478bd9Sstevel@tonic-gate return (err); 51507c478bd9Sstevel@tonic-gate } 51517c478bd9Sstevel@tonic-gate 51527c478bd9Sstevel@tonic-gate /* ARGSUSED */ 51537c478bd9Sstevel@tonic-gate static int 51547c478bd9Sstevel@tonic-gate ufs_map(struct vnode *vp, 51557c478bd9Sstevel@tonic-gate offset_t off, 51567c478bd9Sstevel@tonic-gate struct as *as, 51577c478bd9Sstevel@tonic-gate caddr_t *addrp, 51587c478bd9Sstevel@tonic-gate size_t len, 51597c478bd9Sstevel@tonic-gate uchar_t prot, 51607c478bd9Sstevel@tonic-gate uchar_t maxprot, 51617c478bd9Sstevel@tonic-gate uint_t flags, 51627c478bd9Sstevel@tonic-gate struct cred *cr) 51637c478bd9Sstevel@tonic-gate { 51647c478bd9Sstevel@tonic-gate struct segvn_crargs vn_a; 51657c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp = VTOI(vp)->i_ufsvfs; 51667c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 51677c478bd9Sstevel@tonic-gate int error; 51687c478bd9Sstevel@tonic-gate 51697c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_MAP_START, 51707c478bd9Sstevel@tonic-gate "ufs_map_start:vp %p", vp); 51717c478bd9Sstevel@tonic-gate 51727c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_MAP_MASK); 51737c478bd9Sstevel@tonic-gate if (error) 51747c478bd9Sstevel@tonic-gate goto out; 51757c478bd9Sstevel@tonic-gate 51767c478bd9Sstevel@tonic-gate if (vp->v_flag & VNOMAP) { 51777c478bd9Sstevel@tonic-gate error = ENOSYS; 51787c478bd9Sstevel@tonic-gate goto unlock; 51797c478bd9Sstevel@tonic-gate } 51807c478bd9Sstevel@tonic-gate 51817c478bd9Sstevel@tonic-gate if (off < (offset_t)0 || (offset_t)(off + len) < (offset_t)0) { 51827c478bd9Sstevel@tonic-gate error = ENXIO; 51837c478bd9Sstevel@tonic-gate goto unlock; 51847c478bd9Sstevel@tonic-gate } 51857c478bd9Sstevel@tonic-gate 51867c478bd9Sstevel@tonic-gate if (vp->v_type != VREG) { 51877c478bd9Sstevel@tonic-gate error = ENODEV; 51887c478bd9Sstevel@tonic-gate goto unlock; 51897c478bd9Sstevel@tonic-gate } 51907c478bd9Sstevel@tonic-gate 51917c478bd9Sstevel@tonic-gate /* 51927c478bd9Sstevel@tonic-gate * If file is being locked, disallow mapping. 51937c478bd9Sstevel@tonic-gate */ 51947c478bd9Sstevel@tonic-gate if (vn_has_mandatory_locks(vp, VTOI(vp)->i_mode)) { 51957c478bd9Sstevel@tonic-gate error = EAGAIN; 51967c478bd9Sstevel@tonic-gate goto unlock; 51977c478bd9Sstevel@tonic-gate } 51987c478bd9Sstevel@tonic-gate 51997c478bd9Sstevel@tonic-gate as_rangelock(as); 52007c478bd9Sstevel@tonic-gate if ((flags & MAP_FIXED) == 0) { 52017c478bd9Sstevel@tonic-gate map_addr(addrp, len, off, 1, flags); 52027c478bd9Sstevel@tonic-gate if (*addrp == NULL) { 52037c478bd9Sstevel@tonic-gate as_rangeunlock(as); 52047c478bd9Sstevel@tonic-gate error = ENOMEM; 52057c478bd9Sstevel@tonic-gate goto unlock; 52067c478bd9Sstevel@tonic-gate } 52077c478bd9Sstevel@tonic-gate } else { 52087c478bd9Sstevel@tonic-gate /* 52097c478bd9Sstevel@tonic-gate * User specified address - blow away any previous mappings 52107c478bd9Sstevel@tonic-gate */ 52117c478bd9Sstevel@tonic-gate (void) as_unmap(as, *addrp, len); 52127c478bd9Sstevel@tonic-gate } 52137c478bd9Sstevel@tonic-gate 52147c478bd9Sstevel@tonic-gate vn_a.vp = vp; 52157c478bd9Sstevel@tonic-gate vn_a.offset = (u_offset_t)off; 52167c478bd9Sstevel@tonic-gate vn_a.type = flags & MAP_TYPE; 52177c478bd9Sstevel@tonic-gate vn_a.prot = prot; 52187c478bd9Sstevel@tonic-gate vn_a.maxprot = maxprot; 52197c478bd9Sstevel@tonic-gate vn_a.cred = cr; 52207c478bd9Sstevel@tonic-gate vn_a.amp = NULL; 52217c478bd9Sstevel@tonic-gate vn_a.flags = flags & ~MAP_TYPE; 52227c478bd9Sstevel@tonic-gate vn_a.szc = 0; 52237c478bd9Sstevel@tonic-gate vn_a.lgrp_mem_policy_flags = 0; 52247c478bd9Sstevel@tonic-gate 52257c478bd9Sstevel@tonic-gate error = as_map(as, *addrp, len, segvn_create, &vn_a); 52267c478bd9Sstevel@tonic-gate as_rangeunlock(as); 52277c478bd9Sstevel@tonic-gate 52287c478bd9Sstevel@tonic-gate unlock: 52297c478bd9Sstevel@tonic-gate if (ulp) { 52307c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 52317c478bd9Sstevel@tonic-gate } 52327c478bd9Sstevel@tonic-gate out: 52337c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_MAP_END, 52347c478bd9Sstevel@tonic-gate "ufs_map_end:vp %p error %d", vp, error); 52357c478bd9Sstevel@tonic-gate return (error); 52367c478bd9Sstevel@tonic-gate } 52377c478bd9Sstevel@tonic-gate 52387c478bd9Sstevel@tonic-gate /* ARGSUSED */ 52397c478bd9Sstevel@tonic-gate static int 52407c478bd9Sstevel@tonic-gate ufs_addmap(struct vnode *vp, 52417c478bd9Sstevel@tonic-gate offset_t off, 52427c478bd9Sstevel@tonic-gate struct as *as, 52437c478bd9Sstevel@tonic-gate caddr_t addr, 52447c478bd9Sstevel@tonic-gate size_t len, 52457c478bd9Sstevel@tonic-gate uchar_t prot, 52467c478bd9Sstevel@tonic-gate uchar_t maxprot, 52477c478bd9Sstevel@tonic-gate uint_t flags, 52487c478bd9Sstevel@tonic-gate struct cred *cr) 52497c478bd9Sstevel@tonic-gate { 52507c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 52517c478bd9Sstevel@tonic-gate 52527c478bd9Sstevel@tonic-gate if (vp->v_flag & VNOMAP) { 52537c478bd9Sstevel@tonic-gate return (ENOSYS); 52547c478bd9Sstevel@tonic-gate } 52557c478bd9Sstevel@tonic-gate 52567c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 52577c478bd9Sstevel@tonic-gate ip->i_mapcnt += btopr(len); 52587c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 52597c478bd9Sstevel@tonic-gate return (0); 52607c478bd9Sstevel@tonic-gate } 52617c478bd9Sstevel@tonic-gate 52627c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 52637c478bd9Sstevel@tonic-gate static int 52647c478bd9Sstevel@tonic-gate ufs_delmap(struct vnode *vp, offset_t off, struct as *as, caddr_t addr, 52657c478bd9Sstevel@tonic-gate size_t len, uint_t prot, uint_t maxprot, uint_t flags, 52667c478bd9Sstevel@tonic-gate struct cred *cr) 52677c478bd9Sstevel@tonic-gate { 52687c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 52697c478bd9Sstevel@tonic-gate 52707c478bd9Sstevel@tonic-gate if (vp->v_flag & VNOMAP) { 52717c478bd9Sstevel@tonic-gate return (ENOSYS); 52727c478bd9Sstevel@tonic-gate } 52737c478bd9Sstevel@tonic-gate 52747c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 52757c478bd9Sstevel@tonic-gate ip->i_mapcnt -= btopr(len); /* Count released mappings */ 52767c478bd9Sstevel@tonic-gate ASSERT(ip->i_mapcnt >= 0); 52777c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 52787c478bd9Sstevel@tonic-gate return (0); 52797c478bd9Sstevel@tonic-gate } 52807c478bd9Sstevel@tonic-gate /* 52817c478bd9Sstevel@tonic-gate * Return the answer requested to poll() for non-device files 52827c478bd9Sstevel@tonic-gate */ 52837c478bd9Sstevel@tonic-gate struct pollhead ufs_pollhd; 52847c478bd9Sstevel@tonic-gate 52857c478bd9Sstevel@tonic-gate /* ARGSUSED */ 52867c478bd9Sstevel@tonic-gate int 52877c478bd9Sstevel@tonic-gate ufs_poll(vnode_t *vp, short ev, int any, short *revp, struct pollhead **phpp) 52887c478bd9Sstevel@tonic-gate { 52897c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp; 52907c478bd9Sstevel@tonic-gate 52917c478bd9Sstevel@tonic-gate *revp = 0; 52927c478bd9Sstevel@tonic-gate ufsvfsp = VTOI(vp)->i_ufsvfs; 52937c478bd9Sstevel@tonic-gate 52947c478bd9Sstevel@tonic-gate if (!ufsvfsp) { 52957c478bd9Sstevel@tonic-gate *revp = POLLHUP; 52967c478bd9Sstevel@tonic-gate goto out; 52977c478bd9Sstevel@tonic-gate } 52987c478bd9Sstevel@tonic-gate 52997c478bd9Sstevel@tonic-gate if (ULOCKFS_IS_HLOCK(&ufsvfsp->vfs_ulockfs) || 53007c478bd9Sstevel@tonic-gate ULOCKFS_IS_ELOCK(&ufsvfsp->vfs_ulockfs)) { 53017c478bd9Sstevel@tonic-gate *revp |= POLLERR; 53027c478bd9Sstevel@tonic-gate 53037c478bd9Sstevel@tonic-gate } else { 53047c478bd9Sstevel@tonic-gate if ((ev & POLLOUT) && !ufsvfsp->vfs_fs->fs_ronly && 53057c478bd9Sstevel@tonic-gate !ULOCKFS_IS_WLOCK(&ufsvfsp->vfs_ulockfs)) 53067c478bd9Sstevel@tonic-gate *revp |= POLLOUT; 53077c478bd9Sstevel@tonic-gate 53087c478bd9Sstevel@tonic-gate if ((ev & POLLWRBAND) && !ufsvfsp->vfs_fs->fs_ronly && 53097c478bd9Sstevel@tonic-gate !ULOCKFS_IS_WLOCK(&ufsvfsp->vfs_ulockfs)) 53107c478bd9Sstevel@tonic-gate *revp |= POLLWRBAND; 53117c478bd9Sstevel@tonic-gate 53127c478bd9Sstevel@tonic-gate if (ev & POLLIN) 53137c478bd9Sstevel@tonic-gate *revp |= POLLIN; 53147c478bd9Sstevel@tonic-gate 53157c478bd9Sstevel@tonic-gate if (ev & POLLRDNORM) 53167c478bd9Sstevel@tonic-gate *revp |= POLLRDNORM; 53177c478bd9Sstevel@tonic-gate 53187c478bd9Sstevel@tonic-gate if (ev & POLLRDBAND) 53197c478bd9Sstevel@tonic-gate *revp |= POLLRDBAND; 53207c478bd9Sstevel@tonic-gate } 53217c478bd9Sstevel@tonic-gate 53227c478bd9Sstevel@tonic-gate if ((ev & POLLPRI) && (*revp & (POLLERR|POLLHUP))) 53237c478bd9Sstevel@tonic-gate *revp |= POLLPRI; 53247c478bd9Sstevel@tonic-gate out: 53257c478bd9Sstevel@tonic-gate *phpp = !any && !*revp ? &ufs_pollhd : (struct pollhead *)NULL; 53267c478bd9Sstevel@tonic-gate 53277c478bd9Sstevel@tonic-gate return (0); 53287c478bd9Sstevel@tonic-gate } 53297c478bd9Sstevel@tonic-gate 53307c478bd9Sstevel@tonic-gate /* ARGSUSED */ 53317c478bd9Sstevel@tonic-gate static int 53327c478bd9Sstevel@tonic-gate ufs_l_pathconf(struct vnode *vp, int cmd, ulong_t *valp, struct cred *cr) 53337c478bd9Sstevel@tonic-gate { 53347c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp = VTOI(vp)->i_ufsvfs; 53357c478bd9Sstevel@tonic-gate struct ulockfs *ulp = NULL; 53367c478bd9Sstevel@tonic-gate struct inode *sip = NULL; 53377c478bd9Sstevel@tonic-gate int error; 53387c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 53397c478bd9Sstevel@tonic-gate int issync; 53407c478bd9Sstevel@tonic-gate 53417c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_PATHCONF_MASK); 53427c478bd9Sstevel@tonic-gate if (error) 53437c478bd9Sstevel@tonic-gate return (error); 53447c478bd9Sstevel@tonic-gate 53457c478bd9Sstevel@tonic-gate switch (cmd) { 53467c478bd9Sstevel@tonic-gate /* 53477c478bd9Sstevel@tonic-gate * Have to handle _PC_NAME_MAX here, because the normal way 53487c478bd9Sstevel@tonic-gate * [fs_pathconf() -> VOP_STATVFS() -> ufs_statvfs()] 53497c478bd9Sstevel@tonic-gate * results in a lock ordering reversal between 53507c478bd9Sstevel@tonic-gate * ufs_lockfs_{begin,end}() and 53517c478bd9Sstevel@tonic-gate * ufs_thread_{suspend,continue}(). 53527c478bd9Sstevel@tonic-gate * 53537c478bd9Sstevel@tonic-gate * Keep in sync with ufs_statvfs(). 53547c478bd9Sstevel@tonic-gate */ 53557c478bd9Sstevel@tonic-gate case _PC_NAME_MAX: 53567c478bd9Sstevel@tonic-gate *valp = MAXNAMLEN; 53577c478bd9Sstevel@tonic-gate break; 53587c478bd9Sstevel@tonic-gate 53597c478bd9Sstevel@tonic-gate case _PC_FILESIZEBITS: 53607c478bd9Sstevel@tonic-gate if (ufsvfsp->vfs_lfflags & UFS_LARGEFILES) 53617c478bd9Sstevel@tonic-gate *valp = UFS_FILESIZE_BITS; 53627c478bd9Sstevel@tonic-gate else 53637c478bd9Sstevel@tonic-gate *valp = 32; 53647c478bd9Sstevel@tonic-gate break; 53657c478bd9Sstevel@tonic-gate 53667c478bd9Sstevel@tonic-gate case _PC_XATTR_EXISTS: 53677c478bd9Sstevel@tonic-gate if (vp->v_vfsp->vfs_flag & VFS_XATTR) { 53687c478bd9Sstevel@tonic-gate 53697c478bd9Sstevel@tonic-gate error = ufs_xattr_getattrdir(vp, &sip, LOOKUP_XATTR, 53707c478bd9Sstevel@tonic-gate cr); 53717c478bd9Sstevel@tonic-gate if (error == 0 && sip != NULL) { 53727c478bd9Sstevel@tonic-gate /* Start transaction */ 53737c478bd9Sstevel@tonic-gate if (ulp) { 53747c478bd9Sstevel@tonic-gate TRANS_BEGIN_CSYNC(ufsvfsp, issync, 53757c478bd9Sstevel@tonic-gate TOP_RMDIR, TOP_RMDIR_SIZE); 53767c478bd9Sstevel@tonic-gate } 53777c478bd9Sstevel@tonic-gate /* 53787c478bd9Sstevel@tonic-gate * Is directory empty 53797c478bd9Sstevel@tonic-gate */ 53807c478bd9Sstevel@tonic-gate rw_enter(&sip->i_rwlock, RW_WRITER); 53817c478bd9Sstevel@tonic-gate rw_enter(&sip->i_contents, RW_WRITER); 53827c478bd9Sstevel@tonic-gate if (ufs_xattrdirempty(sip, 53837c478bd9Sstevel@tonic-gate sip->i_number, CRED())) { 53847c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_WRITER); 53857c478bd9Sstevel@tonic-gate ufs_unhook_shadow(ip, sip); 53867c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 53877c478bd9Sstevel@tonic-gate 53887c478bd9Sstevel@tonic-gate *valp = 0; 53897c478bd9Sstevel@tonic-gate 53907c478bd9Sstevel@tonic-gate } else 53917c478bd9Sstevel@tonic-gate *valp = 1; 53927c478bd9Sstevel@tonic-gate rw_exit(&sip->i_contents); 53937c478bd9Sstevel@tonic-gate rw_exit(&sip->i_rwlock); 53947c478bd9Sstevel@tonic-gate if (ulp) { 53957c478bd9Sstevel@tonic-gate TRANS_END_CSYNC(ufsvfsp, error, issync, 53967c478bd9Sstevel@tonic-gate TOP_RMDIR, TOP_RMDIR_SIZE); 53977c478bd9Sstevel@tonic-gate } 53987c478bd9Sstevel@tonic-gate VN_RELE(ITOV(sip)); 53997c478bd9Sstevel@tonic-gate } else if (error == ENOENT) { 54007c478bd9Sstevel@tonic-gate *valp = 0; 54017c478bd9Sstevel@tonic-gate error = 0; 54027c478bd9Sstevel@tonic-gate } 54037c478bd9Sstevel@tonic-gate } else { 54047c478bd9Sstevel@tonic-gate error = fs_pathconf(vp, cmd, valp, cr); 54057c478bd9Sstevel@tonic-gate } 54067c478bd9Sstevel@tonic-gate break; 54077c478bd9Sstevel@tonic-gate 54087c478bd9Sstevel@tonic-gate case _PC_ACL_ENABLED: 54097c478bd9Sstevel@tonic-gate *valp = _ACL_ACLENT_ENABLED; 54107c478bd9Sstevel@tonic-gate break; 54117c478bd9Sstevel@tonic-gate 54127c478bd9Sstevel@tonic-gate case _PC_MIN_HOLE_SIZE: 54137c478bd9Sstevel@tonic-gate *valp = (ulong_t)ip->i_fs->fs_bsize; 54147c478bd9Sstevel@tonic-gate break; 54157c478bd9Sstevel@tonic-gate 54167c478bd9Sstevel@tonic-gate default: 54177c478bd9Sstevel@tonic-gate error = fs_pathconf(vp, cmd, valp, cr); 54187c478bd9Sstevel@tonic-gate } 54197c478bd9Sstevel@tonic-gate 54207c478bd9Sstevel@tonic-gate if (ulp != NULL) { 54217c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 54227c478bd9Sstevel@tonic-gate } 54237c478bd9Sstevel@tonic-gate return (error); 54247c478bd9Sstevel@tonic-gate } 54257c478bd9Sstevel@tonic-gate 54267c478bd9Sstevel@tonic-gate int ufs_pageio_writes, ufs_pageio_reads; 54277c478bd9Sstevel@tonic-gate 54287c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 54297c478bd9Sstevel@tonic-gate static int 54307c478bd9Sstevel@tonic-gate ufs_pageio(struct vnode *vp, page_t *pp, u_offset_t io_off, size_t io_len, 54317c478bd9Sstevel@tonic-gate int flags, struct cred *cr) 54327c478bd9Sstevel@tonic-gate { 54337c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 54347c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp; 54357c478bd9Sstevel@tonic-gate page_t *npp = NULL, *opp = NULL, *cpp = pp; 54367c478bd9Sstevel@tonic-gate struct buf *bp; 54377c478bd9Sstevel@tonic-gate daddr_t bn; 54387c478bd9Sstevel@tonic-gate size_t done_len = 0, cur_len = 0; 54397c478bd9Sstevel@tonic-gate int err = 0; 54407c478bd9Sstevel@tonic-gate int contig = 0; 54417c478bd9Sstevel@tonic-gate int dolock; 54427c478bd9Sstevel@tonic-gate int vmpss = 0; 5443*bc69f433Saguzovsk struct ulockfs *ulp; 54447c478bd9Sstevel@tonic-gate 54457c478bd9Sstevel@tonic-gate if ((flags & B_READ) && pp != NULL && pp->p_vnode == vp && 54467c478bd9Sstevel@tonic-gate vp->v_mpssdata != NULL) { 54477c478bd9Sstevel@tonic-gate vmpss = 1; 54487c478bd9Sstevel@tonic-gate } 54497c478bd9Sstevel@tonic-gate 54507c478bd9Sstevel@tonic-gate dolock = (rw_owner(&ip->i_contents) != curthread); 54517c478bd9Sstevel@tonic-gate /* 54527c478bd9Sstevel@tonic-gate * We need a better check. Ideally, we would use another 54537c478bd9Sstevel@tonic-gate * vnodeops so that hlocked and forcibly unmounted file 54547c478bd9Sstevel@tonic-gate * systems would return EIO where appropriate and w/o the 54557c478bd9Sstevel@tonic-gate * need for these checks. 54567c478bd9Sstevel@tonic-gate */ 54577c478bd9Sstevel@tonic-gate if ((ufsvfsp = ip->i_ufsvfs) == NULL) 54587c478bd9Sstevel@tonic-gate return (EIO); 54597c478bd9Sstevel@tonic-gate 5460*bc69f433Saguzovsk /* 5461*bc69f433Saguzovsk * For vmpss (pp can be NULL) case respect the quiesce protocol. 5462*bc69f433Saguzovsk * ul_lock must be taken before locking pages so we can't use it here 5463*bc69f433Saguzovsk * if pp is non NULL because segvn already locked pages 5464*bc69f433Saguzovsk * SE_EXCL. Instead we rely on the fact that a forced umount or 5465*bc69f433Saguzovsk * applying a filesystem lock via ufs_fiolfs() will block in the 5466*bc69f433Saguzovsk * implicit call to ufs_flush() until we unlock the pages after the 5467*bc69f433Saguzovsk * return to segvn. Other ufs_quiesce() callers keep ufs_quiesce_pend 5468*bc69f433Saguzovsk * above 0 until they are done. We have to be careful not to increment 5469*bc69f433Saguzovsk * ul_vnops_cnt here after forceful unmount hlocks the file system. 5470*bc69f433Saguzovsk * 5471*bc69f433Saguzovsk * If pp is NULL use ul_lock to make sure we don't increment 5472*bc69f433Saguzovsk * ul_vnops_cnt after forceful unmount hlocks the file system. 5473*bc69f433Saguzovsk */ 5474*bc69f433Saguzovsk if (vmpss || pp == NULL) { 5475*bc69f433Saguzovsk ulp = &ufsvfsp->vfs_ulockfs; 5476*bc69f433Saguzovsk if (pp == NULL) 5477*bc69f433Saguzovsk mutex_enter(&ulp->ul_lock); 5478*bc69f433Saguzovsk if (ulp->ul_fs_lock & ULOCKFS_GETREAD_MASK) { 5479*bc69f433Saguzovsk if (pp == NULL) { 5480*bc69f433Saguzovsk mutex_exit(&ulp->ul_lock); 5481*bc69f433Saguzovsk } 5482*bc69f433Saguzovsk return (vmpss ? EIO : EINVAL); 5483*bc69f433Saguzovsk } 5484*bc69f433Saguzovsk atomic_add_long(&ulp->ul_vnops_cnt, 1); 5485*bc69f433Saguzovsk if (pp == NULL) 5486*bc69f433Saguzovsk mutex_exit(&ulp->ul_lock); 5487*bc69f433Saguzovsk if (ufs_quiesce_pend) { 5488*bc69f433Saguzovsk if (!atomic_add_long_nv(&ulp->ul_vnops_cnt, -1)) 5489*bc69f433Saguzovsk cv_broadcast(&ulp->ul_cv); 5490*bc69f433Saguzovsk return (vmpss ? EIO : EINVAL); 5491*bc69f433Saguzovsk } 5492*bc69f433Saguzovsk } 5493*bc69f433Saguzovsk 54947c478bd9Sstevel@tonic-gate if (dolock) { 54957c478bd9Sstevel@tonic-gate /* 54967c478bd9Sstevel@tonic-gate * segvn may call VOP_PAGEIO() instead of VOP_GETPAGE() to 54977c478bd9Sstevel@tonic-gate * handle a fault against a segment that maps vnode pages with 54987c478bd9Sstevel@tonic-gate * large mappings. Segvn creates pages and holds them locked 54997c478bd9Sstevel@tonic-gate * SE_EXCL during VOP_PAGEIO() call. In this case we have to 55007c478bd9Sstevel@tonic-gate * use rw_tryenter() to avoid a potential deadlock since in 55017c478bd9Sstevel@tonic-gate * lock order i_contents needs to be taken first. 55027c478bd9Sstevel@tonic-gate * Segvn will retry via VOP_GETPAGE() if VOP_PAGEIO() fails. 55037c478bd9Sstevel@tonic-gate */ 55047c478bd9Sstevel@tonic-gate if (!vmpss) { 55057c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_READER); 55067c478bd9Sstevel@tonic-gate } else if (!rw_tryenter(&ip->i_contents, RW_READER)) { 5507*bc69f433Saguzovsk if (!atomic_add_long_nv(&ulp->ul_vnops_cnt, -1)) 5508*bc69f433Saguzovsk cv_broadcast(&ulp->ul_cv); 55097c478bd9Sstevel@tonic-gate return (EDEADLK); 55107c478bd9Sstevel@tonic-gate } 55117c478bd9Sstevel@tonic-gate } 55127c478bd9Sstevel@tonic-gate 551337fbc076Saguzovsk /* 551437fbc076Saguzovsk * Return an error to segvn because the pagefault request is beyond 551537fbc076Saguzovsk * PAGESIZE rounded EOF. 551637fbc076Saguzovsk */ 551737fbc076Saguzovsk if (vmpss && btopr(io_off + io_len) > btopr(ip->i_size)) { 551837fbc076Saguzovsk if (dolock) 551937fbc076Saguzovsk rw_exit(&ip->i_contents); 5520*bc69f433Saguzovsk if (!atomic_add_long_nv(&ulp->ul_vnops_cnt, -1)) 5521*bc69f433Saguzovsk cv_broadcast(&ulp->ul_cv); 552237fbc076Saguzovsk return (EFAULT); 552337fbc076Saguzovsk } 552437fbc076Saguzovsk 55257c478bd9Sstevel@tonic-gate if (pp == NULL) { 55267c478bd9Sstevel@tonic-gate if (bmap_has_holes(ip)) { 55277c478bd9Sstevel@tonic-gate err = ENOSYS; 55287c478bd9Sstevel@tonic-gate } else { 55297c478bd9Sstevel@tonic-gate err = EINVAL; 55307c478bd9Sstevel@tonic-gate } 55317c478bd9Sstevel@tonic-gate if (dolock) 55327c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 5533*bc69f433Saguzovsk if (!atomic_add_long_nv(&ulp->ul_vnops_cnt, -1)) 5534*bc69f433Saguzovsk cv_broadcast(&ulp->ul_cv); 55357c478bd9Sstevel@tonic-gate return (err); 55367c478bd9Sstevel@tonic-gate } 55377c478bd9Sstevel@tonic-gate 55387c478bd9Sstevel@tonic-gate /* 55397c478bd9Sstevel@tonic-gate * Break the io request into chunks, one for each contiguous 55407c478bd9Sstevel@tonic-gate * stretch of disk blocks in the target file. 55417c478bd9Sstevel@tonic-gate */ 55427c478bd9Sstevel@tonic-gate while (done_len < io_len) { 55437c478bd9Sstevel@tonic-gate ASSERT(cpp); 55447c478bd9Sstevel@tonic-gate contig = 0; 55457c478bd9Sstevel@tonic-gate if (err = bmap_read(ip, (u_offset_t)(io_off + done_len), 55467c478bd9Sstevel@tonic-gate &bn, &contig)) 55477c478bd9Sstevel@tonic-gate break; 55487c478bd9Sstevel@tonic-gate 55497c478bd9Sstevel@tonic-gate if (bn == UFS_HOLE) { /* No holey swapfiles */ 55507c478bd9Sstevel@tonic-gate if (vmpss) { 55517c478bd9Sstevel@tonic-gate err = EFAULT; 55527c478bd9Sstevel@tonic-gate break; 55537c478bd9Sstevel@tonic-gate } 55547c478bd9Sstevel@tonic-gate err = ufs_fault(ITOV(ip), "ufs_pageio: bn == UFS_HOLE"); 55557c478bd9Sstevel@tonic-gate break; 55567c478bd9Sstevel@tonic-gate } 55577c478bd9Sstevel@tonic-gate 55587c478bd9Sstevel@tonic-gate cur_len = MIN(io_len - done_len, contig); 55597c478bd9Sstevel@tonic-gate /* 55607c478bd9Sstevel@tonic-gate * Zero out a page beyond EOF, when the last block of 55617c478bd9Sstevel@tonic-gate * a file is a UFS fragment so that ufs_pageio() can be used 55627c478bd9Sstevel@tonic-gate * instead of ufs_getpage() to handle faults against 55637c478bd9Sstevel@tonic-gate * segvn segments that use large pages. 55647c478bd9Sstevel@tonic-gate */ 55657c478bd9Sstevel@tonic-gate page_list_break(&cpp, &npp, btopr(cur_len)); 55667c478bd9Sstevel@tonic-gate if ((flags & B_READ) && (cur_len & PAGEOFFSET)) { 55677c478bd9Sstevel@tonic-gate size_t xlen = cur_len & PAGEOFFSET; 55687c478bd9Sstevel@tonic-gate pagezero(cpp->p_prev, xlen, PAGESIZE - xlen); 55697c478bd9Sstevel@tonic-gate } 55707c478bd9Sstevel@tonic-gate 55717c478bd9Sstevel@tonic-gate bp = pageio_setup(cpp, cur_len, ip->i_devvp, flags); 55727c478bd9Sstevel@tonic-gate ASSERT(bp != NULL); 55737c478bd9Sstevel@tonic-gate 55747c478bd9Sstevel@tonic-gate bp->b_edev = ip->i_dev; 55757c478bd9Sstevel@tonic-gate bp->b_dev = cmpdev(ip->i_dev); 55767c478bd9Sstevel@tonic-gate bp->b_blkno = bn; 55777c478bd9Sstevel@tonic-gate bp->b_un.b_addr = (caddr_t)0; 55787c478bd9Sstevel@tonic-gate bp->b_file = ip->i_vnode; 55797c478bd9Sstevel@tonic-gate 55807c478bd9Sstevel@tonic-gate ufsvfsp->vfs_iotstamp = lbolt; 55817c478bd9Sstevel@tonic-gate ub.ub_pageios.value.ul++; 55827c478bd9Sstevel@tonic-gate if (ufsvfsp->vfs_snapshot) 55837c478bd9Sstevel@tonic-gate fssnap_strategy(&(ufsvfsp->vfs_snapshot), bp); 55847c478bd9Sstevel@tonic-gate else 55857c478bd9Sstevel@tonic-gate (void) bdev_strategy(bp); 55867c478bd9Sstevel@tonic-gate 55877c478bd9Sstevel@tonic-gate if (flags & B_READ) 55887c478bd9Sstevel@tonic-gate ufs_pageio_reads++; 55897c478bd9Sstevel@tonic-gate else 55907c478bd9Sstevel@tonic-gate ufs_pageio_writes++; 55917c478bd9Sstevel@tonic-gate if (flags & B_READ) 55927c478bd9Sstevel@tonic-gate lwp_stat_update(LWP_STAT_INBLK, 1); 55937c478bd9Sstevel@tonic-gate else 55947c478bd9Sstevel@tonic-gate lwp_stat_update(LWP_STAT_OUBLK, 1); 55957c478bd9Sstevel@tonic-gate /* 55967c478bd9Sstevel@tonic-gate * If the request is not B_ASYNC, wait for i/o to complete 55977c478bd9Sstevel@tonic-gate * and re-assemble the page list to return to the caller. 55987c478bd9Sstevel@tonic-gate * If it is B_ASYNC we leave the page list in pieces and 55997c478bd9Sstevel@tonic-gate * cleanup() will dispose of them. 56007c478bd9Sstevel@tonic-gate */ 56017c478bd9Sstevel@tonic-gate if ((flags & B_ASYNC) == 0) { 56027c478bd9Sstevel@tonic-gate err = biowait(bp); 56037c478bd9Sstevel@tonic-gate pageio_done(bp); 56047c478bd9Sstevel@tonic-gate if (err) 56057c478bd9Sstevel@tonic-gate break; 56067c478bd9Sstevel@tonic-gate page_list_concat(&opp, &cpp); 56077c478bd9Sstevel@tonic-gate } 56087c478bd9Sstevel@tonic-gate cpp = npp; 56097c478bd9Sstevel@tonic-gate npp = NULL; 56107c478bd9Sstevel@tonic-gate if (flags & B_READ) 56117c478bd9Sstevel@tonic-gate cur_len = P2ROUNDUP_TYPED(cur_len, PAGESIZE, size_t); 56127c478bd9Sstevel@tonic-gate done_len += cur_len; 56137c478bd9Sstevel@tonic-gate } 56147c478bd9Sstevel@tonic-gate ASSERT(err || (cpp == NULL && npp == NULL && done_len == io_len)); 56157c478bd9Sstevel@tonic-gate if (err) { 56167c478bd9Sstevel@tonic-gate if (flags & B_ASYNC) { 56177c478bd9Sstevel@tonic-gate /* Cleanup unprocessed parts of list */ 56187c478bd9Sstevel@tonic-gate page_list_concat(&cpp, &npp); 56197c478bd9Sstevel@tonic-gate if (flags & B_READ) 56207c478bd9Sstevel@tonic-gate pvn_read_done(cpp, B_ERROR); 56217c478bd9Sstevel@tonic-gate else 56227c478bd9Sstevel@tonic-gate pvn_write_done(cpp, B_ERROR); 56237c478bd9Sstevel@tonic-gate } else { 56247c478bd9Sstevel@tonic-gate /* Re-assemble list and let caller clean up */ 56257c478bd9Sstevel@tonic-gate page_list_concat(&opp, &cpp); 56267c478bd9Sstevel@tonic-gate page_list_concat(&opp, &npp); 56277c478bd9Sstevel@tonic-gate } 56287c478bd9Sstevel@tonic-gate } 5629*bc69f433Saguzovsk 5630*bc69f433Saguzovsk if (vmpss && !(ip->i_flag & IACC) && !ULOCKFS_IS_NOIACC(ulp) && 5631*bc69f433Saguzovsk ufsvfsp->vfs_fs->fs_ronly == 0 && !ufsvfsp->vfs_noatime) { 5632*bc69f433Saguzovsk mutex_enter(&ip->i_tlock); 5633*bc69f433Saguzovsk ip->i_flag |= IACC; 5634*bc69f433Saguzovsk ITIMES_NOLOCK(ip); 5635*bc69f433Saguzovsk mutex_exit(&ip->i_tlock); 5636*bc69f433Saguzovsk } 5637*bc69f433Saguzovsk 56387c478bd9Sstevel@tonic-gate if (dolock) 56397c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 5640*bc69f433Saguzovsk if (vmpss && !atomic_add_long_nv(&ulp->ul_vnops_cnt, -1)) 5641*bc69f433Saguzovsk cv_broadcast(&ulp->ul_cv); 56427c478bd9Sstevel@tonic-gate return (err); 56437c478bd9Sstevel@tonic-gate } 56447c478bd9Sstevel@tonic-gate 56457c478bd9Sstevel@tonic-gate /* 56467c478bd9Sstevel@tonic-gate * Called when the kernel is in a frozen state to dump data 56477c478bd9Sstevel@tonic-gate * directly to the device. It uses a private dump data structure, 56487c478bd9Sstevel@tonic-gate * set up by dump_ctl, to locate the correct disk block to which to dump. 56497c478bd9Sstevel@tonic-gate */ 56507c478bd9Sstevel@tonic-gate static int 56517c478bd9Sstevel@tonic-gate ufs_dump(vnode_t *vp, caddr_t addr, int ldbn, int dblks) 56527c478bd9Sstevel@tonic-gate { 56537c478bd9Sstevel@tonic-gate u_offset_t file_size; 56547c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 56557c478bd9Sstevel@tonic-gate struct fs *fs = ip->i_fs; 56567c478bd9Sstevel@tonic-gate daddr_t dbn, lfsbn; 56577c478bd9Sstevel@tonic-gate int disk_blks = fs->fs_bsize >> DEV_BSHIFT; 56587c478bd9Sstevel@tonic-gate int error = 0; 56597c478bd9Sstevel@tonic-gate int ndbs, nfsbs; 56607c478bd9Sstevel@tonic-gate 56617c478bd9Sstevel@tonic-gate /* 56627c478bd9Sstevel@tonic-gate * forced unmount case 56637c478bd9Sstevel@tonic-gate */ 56647c478bd9Sstevel@tonic-gate if (ip->i_ufsvfs == NULL) 56657c478bd9Sstevel@tonic-gate return (EIO); 56667c478bd9Sstevel@tonic-gate /* 56677c478bd9Sstevel@tonic-gate * Validate the inode that it has not been modified since 56687c478bd9Sstevel@tonic-gate * the dump structure is allocated. 56697c478bd9Sstevel@tonic-gate */ 56707c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 56717c478bd9Sstevel@tonic-gate if ((dump_info == NULL) || 56727c478bd9Sstevel@tonic-gate (dump_info->ip != ip) || 56737c478bd9Sstevel@tonic-gate (dump_info->time.tv_sec != ip->i_mtime.tv_sec) || 56747c478bd9Sstevel@tonic-gate (dump_info->time.tv_usec != ip->i_mtime.tv_usec)) { 56757c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 56767c478bd9Sstevel@tonic-gate return (-1); 56777c478bd9Sstevel@tonic-gate } 56787c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 56797c478bd9Sstevel@tonic-gate 56807c478bd9Sstevel@tonic-gate /* 56817c478bd9Sstevel@tonic-gate * See that the file has room for this write 56827c478bd9Sstevel@tonic-gate */ 56837c478bd9Sstevel@tonic-gate UFS_GET_ISIZE(&file_size, ip); 56847c478bd9Sstevel@tonic-gate 56857c478bd9Sstevel@tonic-gate if (ldbtob((offset_t)(ldbn + dblks)) > file_size) 56867c478bd9Sstevel@tonic-gate return (ENOSPC); 56877c478bd9Sstevel@tonic-gate 56887c478bd9Sstevel@tonic-gate /* 56897c478bd9Sstevel@tonic-gate * Find the physical disk block numbers from the dump 56907c478bd9Sstevel@tonic-gate * private data structure directly and write out the data 56917c478bd9Sstevel@tonic-gate * in contiguous block lumps 56927c478bd9Sstevel@tonic-gate */ 56937c478bd9Sstevel@tonic-gate while (dblks > 0 && !error) { 56947c478bd9Sstevel@tonic-gate lfsbn = (daddr_t)lblkno(fs, ldbtob((offset_t)ldbn)); 56957c478bd9Sstevel@tonic-gate dbn = fsbtodb(fs, dump_info->dblk[lfsbn]) + ldbn % disk_blks; 56967c478bd9Sstevel@tonic-gate nfsbs = 1; 56977c478bd9Sstevel@tonic-gate ndbs = disk_blks - ldbn % disk_blks; 56987c478bd9Sstevel@tonic-gate while (ndbs < dblks && fsbtodb(fs, dump_info->dblk[lfsbn + 56997c478bd9Sstevel@tonic-gate nfsbs]) == dbn + ndbs) { 57007c478bd9Sstevel@tonic-gate nfsbs++; 57017c478bd9Sstevel@tonic-gate ndbs += disk_blks; 57027c478bd9Sstevel@tonic-gate } 57037c478bd9Sstevel@tonic-gate if (ndbs > dblks) 57047c478bd9Sstevel@tonic-gate ndbs = dblks; 57057c478bd9Sstevel@tonic-gate error = bdev_dump(ip->i_dev, addr, dbn, ndbs); 57067c478bd9Sstevel@tonic-gate addr += ldbtob((offset_t)ndbs); 57077c478bd9Sstevel@tonic-gate dblks -= ndbs; 57087c478bd9Sstevel@tonic-gate ldbn += ndbs; 57097c478bd9Sstevel@tonic-gate } 57107c478bd9Sstevel@tonic-gate return (error); 57117c478bd9Sstevel@tonic-gate 57127c478bd9Sstevel@tonic-gate } 57137c478bd9Sstevel@tonic-gate 57147c478bd9Sstevel@tonic-gate /* 57157c478bd9Sstevel@tonic-gate * Prepare the file system before and after the dump operation. 57167c478bd9Sstevel@tonic-gate * 57177c478bd9Sstevel@tonic-gate * action = DUMP_ALLOC: 57187c478bd9Sstevel@tonic-gate * Preparation before dump, allocate dump private data structure 57197c478bd9Sstevel@tonic-gate * to hold all the direct and indirect block info for dump. 57207c478bd9Sstevel@tonic-gate * 57217c478bd9Sstevel@tonic-gate * action = DUMP_FREE: 57227c478bd9Sstevel@tonic-gate * Clean up after dump, deallocate the dump private data structure. 57237c478bd9Sstevel@tonic-gate * 57247c478bd9Sstevel@tonic-gate * action = DUMP_SCAN: 57257c478bd9Sstevel@tonic-gate * Scan dump_info for *blkp DEV_BSIZE blocks of contig fs space; 57267c478bd9Sstevel@tonic-gate * if found, the starting file-relative DEV_BSIZE lbn is written 57277c478bd9Sstevel@tonic-gate * to *bklp; that lbn is intended for use with VOP_DUMP() 57287c478bd9Sstevel@tonic-gate */ 57297c478bd9Sstevel@tonic-gate static int 57307c478bd9Sstevel@tonic-gate ufs_dumpctl(vnode_t *vp, int action, int *blkp) 57317c478bd9Sstevel@tonic-gate { 57327c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 57337c478bd9Sstevel@tonic-gate ufsvfs_t *ufsvfsp = ip->i_ufsvfs; 57347c478bd9Sstevel@tonic-gate struct fs *fs; 57357c478bd9Sstevel@tonic-gate daddr32_t *dblk, *storeblk; 57367c478bd9Sstevel@tonic-gate daddr32_t *nextblk, *endblk; 57377c478bd9Sstevel@tonic-gate struct buf *bp; 57387c478bd9Sstevel@tonic-gate int i, entry, entries; 57397c478bd9Sstevel@tonic-gate int n, ncontig; 57407c478bd9Sstevel@tonic-gate 57417c478bd9Sstevel@tonic-gate /* 57427c478bd9Sstevel@tonic-gate * check for forced unmount 57437c478bd9Sstevel@tonic-gate */ 57447c478bd9Sstevel@tonic-gate if (ufsvfsp == NULL) 57457c478bd9Sstevel@tonic-gate return (EIO); 57467c478bd9Sstevel@tonic-gate 57477c478bd9Sstevel@tonic-gate if (action == DUMP_ALLOC) { 57487c478bd9Sstevel@tonic-gate /* 57497c478bd9Sstevel@tonic-gate * alloc and record dump_info 57507c478bd9Sstevel@tonic-gate */ 57517c478bd9Sstevel@tonic-gate if (dump_info != NULL) 57527c478bd9Sstevel@tonic-gate return (EINVAL); 57537c478bd9Sstevel@tonic-gate 57547c478bd9Sstevel@tonic-gate ASSERT(vp->v_type == VREG); 57557c478bd9Sstevel@tonic-gate fs = ufsvfsp->vfs_fs; 57567c478bd9Sstevel@tonic-gate 57577c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_READER); 57587c478bd9Sstevel@tonic-gate 57597c478bd9Sstevel@tonic-gate if (bmap_has_holes(ip)) { 57607c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 57617c478bd9Sstevel@tonic-gate return (EFAULT); 57627c478bd9Sstevel@tonic-gate } 57637c478bd9Sstevel@tonic-gate 57647c478bd9Sstevel@tonic-gate /* 57657c478bd9Sstevel@tonic-gate * calculate and allocate space needed according to i_size 57667c478bd9Sstevel@tonic-gate */ 57677c478bd9Sstevel@tonic-gate entries = (int)lblkno(fs, blkroundup(fs, ip->i_size)); 57687c478bd9Sstevel@tonic-gate if ((dump_info = (struct dump *) 57697c478bd9Sstevel@tonic-gate kmem_alloc(sizeof (struct dump) + 57707c478bd9Sstevel@tonic-gate (entries - 1) * sizeof (daddr32_t), KM_NOSLEEP)) == NULL) { 57717c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 57727c478bd9Sstevel@tonic-gate return (ENOMEM); 57737c478bd9Sstevel@tonic-gate } 57747c478bd9Sstevel@tonic-gate 57757c478bd9Sstevel@tonic-gate /* Start saving the info */ 57767c478bd9Sstevel@tonic-gate dump_info->fsbs = entries; 57777c478bd9Sstevel@tonic-gate dump_info->ip = ip; 57787c478bd9Sstevel@tonic-gate storeblk = &dump_info->dblk[0]; 57797c478bd9Sstevel@tonic-gate 57807c478bd9Sstevel@tonic-gate /* Direct Blocks */ 57817c478bd9Sstevel@tonic-gate for (entry = 0; entry < NDADDR && entry < entries; entry++) 57827c478bd9Sstevel@tonic-gate *storeblk++ = ip->i_db[entry]; 57837c478bd9Sstevel@tonic-gate 57847c478bd9Sstevel@tonic-gate /* Indirect Blocks */ 57857c478bd9Sstevel@tonic-gate for (i = 0; i < NIADDR; i++) { 57867c478bd9Sstevel@tonic-gate int error = 0; 57877c478bd9Sstevel@tonic-gate 57887c478bd9Sstevel@tonic-gate bp = UFS_BREAD(ufsvfsp, 57897c478bd9Sstevel@tonic-gate ip->i_dev, fsbtodb(fs, ip->i_ib[i]), 57907c478bd9Sstevel@tonic-gate fs->fs_bsize); 57917c478bd9Sstevel@tonic-gate if (bp->b_flags & B_ERROR) 57927c478bd9Sstevel@tonic-gate error = EIO; 57937c478bd9Sstevel@tonic-gate else { 57947c478bd9Sstevel@tonic-gate dblk = bp->b_un.b_daddr; 57957c478bd9Sstevel@tonic-gate if ((storeblk = save_dblks(ip, ufsvfsp, 57967c478bd9Sstevel@tonic-gate storeblk, dblk, i, entries)) == NULL) 57977c478bd9Sstevel@tonic-gate error = EIO; 57987c478bd9Sstevel@tonic-gate } 57997c478bd9Sstevel@tonic-gate 58007c478bd9Sstevel@tonic-gate brelse(bp); 58017c478bd9Sstevel@tonic-gate 58027c478bd9Sstevel@tonic-gate if (error != 0) { 58037c478bd9Sstevel@tonic-gate kmem_free(dump_info, sizeof (struct dump) + 58047c478bd9Sstevel@tonic-gate (entries - 1) * sizeof (daddr32_t)); 58057c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 58067c478bd9Sstevel@tonic-gate dump_info = NULL; 58077c478bd9Sstevel@tonic-gate return (error); 58087c478bd9Sstevel@tonic-gate } 58097c478bd9Sstevel@tonic-gate } 58107c478bd9Sstevel@tonic-gate /* and time stamp the information */ 58117c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 58127c478bd9Sstevel@tonic-gate dump_info->time = ip->i_mtime; 58137c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 58147c478bd9Sstevel@tonic-gate 58157c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 58167c478bd9Sstevel@tonic-gate } else if (action == DUMP_FREE) { 58177c478bd9Sstevel@tonic-gate /* 58187c478bd9Sstevel@tonic-gate * free dump_info 58197c478bd9Sstevel@tonic-gate */ 58207c478bd9Sstevel@tonic-gate if (dump_info == NULL) 58217c478bd9Sstevel@tonic-gate return (EINVAL); 58227c478bd9Sstevel@tonic-gate entries = dump_info->fsbs - 1; 58237c478bd9Sstevel@tonic-gate kmem_free(dump_info, sizeof (struct dump) + 58247c478bd9Sstevel@tonic-gate entries * sizeof (daddr32_t)); 58257c478bd9Sstevel@tonic-gate dump_info = NULL; 58267c478bd9Sstevel@tonic-gate } else if (action == DUMP_SCAN) { 58277c478bd9Sstevel@tonic-gate /* 58287c478bd9Sstevel@tonic-gate * scan dump_info 58297c478bd9Sstevel@tonic-gate */ 58307c478bd9Sstevel@tonic-gate if (dump_info == NULL) 58317c478bd9Sstevel@tonic-gate return (EINVAL); 58327c478bd9Sstevel@tonic-gate 58337c478bd9Sstevel@tonic-gate dblk = dump_info->dblk; 58347c478bd9Sstevel@tonic-gate nextblk = dblk + 1; 58357c478bd9Sstevel@tonic-gate endblk = dblk + dump_info->fsbs - 1; 58367c478bd9Sstevel@tonic-gate fs = ufsvfsp->vfs_fs; 58377c478bd9Sstevel@tonic-gate ncontig = *blkp >> (fs->fs_bshift - DEV_BSHIFT); 58387c478bd9Sstevel@tonic-gate 58397c478bd9Sstevel@tonic-gate /* 58407c478bd9Sstevel@tonic-gate * scan dblk[] entries; contig fs space is found when: 58417c478bd9Sstevel@tonic-gate * ((current blkno + frags per block) == next blkno) 58427c478bd9Sstevel@tonic-gate */ 58437c478bd9Sstevel@tonic-gate n = 0; 58447c478bd9Sstevel@tonic-gate while (n < ncontig && dblk < endblk) { 58457c478bd9Sstevel@tonic-gate if ((*dblk + fs->fs_frag) == *nextblk) 58467c478bd9Sstevel@tonic-gate n++; 58477c478bd9Sstevel@tonic-gate else 58487c478bd9Sstevel@tonic-gate n = 0; 58497c478bd9Sstevel@tonic-gate dblk++; 58507c478bd9Sstevel@tonic-gate nextblk++; 58517c478bd9Sstevel@tonic-gate } 58527c478bd9Sstevel@tonic-gate 58537c478bd9Sstevel@tonic-gate /* 58547c478bd9Sstevel@tonic-gate * index is where size bytes of contig space begins; 58557c478bd9Sstevel@tonic-gate * conversion from index to the file's DEV_BSIZE lbn 58567c478bd9Sstevel@tonic-gate * is equivalent to: (index * fs_bsize) / DEV_BSIZE 58577c478bd9Sstevel@tonic-gate */ 58587c478bd9Sstevel@tonic-gate if (n == ncontig) { 58597c478bd9Sstevel@tonic-gate i = (dblk - dump_info->dblk) - ncontig; 58607c478bd9Sstevel@tonic-gate *blkp = i << (fs->fs_bshift - DEV_BSHIFT); 58617c478bd9Sstevel@tonic-gate } else 58627c478bd9Sstevel@tonic-gate return (EFAULT); 58637c478bd9Sstevel@tonic-gate } 58647c478bd9Sstevel@tonic-gate return (0); 58657c478bd9Sstevel@tonic-gate } 58667c478bd9Sstevel@tonic-gate 58677c478bd9Sstevel@tonic-gate /* 58687c478bd9Sstevel@tonic-gate * Recursive helper function for ufs_dumpctl(). It follows the indirect file 58697c478bd9Sstevel@tonic-gate * system blocks until it reaches the the disk block addresses, which are 58707c478bd9Sstevel@tonic-gate * then stored into the given buffer, storeblk. 58717c478bd9Sstevel@tonic-gate */ 58727c478bd9Sstevel@tonic-gate static daddr32_t * 58737c478bd9Sstevel@tonic-gate save_dblks(struct inode *ip, struct ufsvfs *ufsvfsp, daddr32_t *storeblk, 58747c478bd9Sstevel@tonic-gate daddr32_t *dblk, int level, int entries) 58757c478bd9Sstevel@tonic-gate { 58767c478bd9Sstevel@tonic-gate struct fs *fs = ufsvfsp->vfs_fs; 58777c478bd9Sstevel@tonic-gate struct buf *bp; 58787c478bd9Sstevel@tonic-gate int i; 58797c478bd9Sstevel@tonic-gate 58807c478bd9Sstevel@tonic-gate if (level == 0) { 58817c478bd9Sstevel@tonic-gate for (i = 0; i < NINDIR(fs); i++) { 58827c478bd9Sstevel@tonic-gate if (storeblk - dump_info->dblk >= entries) 58837c478bd9Sstevel@tonic-gate break; 58847c478bd9Sstevel@tonic-gate *storeblk++ = dblk[i]; 58857c478bd9Sstevel@tonic-gate } 58867c478bd9Sstevel@tonic-gate return (storeblk); 58877c478bd9Sstevel@tonic-gate } 58887c478bd9Sstevel@tonic-gate for (i = 0; i < NINDIR(fs); i++) { 58897c478bd9Sstevel@tonic-gate if (storeblk - dump_info->dblk >= entries) 58907c478bd9Sstevel@tonic-gate break; 58917c478bd9Sstevel@tonic-gate bp = UFS_BREAD(ufsvfsp, 58927c478bd9Sstevel@tonic-gate ip->i_dev, fsbtodb(fs, dblk[i]), fs->fs_bsize); 58937c478bd9Sstevel@tonic-gate if (bp->b_flags & B_ERROR) { 58947c478bd9Sstevel@tonic-gate brelse(bp); 58957c478bd9Sstevel@tonic-gate return (NULL); 58967c478bd9Sstevel@tonic-gate } 58977c478bd9Sstevel@tonic-gate storeblk = save_dblks(ip, ufsvfsp, storeblk, bp->b_un.b_daddr, 58987c478bd9Sstevel@tonic-gate level - 1, entries); 58997c478bd9Sstevel@tonic-gate brelse(bp); 59007c478bd9Sstevel@tonic-gate 59017c478bd9Sstevel@tonic-gate if (storeblk == NULL) 59027c478bd9Sstevel@tonic-gate return (NULL); 59037c478bd9Sstevel@tonic-gate } 59047c478bd9Sstevel@tonic-gate return (storeblk); 59057c478bd9Sstevel@tonic-gate } 59067c478bd9Sstevel@tonic-gate 59077c478bd9Sstevel@tonic-gate /* ARGSUSED */ 59087c478bd9Sstevel@tonic-gate static int 59097c478bd9Sstevel@tonic-gate ufs_getsecattr(struct vnode *vp, vsecattr_t *vsap, int flag, 59107c478bd9Sstevel@tonic-gate struct cred *cr) 59117c478bd9Sstevel@tonic-gate { 59127c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 59137c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 59147c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp = ip->i_ufsvfs; 59157c478bd9Sstevel@tonic-gate ulong_t vsa_mask = vsap->vsa_mask; 59167c478bd9Sstevel@tonic-gate int err = EINVAL; 59177c478bd9Sstevel@tonic-gate 59187c478bd9Sstevel@tonic-gate TRACE_3(TR_FAC_UFS, TR_UFS_GETSECATTR_START, 59197c478bd9Sstevel@tonic-gate "ufs_getsecattr_start:vp %p, vsap %p, flags %x", vp, vsap, flag); 59207c478bd9Sstevel@tonic-gate 59217c478bd9Sstevel@tonic-gate vsa_mask &= (VSA_ACL | VSA_ACLCNT | VSA_DFACL | VSA_DFACLCNT); 59227c478bd9Sstevel@tonic-gate 59237c478bd9Sstevel@tonic-gate /* 59247c478bd9Sstevel@tonic-gate * Only grab locks if needed - they're not needed to check vsa_mask 59257c478bd9Sstevel@tonic-gate * or if the mask contains no acl flags. 59267c478bd9Sstevel@tonic-gate */ 59277c478bd9Sstevel@tonic-gate if (vsa_mask != 0) { 59287c478bd9Sstevel@tonic-gate if (err = ufs_lockfs_begin(ufsvfsp, &ulp, 59297c478bd9Sstevel@tonic-gate ULOCKFS_GETATTR_MASK)) 59307c478bd9Sstevel@tonic-gate return (err); 59317c478bd9Sstevel@tonic-gate 59327c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_READER); 59337c478bd9Sstevel@tonic-gate err = ufs_acl_get(ip, vsap, flag, cr); 59347c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 59357c478bd9Sstevel@tonic-gate 59367c478bd9Sstevel@tonic-gate if (ulp) 59377c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 59387c478bd9Sstevel@tonic-gate } 59397c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_GETSECATTR_END, 59407c478bd9Sstevel@tonic-gate "ufs_getsecattr_end:vp %p", vp); 59417c478bd9Sstevel@tonic-gate return (err); 59427c478bd9Sstevel@tonic-gate } 59437c478bd9Sstevel@tonic-gate 59447c478bd9Sstevel@tonic-gate /* ARGSUSED */ 59457c478bd9Sstevel@tonic-gate static int 59467c478bd9Sstevel@tonic-gate ufs_setsecattr(struct vnode *vp, vsecattr_t *vsap, int flag, struct cred *cr) 59477c478bd9Sstevel@tonic-gate { 59487c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 59497c478bd9Sstevel@tonic-gate struct ulockfs *ulp = NULL; 59507c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp = VTOI(vp)->i_ufsvfs; 59517c478bd9Sstevel@tonic-gate ulong_t vsa_mask = vsap->vsa_mask; 59527c478bd9Sstevel@tonic-gate int err; 59537c478bd9Sstevel@tonic-gate int haverwlock = 1; 59547c478bd9Sstevel@tonic-gate int trans_size; 59557c478bd9Sstevel@tonic-gate int donetrans = 0; 59567c478bd9Sstevel@tonic-gate int retry = 1; 59577c478bd9Sstevel@tonic-gate 59587c478bd9Sstevel@tonic-gate 59597c478bd9Sstevel@tonic-gate TRACE_3(TR_FAC_UFS, TR_UFS_SETSECATTR_START, 59607c478bd9Sstevel@tonic-gate "ufs_setsecattr_start:vp %p, vsap %p, flags %x", vp, vsap, flag); 59617c478bd9Sstevel@tonic-gate 59627c478bd9Sstevel@tonic-gate ASSERT(RW_LOCK_HELD(&ip->i_rwlock)); 59637c478bd9Sstevel@tonic-gate 59647c478bd9Sstevel@tonic-gate /* Abort now if the request is either empty or invalid. */ 59657c478bd9Sstevel@tonic-gate vsa_mask &= (VSA_ACL | VSA_ACLCNT | VSA_DFACL | VSA_DFACLCNT); 59667c478bd9Sstevel@tonic-gate if ((vsa_mask == 0) || 59677c478bd9Sstevel@tonic-gate ((vsap->vsa_aclentp == NULL) && 59687c478bd9Sstevel@tonic-gate (vsap->vsa_dfaclentp == NULL))) { 59697c478bd9Sstevel@tonic-gate err = EINVAL; 59707c478bd9Sstevel@tonic-gate goto out; 59717c478bd9Sstevel@tonic-gate } 59727c478bd9Sstevel@tonic-gate 59737c478bd9Sstevel@tonic-gate /* 59747c478bd9Sstevel@tonic-gate * Following convention, if this is a directory then we acquire the 59757c478bd9Sstevel@tonic-gate * inode's i_rwlock after starting a UFS logging transaction; 59767c478bd9Sstevel@tonic-gate * otherwise, we acquire it beforehand. Since we were called (and 59777c478bd9Sstevel@tonic-gate * must therefore return) with the lock held, we will have to drop it, 59787c478bd9Sstevel@tonic-gate * and later reacquire it, if operating on a directory. 59797c478bd9Sstevel@tonic-gate */ 59807c478bd9Sstevel@tonic-gate if (vp->v_type == VDIR) { 59817c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 59827c478bd9Sstevel@tonic-gate haverwlock = 0; 59837c478bd9Sstevel@tonic-gate } else { 59847c478bd9Sstevel@tonic-gate /* Upgrade the lock if required. */ 59857c478bd9Sstevel@tonic-gate if (!rw_write_held(&ip->i_rwlock)) { 59867c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 59877c478bd9Sstevel@tonic-gate rw_enter(&ip->i_rwlock, RW_WRITER); 59887c478bd9Sstevel@tonic-gate } 59897c478bd9Sstevel@tonic-gate } 59907c478bd9Sstevel@tonic-gate 59917c478bd9Sstevel@tonic-gate again: 59927c478bd9Sstevel@tonic-gate ASSERT(!(vp->v_type == VDIR && haverwlock)); 59937c478bd9Sstevel@tonic-gate if (err = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_SETATTR_MASK)) { 59947c478bd9Sstevel@tonic-gate ulp = NULL; 59957c478bd9Sstevel@tonic-gate retry = 0; 59967c478bd9Sstevel@tonic-gate goto out; 59977c478bd9Sstevel@tonic-gate } 59987c478bd9Sstevel@tonic-gate 59997c478bd9Sstevel@tonic-gate /* 60007c478bd9Sstevel@tonic-gate * Check that the file system supports this operation. Note that 60017c478bd9Sstevel@tonic-gate * ufs_lockfs_begin() will have checked that the file system had 60027c478bd9Sstevel@tonic-gate * not been forcibly unmounted. 60037c478bd9Sstevel@tonic-gate */ 60047c478bd9Sstevel@tonic-gate if (ufsvfsp->vfs_fs->fs_ronly) { 60057c478bd9Sstevel@tonic-gate err = EROFS; 60067c478bd9Sstevel@tonic-gate goto out; 60077c478bd9Sstevel@tonic-gate } 60087c478bd9Sstevel@tonic-gate if (ufsvfsp->vfs_nosetsec) { 60097c478bd9Sstevel@tonic-gate err = ENOSYS; 60107c478bd9Sstevel@tonic-gate goto out; 60117c478bd9Sstevel@tonic-gate } 60127c478bd9Sstevel@tonic-gate 60137c478bd9Sstevel@tonic-gate if (ulp) { 60147c478bd9Sstevel@tonic-gate TRANS_BEGIN_ASYNC(ufsvfsp, TOP_SETSECATTR, 60157c478bd9Sstevel@tonic-gate trans_size = TOP_SETSECATTR_SIZE(VTOI(vp))); 60167c478bd9Sstevel@tonic-gate donetrans = 1; 60177c478bd9Sstevel@tonic-gate } 60187c478bd9Sstevel@tonic-gate 60197c478bd9Sstevel@tonic-gate if (vp->v_type == VDIR) { 60207c478bd9Sstevel@tonic-gate rw_enter(&ip->i_rwlock, RW_WRITER); 60217c478bd9Sstevel@tonic-gate haverwlock = 1; 60227c478bd9Sstevel@tonic-gate } 60237c478bd9Sstevel@tonic-gate 60247c478bd9Sstevel@tonic-gate ASSERT(haverwlock); 60257c478bd9Sstevel@tonic-gate 60267c478bd9Sstevel@tonic-gate /* Do the actual work. */ 60277c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_WRITER); 60287c478bd9Sstevel@tonic-gate /* 60297c478bd9Sstevel@tonic-gate * Suppress out of inodes messages if we will retry. 60307c478bd9Sstevel@tonic-gate */ 60317c478bd9Sstevel@tonic-gate if (retry) 60327c478bd9Sstevel@tonic-gate ip->i_flag |= IQUIET; 60337c478bd9Sstevel@tonic-gate err = ufs_acl_set(ip, vsap, flag, cr); 60347c478bd9Sstevel@tonic-gate ip->i_flag &= ~IQUIET; 60357c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 60367c478bd9Sstevel@tonic-gate 60377c478bd9Sstevel@tonic-gate out: 60387c478bd9Sstevel@tonic-gate if (ulp) { 60397c478bd9Sstevel@tonic-gate if (donetrans) { 60407c478bd9Sstevel@tonic-gate /* 60417c478bd9Sstevel@tonic-gate * top_end_async() can eventually call 60427c478bd9Sstevel@tonic-gate * top_end_sync(), which can block. We must 60437c478bd9Sstevel@tonic-gate * therefore observe the lock-ordering protocol 60447c478bd9Sstevel@tonic-gate * here as well. 60457c478bd9Sstevel@tonic-gate */ 60467c478bd9Sstevel@tonic-gate if (vp->v_type == VDIR) { 60477c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 60487c478bd9Sstevel@tonic-gate haverwlock = 0; 60497c478bd9Sstevel@tonic-gate } 60507c478bd9Sstevel@tonic-gate TRANS_END_ASYNC(ufsvfsp, TOP_SETSECATTR, trans_size); 60517c478bd9Sstevel@tonic-gate } 60527c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 60537c478bd9Sstevel@tonic-gate } 60547c478bd9Sstevel@tonic-gate /* 60557c478bd9Sstevel@tonic-gate * If no inodes available, try scaring a logically- 60567c478bd9Sstevel@tonic-gate * free one out of the delete queue to someplace 60577c478bd9Sstevel@tonic-gate * that we can find it. 60587c478bd9Sstevel@tonic-gate */ 60597c478bd9Sstevel@tonic-gate if ((err == ENOSPC) && retry && TRANS_ISTRANS(ufsvfsp)) { 60607c478bd9Sstevel@tonic-gate ufs_delete_drain_wait(ufsvfsp, 1); 60617c478bd9Sstevel@tonic-gate retry = 0; 60627c478bd9Sstevel@tonic-gate if (vp->v_type == VDIR && haverwlock) { 60637c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 60647c478bd9Sstevel@tonic-gate haverwlock = 0; 60657c478bd9Sstevel@tonic-gate } 60667c478bd9Sstevel@tonic-gate goto again; 60677c478bd9Sstevel@tonic-gate } 60687c478bd9Sstevel@tonic-gate /* 60697c478bd9Sstevel@tonic-gate * If we need to reacquire the lock then it is safe to do so 60707c478bd9Sstevel@tonic-gate * as a reader. This is because ufs_rwunlock(), which will be 60717c478bd9Sstevel@tonic-gate * called by our caller after we return, does not differentiate 60727c478bd9Sstevel@tonic-gate * between shared and exclusive locks. 60737c478bd9Sstevel@tonic-gate */ 60747c478bd9Sstevel@tonic-gate if (!haverwlock) { 60757c478bd9Sstevel@tonic-gate ASSERT(vp->v_type == VDIR); 60767c478bd9Sstevel@tonic-gate rw_enter(&ip->i_rwlock, RW_READER); 60777c478bd9Sstevel@tonic-gate } 60787c478bd9Sstevel@tonic-gate 60797c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_SETSECATTR_END, 60807c478bd9Sstevel@tonic-gate "ufs_setsecattr_end:vp %p", vp); 60817c478bd9Sstevel@tonic-gate return (err); 60827c478bd9Sstevel@tonic-gate } 6083