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 55b024a5bSbatschul * Common Development and Distribution License (the "License"). 65b024a5bSbatschul * You may not use this file except in compliance with the License. 77c478bd9Sstevel@tonic-gate * 87c478bd9Sstevel@tonic-gate * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 97c478bd9Sstevel@tonic-gate * or http://www.opensolaris.org/os/licensing. 107c478bd9Sstevel@tonic-gate * See the License for the specific language governing permissions 117c478bd9Sstevel@tonic-gate * and limitations under the License. 127c478bd9Sstevel@tonic-gate * 137c478bd9Sstevel@tonic-gate * When distributing Covered Code, include this CDDL HEADER in each 147c478bd9Sstevel@tonic-gate * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 157c478bd9Sstevel@tonic-gate * If applicable, add the following below this CDDL HEADER, with the 167c478bd9Sstevel@tonic-gate * fields enclosed by brackets "[]" replaced with your own identifying 177c478bd9Sstevel@tonic-gate * information: Portions Copyright [yyyy] [name of copyright owner] 187c478bd9Sstevel@tonic-gate * 197c478bd9Sstevel@tonic-gate * CDDL HEADER END 207c478bd9Sstevel@tonic-gate */ 217c478bd9Sstevel@tonic-gate /* 225b024a5bSbatschul * Copyright 2006 Sun Microsystems, Inc. All rights reserved. 237c478bd9Sstevel@tonic-gate * Use is subject to license terms. 247c478bd9Sstevel@tonic-gate */ 257c478bd9Sstevel@tonic-gate 267c478bd9Sstevel@tonic-gate /* Copyright (c) 1983, 1984, 1985, 1986, 1987, 1988, 1989 AT&T */ 277c478bd9Sstevel@tonic-gate /* All Rights Reserved */ 287c478bd9Sstevel@tonic-gate 297c478bd9Sstevel@tonic-gate /* 307c478bd9Sstevel@tonic-gate * Portions of this source code were derived from Berkeley 4.3 BSD 317c478bd9Sstevel@tonic-gate * under license from the Regents of the University of California. 327c478bd9Sstevel@tonic-gate */ 337c478bd9Sstevel@tonic-gate 347c478bd9Sstevel@tonic-gate #pragma ident "%Z%%M% %I% %E% SMI" 357c478bd9Sstevel@tonic-gate 367c478bd9Sstevel@tonic-gate #include <sys/types.h> 377c478bd9Sstevel@tonic-gate #include <sys/t_lock.h> 387c478bd9Sstevel@tonic-gate #include <sys/ksynch.h> 397c478bd9Sstevel@tonic-gate #include <sys/param.h> 407c478bd9Sstevel@tonic-gate #include <sys/time.h> 417c478bd9Sstevel@tonic-gate #include <sys/systm.h> 427c478bd9Sstevel@tonic-gate #include <sys/sysmacros.h> 437c478bd9Sstevel@tonic-gate #include <sys/resource.h> 447c478bd9Sstevel@tonic-gate #include <sys/signal.h> 457c478bd9Sstevel@tonic-gate #include <sys/cred.h> 467c478bd9Sstevel@tonic-gate #include <sys/user.h> 477c478bd9Sstevel@tonic-gate #include <sys/buf.h> 487c478bd9Sstevel@tonic-gate #include <sys/vfs.h> 497c478bd9Sstevel@tonic-gate #include <sys/vnode.h> 507c478bd9Sstevel@tonic-gate #include <sys/proc.h> 517c478bd9Sstevel@tonic-gate #include <sys/disp.h> 527c478bd9Sstevel@tonic-gate #include <sys/file.h> 537c478bd9Sstevel@tonic-gate #include <sys/fcntl.h> 547c478bd9Sstevel@tonic-gate #include <sys/flock.h> 55bc69f433Saguzovsk #include <sys/atomic.h> 567c478bd9Sstevel@tonic-gate #include <sys/kmem.h> 577c478bd9Sstevel@tonic-gate #include <sys/uio.h> 587c478bd9Sstevel@tonic-gate #include <sys/dnlc.h> 597c478bd9Sstevel@tonic-gate #include <sys/conf.h> 607c478bd9Sstevel@tonic-gate #include <sys/mman.h> 617c478bd9Sstevel@tonic-gate #include <sys/pathname.h> 627c478bd9Sstevel@tonic-gate #include <sys/debug.h> 637c478bd9Sstevel@tonic-gate #include <sys/vmsystm.h> 647c478bd9Sstevel@tonic-gate #include <sys/cmn_err.h> 657c478bd9Sstevel@tonic-gate #include <sys/vtrace.h> 667c478bd9Sstevel@tonic-gate #include <sys/filio.h> 677c478bd9Sstevel@tonic-gate #include <sys/policy.h> 687c478bd9Sstevel@tonic-gate 697c478bd9Sstevel@tonic-gate #include <sys/fs/ufs_fs.h> 707c478bd9Sstevel@tonic-gate #include <sys/fs/ufs_lockfs.h> 717c478bd9Sstevel@tonic-gate #include <sys/fs/ufs_filio.h> 727c478bd9Sstevel@tonic-gate #include <sys/fs/ufs_inode.h> 737c478bd9Sstevel@tonic-gate #include <sys/fs/ufs_fsdir.h> 747c478bd9Sstevel@tonic-gate #include <sys/fs/ufs_quota.h> 757c478bd9Sstevel@tonic-gate #include <sys/fs/ufs_log.h> 767c478bd9Sstevel@tonic-gate #include <sys/fs/ufs_snap.h> 777c478bd9Sstevel@tonic-gate #include <sys/fs/ufs_trans.h> 787c478bd9Sstevel@tonic-gate #include <sys/fs/ufs_panic.h> 797c478bd9Sstevel@tonic-gate #include <sys/fs/ufs_bio.h> 807c478bd9Sstevel@tonic-gate #include <sys/dirent.h> /* must be AFTER <sys/fs/fsdir.h>! */ 817c478bd9Sstevel@tonic-gate #include <sys/errno.h> 827c478bd9Sstevel@tonic-gate #include <sys/fssnap_if.h> 837c478bd9Sstevel@tonic-gate #include <sys/unistd.h> 847c478bd9Sstevel@tonic-gate #include <sys/sunddi.h> 857c478bd9Sstevel@tonic-gate 867c478bd9Sstevel@tonic-gate #include <sys/filio.h> /* _FIOIO */ 877c478bd9Sstevel@tonic-gate 887c478bd9Sstevel@tonic-gate #include <vm/hat.h> 897c478bd9Sstevel@tonic-gate #include <vm/page.h> 907c478bd9Sstevel@tonic-gate #include <vm/pvn.h> 917c478bd9Sstevel@tonic-gate #include <vm/as.h> 927c478bd9Sstevel@tonic-gate #include <vm/seg.h> 937c478bd9Sstevel@tonic-gate #include <vm/seg_map.h> 947c478bd9Sstevel@tonic-gate #include <vm/seg_vn.h> 957c478bd9Sstevel@tonic-gate #include <vm/seg_kmem.h> 967c478bd9Sstevel@tonic-gate #include <vm/rm.h> 977c478bd9Sstevel@tonic-gate #include <sys/swap.h> 987c478bd9Sstevel@tonic-gate 997c478bd9Sstevel@tonic-gate #include <fs/fs_subr.h> 1007c478bd9Sstevel@tonic-gate 1017c478bd9Sstevel@tonic-gate static struct instats ins; 1027c478bd9Sstevel@tonic-gate 1037c478bd9Sstevel@tonic-gate static int ufs_getpage_ra(struct vnode *, u_offset_t, struct seg *, caddr_t); 1047c478bd9Sstevel@tonic-gate static int ufs_getpage_miss(struct vnode *, u_offset_t, size_t, struct seg *, 1057c478bd9Sstevel@tonic-gate caddr_t, struct page **, size_t, enum seg_rw, int); 1067c478bd9Sstevel@tonic-gate static int ufs_open(struct vnode **, int, struct cred *); 1077c478bd9Sstevel@tonic-gate static int ufs_close(struct vnode *, int, int, offset_t, struct cred *); 1087c478bd9Sstevel@tonic-gate static int ufs_read(struct vnode *, struct uio *, int, struct cred *, 1097c478bd9Sstevel@tonic-gate struct caller_context *); 1107c478bd9Sstevel@tonic-gate static int ufs_write(struct vnode *, struct uio *, int, struct cred *, 1117c478bd9Sstevel@tonic-gate struct caller_context *); 1127c478bd9Sstevel@tonic-gate static int ufs_ioctl(struct vnode *, int, intptr_t, int, struct cred *, int *); 1137c478bd9Sstevel@tonic-gate static int ufs_getattr(struct vnode *, struct vattr *, int, struct cred *); 1147c478bd9Sstevel@tonic-gate static int ufs_setattr(struct vnode *, struct vattr *, int, struct cred *, 1157c478bd9Sstevel@tonic-gate caller_context_t *); 1167c478bd9Sstevel@tonic-gate static int ufs_access(struct vnode *, int, int, struct cred *); 1177c478bd9Sstevel@tonic-gate static int ufs_lookup(struct vnode *, char *, struct vnode **, 1187c478bd9Sstevel@tonic-gate struct pathname *, int, struct vnode *, struct cred *); 1197c478bd9Sstevel@tonic-gate static int ufs_create(struct vnode *, char *, struct vattr *, enum vcexcl, 1207c478bd9Sstevel@tonic-gate int, struct vnode **, struct cred *, int); 1217c478bd9Sstevel@tonic-gate static int ufs_remove(struct vnode *, char *, struct cred *); 1227c478bd9Sstevel@tonic-gate static int ufs_link(struct vnode *, struct vnode *, char *, struct cred *); 1237c478bd9Sstevel@tonic-gate static int ufs_rename(struct vnode *, char *, struct vnode *, char *, 1247c478bd9Sstevel@tonic-gate struct cred *); 1257c478bd9Sstevel@tonic-gate static int ufs_mkdir(struct vnode *, char *, struct vattr *, struct vnode **, 1267c478bd9Sstevel@tonic-gate struct cred *); 1277c478bd9Sstevel@tonic-gate static int ufs_rmdir(struct vnode *, char *, struct vnode *, struct cred *); 1287c478bd9Sstevel@tonic-gate static int ufs_readdir(struct vnode *, struct uio *, struct cred *, int *); 1297c478bd9Sstevel@tonic-gate static int ufs_symlink(struct vnode *, char *, struct vattr *, char *, 1307c478bd9Sstevel@tonic-gate struct cred *); 1317c478bd9Sstevel@tonic-gate static int ufs_readlink(struct vnode *, struct uio *, struct cred *); 1327c478bd9Sstevel@tonic-gate static int ufs_fsync(struct vnode *, int, struct cred *); 1337c478bd9Sstevel@tonic-gate static void ufs_inactive(struct vnode *, struct cred *); 1347c478bd9Sstevel@tonic-gate static int ufs_fid(struct vnode *, struct fid *); 1357c478bd9Sstevel@tonic-gate static int ufs_rwlock(struct vnode *, int, caller_context_t *); 1367c478bd9Sstevel@tonic-gate static void ufs_rwunlock(struct vnode *, int, caller_context_t *); 1377c478bd9Sstevel@tonic-gate static int ufs_seek(struct vnode *, offset_t, offset_t *); 1387c478bd9Sstevel@tonic-gate static int ufs_frlock(struct vnode *, int, struct flock64 *, int, offset_t, 1397c478bd9Sstevel@tonic-gate struct flk_callback *, struct cred *); 1407c478bd9Sstevel@tonic-gate static int ufs_space(struct vnode *, int, struct flock64 *, int, offset_t, 1417c478bd9Sstevel@tonic-gate cred_t *, caller_context_t *); 1427c478bd9Sstevel@tonic-gate static int ufs_getpage(struct vnode *, offset_t, size_t, uint_t *, 1437c478bd9Sstevel@tonic-gate struct page **, size_t, struct seg *, caddr_t, 1447c478bd9Sstevel@tonic-gate enum seg_rw, struct cred *); 1457c478bd9Sstevel@tonic-gate static int ufs_putpage(struct vnode *, offset_t, size_t, int, struct cred *); 1467c478bd9Sstevel@tonic-gate static int ufs_putpages(struct vnode *, offset_t, size_t, int, struct cred *); 1477c478bd9Sstevel@tonic-gate static int ufs_map(struct vnode *, offset_t, struct as *, caddr_t *, size_t, 1487c478bd9Sstevel@tonic-gate uchar_t, uchar_t, uint_t, struct cred *); 1497c478bd9Sstevel@tonic-gate static int ufs_addmap(struct vnode *, offset_t, struct as *, caddr_t, size_t, 1507c478bd9Sstevel@tonic-gate uchar_t, uchar_t, uint_t, struct cred *); 1517c478bd9Sstevel@tonic-gate static int ufs_delmap(struct vnode *, offset_t, struct as *, caddr_t, size_t, 1527c478bd9Sstevel@tonic-gate uint_t, uint_t, uint_t, struct cred *); 1537c478bd9Sstevel@tonic-gate static int ufs_poll(vnode_t *, short, int, short *, struct pollhead **); 1547c478bd9Sstevel@tonic-gate static int ufs_dump(vnode_t *, caddr_t, int, int); 1557c478bd9Sstevel@tonic-gate static int ufs_l_pathconf(struct vnode *, int, ulong_t *, struct cred *); 1567c478bd9Sstevel@tonic-gate static int ufs_pageio(struct vnode *, struct page *, u_offset_t, size_t, int, 1577c478bd9Sstevel@tonic-gate struct cred *); 1587c478bd9Sstevel@tonic-gate static int ufs_dump(vnode_t *, caddr_t, int, int); 1597c478bd9Sstevel@tonic-gate static int ufs_dumpctl(vnode_t *, int, int *); 1607c478bd9Sstevel@tonic-gate static daddr32_t *save_dblks(struct inode *, struct ufsvfs *, daddr32_t *, 1617c478bd9Sstevel@tonic-gate daddr32_t *, int, int); 1627c478bd9Sstevel@tonic-gate static int ufs_getsecattr(struct vnode *, vsecattr_t *, int, struct cred *); 1637c478bd9Sstevel@tonic-gate static int ufs_setsecattr(struct vnode *, vsecattr_t *, int, struct cred *); 1647c478bd9Sstevel@tonic-gate 165*02ff05a9Svsakar extern int as_map_locked(struct as *, caddr_t, size_t, int ((*)()), void *); 166*02ff05a9Svsakar 1677c478bd9Sstevel@tonic-gate /* 1687c478bd9Sstevel@tonic-gate * For lockfs: ulockfs begin/end is now inlined in the ufs_xxx functions. 1697c478bd9Sstevel@tonic-gate * 1707c478bd9Sstevel@tonic-gate * XXX - ULOCKFS in fs_pathconf and ufs_ioctl is not inlined yet. 1717c478bd9Sstevel@tonic-gate */ 1727c478bd9Sstevel@tonic-gate struct vnodeops *ufs_vnodeops; 1737c478bd9Sstevel@tonic-gate 1747c478bd9Sstevel@tonic-gate const fs_operation_def_t ufs_vnodeops_template[] = { 1757c478bd9Sstevel@tonic-gate VOPNAME_OPEN, ufs_open, /* will not be blocked by lockfs */ 1767c478bd9Sstevel@tonic-gate VOPNAME_CLOSE, ufs_close, /* will not be blocked by lockfs */ 1777c478bd9Sstevel@tonic-gate VOPNAME_READ, ufs_read, 1787c478bd9Sstevel@tonic-gate VOPNAME_WRITE, ufs_write, 1797c478bd9Sstevel@tonic-gate VOPNAME_IOCTL, ufs_ioctl, 1807c478bd9Sstevel@tonic-gate VOPNAME_GETATTR, ufs_getattr, 1817c478bd9Sstevel@tonic-gate VOPNAME_SETATTR, ufs_setattr, 1827c478bd9Sstevel@tonic-gate VOPNAME_ACCESS, ufs_access, 1837c478bd9Sstevel@tonic-gate VOPNAME_LOOKUP, ufs_lookup, 1847c478bd9Sstevel@tonic-gate VOPNAME_CREATE, ufs_create, 1857c478bd9Sstevel@tonic-gate VOPNAME_REMOVE, ufs_remove, 1867c478bd9Sstevel@tonic-gate VOPNAME_LINK, ufs_link, 1877c478bd9Sstevel@tonic-gate VOPNAME_RENAME, ufs_rename, 1887c478bd9Sstevel@tonic-gate VOPNAME_MKDIR, ufs_mkdir, 1897c478bd9Sstevel@tonic-gate VOPNAME_RMDIR, ufs_rmdir, 1907c478bd9Sstevel@tonic-gate VOPNAME_READDIR, ufs_readdir, 1917c478bd9Sstevel@tonic-gate VOPNAME_SYMLINK, ufs_symlink, 1927c478bd9Sstevel@tonic-gate VOPNAME_READLINK, ufs_readlink, 1937c478bd9Sstevel@tonic-gate VOPNAME_FSYNC, ufs_fsync, 1947c478bd9Sstevel@tonic-gate VOPNAME_INACTIVE, (fs_generic_func_p) ufs_inactive, /* not blocked */ 1957c478bd9Sstevel@tonic-gate VOPNAME_FID, ufs_fid, 1967c478bd9Sstevel@tonic-gate VOPNAME_RWLOCK, ufs_rwlock, /* not blocked */ 1977c478bd9Sstevel@tonic-gate VOPNAME_RWUNLOCK, (fs_generic_func_p) ufs_rwunlock, /* not blocked */ 1987c478bd9Sstevel@tonic-gate VOPNAME_SEEK, ufs_seek, 1997c478bd9Sstevel@tonic-gate VOPNAME_FRLOCK, ufs_frlock, 2007c478bd9Sstevel@tonic-gate VOPNAME_SPACE, ufs_space, 2017c478bd9Sstevel@tonic-gate VOPNAME_GETPAGE, ufs_getpage, 2027c478bd9Sstevel@tonic-gate VOPNAME_PUTPAGE, ufs_putpage, 2037c478bd9Sstevel@tonic-gate VOPNAME_MAP, (fs_generic_func_p) ufs_map, 2047c478bd9Sstevel@tonic-gate VOPNAME_ADDMAP, (fs_generic_func_p) ufs_addmap, /* not blocked */ 2057c478bd9Sstevel@tonic-gate VOPNAME_DELMAP, ufs_delmap, /* will not be blocked by lockfs */ 2067c478bd9Sstevel@tonic-gate VOPNAME_POLL, (fs_generic_func_p) ufs_poll, /* not blocked */ 2077c478bd9Sstevel@tonic-gate VOPNAME_DUMP, ufs_dump, 2087c478bd9Sstevel@tonic-gate VOPNAME_PATHCONF, ufs_l_pathconf, 2097c478bd9Sstevel@tonic-gate VOPNAME_PAGEIO, ufs_pageio, 2107c478bd9Sstevel@tonic-gate VOPNAME_DUMPCTL, ufs_dumpctl, 2117c478bd9Sstevel@tonic-gate VOPNAME_GETSECATTR, ufs_getsecattr, 2127c478bd9Sstevel@tonic-gate VOPNAME_SETSECATTR, ufs_setsecattr, 2137c478bd9Sstevel@tonic-gate VOPNAME_VNEVENT, fs_vnevent_support, 2147c478bd9Sstevel@tonic-gate NULL, NULL 2157c478bd9Sstevel@tonic-gate }; 2167c478bd9Sstevel@tonic-gate 2177c478bd9Sstevel@tonic-gate #define MAX_BACKFILE_COUNT 9999 2187c478bd9Sstevel@tonic-gate 2197c478bd9Sstevel@tonic-gate /* 2207c478bd9Sstevel@tonic-gate * Created by ufs_dumpctl() to store a file's disk block info into memory. 2217c478bd9Sstevel@tonic-gate * Used by ufs_dump() to dump data to disk directly. 2227c478bd9Sstevel@tonic-gate */ 2237c478bd9Sstevel@tonic-gate struct dump { 2247c478bd9Sstevel@tonic-gate struct inode *ip; /* the file we contain */ 2257c478bd9Sstevel@tonic-gate daddr_t fsbs; /* number of blocks stored */ 2267c478bd9Sstevel@tonic-gate struct timeval32 time; /* time stamp for the struct */ 2277c478bd9Sstevel@tonic-gate daddr32_t dblk[1]; /* place holder for block info */ 2287c478bd9Sstevel@tonic-gate }; 2297c478bd9Sstevel@tonic-gate 2307c478bd9Sstevel@tonic-gate static struct dump *dump_info = NULL; 2317c478bd9Sstevel@tonic-gate 2327c478bd9Sstevel@tonic-gate /* 2337c478bd9Sstevel@tonic-gate * Previously there was no special action required for ordinary files. 2347c478bd9Sstevel@tonic-gate * (Devices are handled through the device file system.) 2357c478bd9Sstevel@tonic-gate * Now we support Large Files and Large File API requires open to 2367c478bd9Sstevel@tonic-gate * fail if file is large. 2377c478bd9Sstevel@tonic-gate * We could take care to prevent data corruption 2387c478bd9Sstevel@tonic-gate * by doing an atomic check of size and truncate if file is opened with 2397c478bd9Sstevel@tonic-gate * FTRUNC flag set but traditionally this is being done by the vfs/vnode 2407c478bd9Sstevel@tonic-gate * layers. So taking care of truncation here is a change in the existing 2417c478bd9Sstevel@tonic-gate * semantics of VOP_OPEN and therefore we chose not to implement any thing 2427c478bd9Sstevel@tonic-gate * here. The check for the size of the file > 2GB is being done at the 2437c478bd9Sstevel@tonic-gate * vfs layer in routine vn_open(). 2447c478bd9Sstevel@tonic-gate */ 2457c478bd9Sstevel@tonic-gate 2467c478bd9Sstevel@tonic-gate /* ARGSUSED */ 2477c478bd9Sstevel@tonic-gate static int 2487c478bd9Sstevel@tonic-gate ufs_open(struct vnode **vpp, int flag, struct cred *cr) 2497c478bd9Sstevel@tonic-gate { 2507c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_OPEN, "ufs_open:vpp %p", vpp); 2517c478bd9Sstevel@tonic-gate return (0); 2527c478bd9Sstevel@tonic-gate } 2537c478bd9Sstevel@tonic-gate 2547c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 2557c478bd9Sstevel@tonic-gate static int 2567c478bd9Sstevel@tonic-gate ufs_close(struct vnode *vp, int flag, int count, offset_t offset, 2577c478bd9Sstevel@tonic-gate struct cred *cr) 2587c478bd9Sstevel@tonic-gate { 2597c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_CLOSE, "ufs_close:vp %p", vp); 2607c478bd9Sstevel@tonic-gate 2617c478bd9Sstevel@tonic-gate cleanlocks(vp, ttoproc(curthread)->p_pid, 0); 2627c478bd9Sstevel@tonic-gate cleanshares(vp, ttoproc(curthread)->p_pid); 2637c478bd9Sstevel@tonic-gate 2647c478bd9Sstevel@tonic-gate /* 2657c478bd9Sstevel@tonic-gate * Push partially filled cluster at last close. 2667c478bd9Sstevel@tonic-gate * ``last close'' is approximated because the dnlc 2677c478bd9Sstevel@tonic-gate * may have a hold on the vnode. 2687c478bd9Sstevel@tonic-gate * Checking for VBAD here will also act as a forced umount check. 2697c478bd9Sstevel@tonic-gate */ 2707c478bd9Sstevel@tonic-gate if (vp->v_count <= 2 && vp->v_type != VBAD) { 2717c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 2727c478bd9Sstevel@tonic-gate if (ip->i_delaylen) { 2737c478bd9Sstevel@tonic-gate ins.in_poc.value.ul++; 2747c478bd9Sstevel@tonic-gate (void) ufs_putpages(vp, ip->i_delayoff, ip->i_delaylen, 2757c478bd9Sstevel@tonic-gate B_ASYNC | B_FREE, cr); 2767c478bd9Sstevel@tonic-gate ip->i_delaylen = 0; 2777c478bd9Sstevel@tonic-gate } 2787c478bd9Sstevel@tonic-gate } 2797c478bd9Sstevel@tonic-gate 2807c478bd9Sstevel@tonic-gate return (0); 2817c478bd9Sstevel@tonic-gate } 2827c478bd9Sstevel@tonic-gate 2837c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 2847c478bd9Sstevel@tonic-gate static int 2857c478bd9Sstevel@tonic-gate ufs_read(struct vnode *vp, struct uio *uiop, int ioflag, struct cred *cr, 2867c478bd9Sstevel@tonic-gate struct caller_context *ct) 2877c478bd9Sstevel@tonic-gate { 2887c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 2897c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp; 2907c478bd9Sstevel@tonic-gate struct ulockfs *ulp = NULL; 2917c478bd9Sstevel@tonic-gate int error = 0; 2927c478bd9Sstevel@tonic-gate int intrans = 0; 2937c478bd9Sstevel@tonic-gate 2947c478bd9Sstevel@tonic-gate ASSERT(RW_READ_HELD(&ip->i_rwlock)); 2957c478bd9Sstevel@tonic-gate TRACE_3(TR_FAC_UFS, TR_UFS_READ_START, 2967c478bd9Sstevel@tonic-gate "ufs_read_start:vp %p uiop %p ioflag %x", 2977c478bd9Sstevel@tonic-gate vp, uiop, ioflag); 2987c478bd9Sstevel@tonic-gate 2997c478bd9Sstevel@tonic-gate /* 3007c478bd9Sstevel@tonic-gate * Mandatory locking needs to be done before ufs_lockfs_begin() 3017c478bd9Sstevel@tonic-gate * and TRANS_BEGIN_SYNC() calls since mandatory locks can sleep. 3027c478bd9Sstevel@tonic-gate */ 3037c478bd9Sstevel@tonic-gate if (MANDLOCK(vp, ip->i_mode)) { 3047c478bd9Sstevel@tonic-gate /* 3057c478bd9Sstevel@tonic-gate * ufs_getattr ends up being called by chklock 3067c478bd9Sstevel@tonic-gate */ 3077c478bd9Sstevel@tonic-gate error = chklock(vp, FREAD, uiop->uio_loffset, 3087c478bd9Sstevel@tonic-gate uiop->uio_resid, uiop->uio_fmode, ct); 3097c478bd9Sstevel@tonic-gate if (error) 3107c478bd9Sstevel@tonic-gate goto out; 3117c478bd9Sstevel@tonic-gate } 3127c478bd9Sstevel@tonic-gate 3137c478bd9Sstevel@tonic-gate ufsvfsp = ip->i_ufsvfs; 3147c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_READ_MASK); 3157c478bd9Sstevel@tonic-gate if (error) 3167c478bd9Sstevel@tonic-gate goto out; 3177c478bd9Sstevel@tonic-gate 3187c478bd9Sstevel@tonic-gate /* 3197c478bd9Sstevel@tonic-gate * In the case that a directory is opened for reading as a file 3207c478bd9Sstevel@tonic-gate * (eg "cat .") with the O_RSYNC, O_SYNC and O_DSYNC flags set. 3217c478bd9Sstevel@tonic-gate * The locking order had to be changed to avoid a deadlock with 3227c478bd9Sstevel@tonic-gate * an update taking place on that directory at the same time. 3237c478bd9Sstevel@tonic-gate */ 3247c478bd9Sstevel@tonic-gate if ((ip->i_mode & IFMT) == IFDIR) { 3257c478bd9Sstevel@tonic-gate 3267c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_READER); 3277c478bd9Sstevel@tonic-gate error = rdip(ip, uiop, ioflag, cr); 3287c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 3297c478bd9Sstevel@tonic-gate 3307c478bd9Sstevel@tonic-gate if (error) { 3317c478bd9Sstevel@tonic-gate if (ulp) 3327c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 3337c478bd9Sstevel@tonic-gate goto out; 3347c478bd9Sstevel@tonic-gate } 3357c478bd9Sstevel@tonic-gate 3367c478bd9Sstevel@tonic-gate if (ulp && (ioflag & FRSYNC) && (ioflag & (FSYNC | FDSYNC)) && 3377c478bd9Sstevel@tonic-gate TRANS_ISTRANS(ufsvfsp)) { 3387c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 3397c478bd9Sstevel@tonic-gate TRANS_BEGIN_SYNC(ufsvfsp, TOP_READ_SYNC, TOP_READ_SIZE, 3407c478bd9Sstevel@tonic-gate error); 3417c478bd9Sstevel@tonic-gate ASSERT(!error); 3427c478bd9Sstevel@tonic-gate TRANS_END_SYNC(ufsvfsp, error, TOP_READ_SYNC, 3437c478bd9Sstevel@tonic-gate TOP_READ_SIZE); 3447c478bd9Sstevel@tonic-gate rw_enter(&ip->i_rwlock, RW_READER); 3457c478bd9Sstevel@tonic-gate } 3467c478bd9Sstevel@tonic-gate } else { 3477c478bd9Sstevel@tonic-gate /* 3487c478bd9Sstevel@tonic-gate * Only transact reads to files opened for sync-read and 3497c478bd9Sstevel@tonic-gate * sync-write on a file system that is not write locked. 3507c478bd9Sstevel@tonic-gate * 3517c478bd9Sstevel@tonic-gate * The ``not write locked'' check prevents problems with 3527c478bd9Sstevel@tonic-gate * enabling/disabling logging on a busy file system. E.g., 3537c478bd9Sstevel@tonic-gate * logging exists at the beginning of the read but does not 3547c478bd9Sstevel@tonic-gate * at the end. 3557c478bd9Sstevel@tonic-gate * 3567c478bd9Sstevel@tonic-gate */ 3577c478bd9Sstevel@tonic-gate if (ulp && (ioflag & FRSYNC) && (ioflag & (FSYNC | FDSYNC)) && 3587c478bd9Sstevel@tonic-gate TRANS_ISTRANS(ufsvfsp)) { 3597c478bd9Sstevel@tonic-gate TRANS_BEGIN_SYNC(ufsvfsp, TOP_READ_SYNC, TOP_READ_SIZE, 3607c478bd9Sstevel@tonic-gate error); 3617c478bd9Sstevel@tonic-gate ASSERT(!error); 3627c478bd9Sstevel@tonic-gate intrans = 1; 3637c478bd9Sstevel@tonic-gate } 3647c478bd9Sstevel@tonic-gate 3657c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_READER); 3667c478bd9Sstevel@tonic-gate error = rdip(ip, uiop, ioflag, cr); 3677c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 3687c478bd9Sstevel@tonic-gate 3697c478bd9Sstevel@tonic-gate if (intrans) { 3707c478bd9Sstevel@tonic-gate TRANS_END_SYNC(ufsvfsp, error, TOP_READ_SYNC, 3717c478bd9Sstevel@tonic-gate TOP_READ_SIZE); 3727c478bd9Sstevel@tonic-gate } 3737c478bd9Sstevel@tonic-gate } 3747c478bd9Sstevel@tonic-gate 3757c478bd9Sstevel@tonic-gate if (ulp) { 3767c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 3777c478bd9Sstevel@tonic-gate } 3787c478bd9Sstevel@tonic-gate out: 3797c478bd9Sstevel@tonic-gate 3807c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_READ_END, 3817c478bd9Sstevel@tonic-gate "ufs_read_end:vp %p error %d", vp, error); 3827c478bd9Sstevel@tonic-gate return (error); 3837c478bd9Sstevel@tonic-gate } 3847c478bd9Sstevel@tonic-gate 3857c478bd9Sstevel@tonic-gate extern int ufs_HW; /* high water mark */ 3867c478bd9Sstevel@tonic-gate extern int ufs_LW; /* low water mark */ 3877c478bd9Sstevel@tonic-gate int ufs_WRITES = 1; /* XXX - enable/disable */ 3887c478bd9Sstevel@tonic-gate int ufs_throttles = 0; /* throttling count */ 3897c478bd9Sstevel@tonic-gate int ufs_allow_shared_writes = 1; /* directio shared writes */ 3907c478bd9Sstevel@tonic-gate 3917c478bd9Sstevel@tonic-gate static int 3927c478bd9Sstevel@tonic-gate ufs_check_rewrite(struct inode *ip, struct uio *uiop, int ioflag) 3937c478bd9Sstevel@tonic-gate { 394f90bab20Sswilcox int shared_write; 395f90bab20Sswilcox 396f90bab20Sswilcox /* 397f90bab20Sswilcox * If the FDSYNC flag is set then ignore the global 398f90bab20Sswilcox * ufs_allow_shared_writes in this case. 399f90bab20Sswilcox */ 400f90bab20Sswilcox shared_write = (ioflag & FDSYNC) | ufs_allow_shared_writes; 4017c478bd9Sstevel@tonic-gate 4027c478bd9Sstevel@tonic-gate /* 4037c478bd9Sstevel@tonic-gate * Filter to determine if this request is suitable as a 4047c478bd9Sstevel@tonic-gate * concurrent rewrite. This write must not allocate blocks 4057c478bd9Sstevel@tonic-gate * by extending the file or filling in holes. No use trying 4067c478bd9Sstevel@tonic-gate * through FSYNC descriptors as the inode will be synchronously 4077c478bd9Sstevel@tonic-gate * updated after the write. The uio structure has not yet been 4087c478bd9Sstevel@tonic-gate * checked for sanity, so assume nothing. 4097c478bd9Sstevel@tonic-gate */ 4107c478bd9Sstevel@tonic-gate return (((ip->i_mode & IFMT) == IFREG) && !(ioflag & FAPPEND) && 4117c478bd9Sstevel@tonic-gate (uiop->uio_loffset >= (offset_t)0) && 4127c478bd9Sstevel@tonic-gate (uiop->uio_loffset < ip->i_size) && (uiop->uio_resid > 0) && 4137c478bd9Sstevel@tonic-gate ((ip->i_size - uiop->uio_loffset) >= uiop->uio_resid) && 4147c478bd9Sstevel@tonic-gate !(ioflag & FSYNC) && !bmap_has_holes(ip) && 415f90bab20Sswilcox shared_write); 4167c478bd9Sstevel@tonic-gate } 4177c478bd9Sstevel@tonic-gate 4187c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 4197c478bd9Sstevel@tonic-gate static int 4207c478bd9Sstevel@tonic-gate ufs_write(struct vnode *vp, struct uio *uiop, int ioflag, cred_t *cr, 4217c478bd9Sstevel@tonic-gate caller_context_t *ct) 4227c478bd9Sstevel@tonic-gate { 4237c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 4247c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp; 4257c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 4267c478bd9Sstevel@tonic-gate int retry = 1; 4277c478bd9Sstevel@tonic-gate int error, resv, resid = 0; 4287c478bd9Sstevel@tonic-gate int directio_status; 4297c478bd9Sstevel@tonic-gate int exclusive; 430f90bab20Sswilcox int rewriteflg; 4317c478bd9Sstevel@tonic-gate long start_resid = uiop->uio_resid; 4327c478bd9Sstevel@tonic-gate 4337c478bd9Sstevel@tonic-gate TRACE_3(TR_FAC_UFS, TR_UFS_WRITE_START, 4347c478bd9Sstevel@tonic-gate "ufs_write_start:vp %p uiop %p ioflag %x", 4357c478bd9Sstevel@tonic-gate vp, uiop, ioflag); 4367c478bd9Sstevel@tonic-gate 4377c478bd9Sstevel@tonic-gate ASSERT(RW_LOCK_HELD(&ip->i_rwlock)); 4387c478bd9Sstevel@tonic-gate 4397c478bd9Sstevel@tonic-gate retry_mandlock: 4407c478bd9Sstevel@tonic-gate /* 4417c478bd9Sstevel@tonic-gate * Mandatory locking needs to be done before ufs_lockfs_begin() 4427c478bd9Sstevel@tonic-gate * and TRANS_BEGIN_[A]SYNC() calls since mandatory locks can sleep. 4437c478bd9Sstevel@tonic-gate * Check for forced unmounts normally done in ufs_lockfs_begin(). 4447c478bd9Sstevel@tonic-gate */ 4457c478bd9Sstevel@tonic-gate if ((ufsvfsp = ip->i_ufsvfs) == NULL) { 4467c478bd9Sstevel@tonic-gate error = EIO; 4477c478bd9Sstevel@tonic-gate goto out; 4487c478bd9Sstevel@tonic-gate } 4497c478bd9Sstevel@tonic-gate if (MANDLOCK(vp, ip->i_mode)) { 4507c478bd9Sstevel@tonic-gate 4517c478bd9Sstevel@tonic-gate ASSERT(RW_WRITE_HELD(&ip->i_rwlock)); 4527c478bd9Sstevel@tonic-gate 4537c478bd9Sstevel@tonic-gate /* 4547c478bd9Sstevel@tonic-gate * ufs_getattr ends up being called by chklock 4557c478bd9Sstevel@tonic-gate */ 4567c478bd9Sstevel@tonic-gate error = chklock(vp, FWRITE, uiop->uio_loffset, 4577c478bd9Sstevel@tonic-gate uiop->uio_resid, uiop->uio_fmode, ct); 4587c478bd9Sstevel@tonic-gate if (error) 4597c478bd9Sstevel@tonic-gate goto out; 4607c478bd9Sstevel@tonic-gate } 4617c478bd9Sstevel@tonic-gate 4627c478bd9Sstevel@tonic-gate /* i_rwlock can change in chklock */ 4637c478bd9Sstevel@tonic-gate exclusive = rw_write_held(&ip->i_rwlock); 464f90bab20Sswilcox rewriteflg = ufs_check_rewrite(ip, uiop, ioflag); 4657c478bd9Sstevel@tonic-gate 4667c478bd9Sstevel@tonic-gate /* 4677c478bd9Sstevel@tonic-gate * Check for fast-path special case of directio re-writes. 4687c478bd9Sstevel@tonic-gate */ 4697c478bd9Sstevel@tonic-gate if ((ip->i_flag & IDIRECTIO || ufsvfsp->vfs_forcedirectio) && 470f90bab20Sswilcox !exclusive && rewriteflg) { 4717c478bd9Sstevel@tonic-gate 4727c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_WRITE_MASK); 4737c478bd9Sstevel@tonic-gate if (error) 4747c478bd9Sstevel@tonic-gate goto out; 4757c478bd9Sstevel@tonic-gate 4767c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_READER); 4777c478bd9Sstevel@tonic-gate error = ufs_directio_write(ip, uiop, ioflag, 1, cr, 4787c478bd9Sstevel@tonic-gate &directio_status); 4797c478bd9Sstevel@tonic-gate if (directio_status == DIRECTIO_SUCCESS) { 4807c478bd9Sstevel@tonic-gate uint_t i_flag_save; 4817c478bd9Sstevel@tonic-gate 4827c478bd9Sstevel@tonic-gate if (start_resid != uiop->uio_resid) 4837c478bd9Sstevel@tonic-gate error = 0; 4847c478bd9Sstevel@tonic-gate /* 4857c478bd9Sstevel@tonic-gate * Special treatment of access times for re-writes. 4867c478bd9Sstevel@tonic-gate * If IMOD is not already set, then convert it 4877c478bd9Sstevel@tonic-gate * to IMODACC for this operation. This defers 4887c478bd9Sstevel@tonic-gate * entering a delta into the log until the inode 4897c478bd9Sstevel@tonic-gate * is flushed. This mimics what is done for read 4907c478bd9Sstevel@tonic-gate * operations and inode access time. 4917c478bd9Sstevel@tonic-gate */ 4927c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 4937c478bd9Sstevel@tonic-gate i_flag_save = ip->i_flag; 4947c478bd9Sstevel@tonic-gate ip->i_flag |= IUPD | ICHG; 4957c478bd9Sstevel@tonic-gate ip->i_seq++; 4967c478bd9Sstevel@tonic-gate ITIMES_NOLOCK(ip); 4977c478bd9Sstevel@tonic-gate if ((i_flag_save & IMOD) == 0) { 4987c478bd9Sstevel@tonic-gate ip->i_flag &= ~IMOD; 4997c478bd9Sstevel@tonic-gate ip->i_flag |= IMODACC; 5007c478bd9Sstevel@tonic-gate } 5017c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 5027c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 5037c478bd9Sstevel@tonic-gate if (ulp) 5047c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 5057c478bd9Sstevel@tonic-gate goto out; 5067c478bd9Sstevel@tonic-gate } 5077c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 5087c478bd9Sstevel@tonic-gate if (ulp) 5097c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 5107c478bd9Sstevel@tonic-gate } 5117c478bd9Sstevel@tonic-gate 5127c478bd9Sstevel@tonic-gate if (!exclusive && !rw_tryupgrade(&ip->i_rwlock)) { 5137c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 5147c478bd9Sstevel@tonic-gate rw_enter(&ip->i_rwlock, RW_WRITER); 5157c478bd9Sstevel@tonic-gate /* 5167c478bd9Sstevel@tonic-gate * Mandatory locking could have been enabled 5177c478bd9Sstevel@tonic-gate * after dropping the i_rwlock. 5187c478bd9Sstevel@tonic-gate */ 5197c478bd9Sstevel@tonic-gate if (MANDLOCK(vp, ip->i_mode)) 5207c478bd9Sstevel@tonic-gate goto retry_mandlock; 5217c478bd9Sstevel@tonic-gate } 5227c478bd9Sstevel@tonic-gate 5237c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_WRITE_MASK); 5247c478bd9Sstevel@tonic-gate if (error) 5257c478bd9Sstevel@tonic-gate goto out; 5267c478bd9Sstevel@tonic-gate 5277c478bd9Sstevel@tonic-gate /* 5287c478bd9Sstevel@tonic-gate * Amount of log space needed for this write 5297c478bd9Sstevel@tonic-gate */ 530f90bab20Sswilcox if (!rewriteflg || !(ioflag & FDSYNC)) 531f90bab20Sswilcox TRANS_WRITE_RESV(ip, uiop, ulp, &resv, &resid); 5327c478bd9Sstevel@tonic-gate 5337c478bd9Sstevel@tonic-gate /* 5347c478bd9Sstevel@tonic-gate * Throttle writes. 5357c478bd9Sstevel@tonic-gate */ 5367c478bd9Sstevel@tonic-gate if (ufs_WRITES && (ip->i_writes > ufs_HW)) { 5377c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 5387c478bd9Sstevel@tonic-gate while (ip->i_writes > ufs_HW) { 5397c478bd9Sstevel@tonic-gate ufs_throttles++; 5407c478bd9Sstevel@tonic-gate cv_wait(&ip->i_wrcv, &ip->i_tlock); 5417c478bd9Sstevel@tonic-gate } 5427c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 5437c478bd9Sstevel@tonic-gate } 5447c478bd9Sstevel@tonic-gate 5457c478bd9Sstevel@tonic-gate /* 5467c478bd9Sstevel@tonic-gate * Enter Transaction 547f90bab20Sswilcox * 548f90bab20Sswilcox * If the write is a rewrite there is no need to open a transaction 549f90bab20Sswilcox * if the FDSYNC flag is set and not the FSYNC. In this case just 550f90bab20Sswilcox * set the IMODACC flag to modify do the update at a later time 551f90bab20Sswilcox * thus avoiding the overhead of the logging transaction that is 552f90bab20Sswilcox * not required. 5537c478bd9Sstevel@tonic-gate */ 5547c478bd9Sstevel@tonic-gate if (ioflag & (FSYNC|FDSYNC)) { 5557c478bd9Sstevel@tonic-gate if (ulp) { 556f90bab20Sswilcox if (rewriteflg) { 557f90bab20Sswilcox uint_t i_flag_save; 558f90bab20Sswilcox 559f90bab20Sswilcox rw_enter(&ip->i_contents, RW_READER); 560f90bab20Sswilcox mutex_enter(&ip->i_tlock); 561f90bab20Sswilcox i_flag_save = ip->i_flag; 562f90bab20Sswilcox ip->i_flag |= IUPD | ICHG; 563f90bab20Sswilcox ip->i_seq++; 564f90bab20Sswilcox ITIMES_NOLOCK(ip); 565f90bab20Sswilcox if ((i_flag_save & IMOD) == 0) { 566f90bab20Sswilcox ip->i_flag &= ~IMOD; 567f90bab20Sswilcox ip->i_flag |= IMODACC; 568f90bab20Sswilcox } 569f90bab20Sswilcox mutex_exit(&ip->i_tlock); 570f90bab20Sswilcox rw_exit(&ip->i_contents); 571f90bab20Sswilcox } else { 572f90bab20Sswilcox int terr = 0; 573f90bab20Sswilcox TRANS_BEGIN_SYNC(ufsvfsp, TOP_WRITE_SYNC, resv, 574f90bab20Sswilcox terr); 575f90bab20Sswilcox ASSERT(!terr); 576f90bab20Sswilcox } 5777c478bd9Sstevel@tonic-gate } 5787c478bd9Sstevel@tonic-gate } else { 5797c478bd9Sstevel@tonic-gate if (ulp) 5807c478bd9Sstevel@tonic-gate TRANS_BEGIN_ASYNC(ufsvfsp, TOP_WRITE, resv); 5817c478bd9Sstevel@tonic-gate } 5827c478bd9Sstevel@tonic-gate 5837c478bd9Sstevel@tonic-gate /* 5847c478bd9Sstevel@tonic-gate * Write the file 5857c478bd9Sstevel@tonic-gate */ 5867c478bd9Sstevel@tonic-gate rw_enter(&ufsvfsp->vfs_dqrwlock, RW_READER); 5877c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_WRITER); 5887c478bd9Sstevel@tonic-gate if ((ioflag & FAPPEND) != 0 && (ip->i_mode & IFMT) == IFREG) { 5897c478bd9Sstevel@tonic-gate /* 5907c478bd9Sstevel@tonic-gate * In append mode start at end of file. 5917c478bd9Sstevel@tonic-gate */ 5927c478bd9Sstevel@tonic-gate uiop->uio_loffset = ip->i_size; 5937c478bd9Sstevel@tonic-gate } 5947c478bd9Sstevel@tonic-gate 5957c478bd9Sstevel@tonic-gate /* 5967c478bd9Sstevel@tonic-gate * Mild optimisation, don't call ufs_trans_write() unless we have to 5977c478bd9Sstevel@tonic-gate * Also, suppress file system full messages if we will retry. 5987c478bd9Sstevel@tonic-gate */ 5997c478bd9Sstevel@tonic-gate if (retry) 6007c478bd9Sstevel@tonic-gate ip->i_flag |= IQUIET; 6017c478bd9Sstevel@tonic-gate if (resid) { 6027c478bd9Sstevel@tonic-gate TRANS_WRITE(ip, uiop, ioflag, error, ulp, cr, resv, resid); 6037c478bd9Sstevel@tonic-gate } else { 6047c478bd9Sstevel@tonic-gate error = wrip(ip, uiop, ioflag, cr); 6057c478bd9Sstevel@tonic-gate } 6067c478bd9Sstevel@tonic-gate ip->i_flag &= ~IQUIET; 6077c478bd9Sstevel@tonic-gate 6087c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 6097c478bd9Sstevel@tonic-gate rw_exit(&ufsvfsp->vfs_dqrwlock); 6107c478bd9Sstevel@tonic-gate 6117c478bd9Sstevel@tonic-gate /* 6127c478bd9Sstevel@tonic-gate * Leave Transaction 6137c478bd9Sstevel@tonic-gate */ 6147c478bd9Sstevel@tonic-gate if (ulp) { 6157c478bd9Sstevel@tonic-gate if (ioflag & (FSYNC|FDSYNC)) { 616f90bab20Sswilcox if (!rewriteflg) { 617f90bab20Sswilcox int terr = 0; 618f90bab20Sswilcox 619f90bab20Sswilcox TRANS_END_SYNC(ufsvfsp, terr, TOP_WRITE_SYNC, 620f90bab20Sswilcox resv); 621f90bab20Sswilcox if (error == 0) 622f90bab20Sswilcox error = terr; 623f90bab20Sswilcox } 6247c478bd9Sstevel@tonic-gate } else { 6257c478bd9Sstevel@tonic-gate TRANS_END_ASYNC(ufsvfsp, TOP_WRITE, resv); 6267c478bd9Sstevel@tonic-gate } 6277c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 6287c478bd9Sstevel@tonic-gate } 6297c478bd9Sstevel@tonic-gate out: 6307c478bd9Sstevel@tonic-gate if ((error == ENOSPC) && retry && TRANS_ISTRANS(ufsvfsp)) { 6317c478bd9Sstevel@tonic-gate /* 6327c478bd9Sstevel@tonic-gate * Any blocks tied up in pending deletes? 6337c478bd9Sstevel@tonic-gate */ 6347c478bd9Sstevel@tonic-gate ufs_delete_drain_wait(ufsvfsp, 1); 6357c478bd9Sstevel@tonic-gate retry = 0; 6367c478bd9Sstevel@tonic-gate goto retry_mandlock; 6377c478bd9Sstevel@tonic-gate } 6387c478bd9Sstevel@tonic-gate 6397c478bd9Sstevel@tonic-gate if (error == ENOSPC && (start_resid != uiop->uio_resid)) 6407c478bd9Sstevel@tonic-gate error = 0; 6417c478bd9Sstevel@tonic-gate 6427c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_WRITE_END, 6437c478bd9Sstevel@tonic-gate "ufs_write_end:vp %p error %d", vp, error); 6447c478bd9Sstevel@tonic-gate return (error); 6457c478bd9Sstevel@tonic-gate } 6467c478bd9Sstevel@tonic-gate 6477c478bd9Sstevel@tonic-gate /* 6487c478bd9Sstevel@tonic-gate * Don't cache write blocks to files with the sticky bit set. 6497c478bd9Sstevel@tonic-gate * Used to keep swap files from blowing the page cache on a server. 6507c478bd9Sstevel@tonic-gate */ 6517c478bd9Sstevel@tonic-gate int stickyhack = 1; 6527c478bd9Sstevel@tonic-gate 6537c478bd9Sstevel@tonic-gate /* 6547c478bd9Sstevel@tonic-gate * Free behind hacks. The pager is busted. 6557c478bd9Sstevel@tonic-gate * XXX - need to pass the information down to writedone() in a flag like B_SEQ 6567c478bd9Sstevel@tonic-gate * or B_FREE_IF_TIGHT_ON_MEMORY. 6577c478bd9Sstevel@tonic-gate */ 6587c478bd9Sstevel@tonic-gate int freebehind = 1; 659562eee46Srbourbon int smallfile = 0; 660562eee46Srbourbon u_offset_t smallfile64 = 32 * 1024; 6617c478bd9Sstevel@tonic-gate 6627c478bd9Sstevel@tonic-gate /* 6637c478bd9Sstevel@tonic-gate * While we should, in most cases, cache the pages for write, we 6647c478bd9Sstevel@tonic-gate * may also want to cache the pages for read as long as they are 6657c478bd9Sstevel@tonic-gate * frequently re-usable. 6667c478bd9Sstevel@tonic-gate * 6677c478bd9Sstevel@tonic-gate * If cache_read_ahead = 1, the pages for read will go to the tail 6687c478bd9Sstevel@tonic-gate * of the cache list when they are released, otherwise go to the head. 6697c478bd9Sstevel@tonic-gate */ 6707c478bd9Sstevel@tonic-gate int cache_read_ahead = 0; 6717c478bd9Sstevel@tonic-gate 672562eee46Srbourbon /* 673562eee46Srbourbon * Freebehind exists so that as we read large files sequentially we 674562eee46Srbourbon * don't consume most of memory with pages from a few files. It takes 675562eee46Srbourbon * longer to re-read from disk multiple small files as it does reading 676562eee46Srbourbon * one large one sequentially. As system memory grows customers need 677562eee46Srbourbon * to retain bigger chunks of files in memory. The advent of the 678562eee46Srbourbon * cachelist opens up of the possibility freeing pages to the head or 679562eee46Srbourbon * tail of the list. 680562eee46Srbourbon * 681562eee46Srbourbon * Not freeing a page is a bet that the page will be read again before 682562eee46Srbourbon * it's segmap slot is needed for something else. If we loose the bet, 683562eee46Srbourbon * it means some other thread is burdened with the page free we did 684562eee46Srbourbon * not do. If we win we save a free and reclaim. 685562eee46Srbourbon * 686562eee46Srbourbon * Freeing it at the tail vs the head of cachelist is a bet that the 687562eee46Srbourbon * page will survive until the next read. It's also saying that this 688562eee46Srbourbon * page is more likely to be re-used than a page freed some time ago 689562eee46Srbourbon * and never reclaimed. 690562eee46Srbourbon * 691562eee46Srbourbon * Freebehind maintains a range of file offset [smallfile1; smallfile2] 692562eee46Srbourbon * 693562eee46Srbourbon * 0 < offset < smallfile1 : pages are not freed. 694562eee46Srbourbon * smallfile1 < offset < smallfile2 : pages freed to tail of cachelist. 695562eee46Srbourbon * smallfile2 < offset : pages freed to head of cachelist. 696562eee46Srbourbon * 697562eee46Srbourbon * The range is computed at most once per second and depends on 698562eee46Srbourbon * freemem and ncpus_online. Both parameters are bounded to be 699562eee46Srbourbon * >= smallfile && >= smallfile64. 700562eee46Srbourbon * 701562eee46Srbourbon * smallfile1 = (free memory / ncpu) / 1000 702562eee46Srbourbon * smallfile2 = (free memory / ncpu) / 10 703562eee46Srbourbon * 704562eee46Srbourbon * A few examples values: 705562eee46Srbourbon * 706562eee46Srbourbon * Free Mem (in Bytes) [smallfile1; smallfile2] [smallfile1; smallfile2] 707562eee46Srbourbon * ncpus_online = 4 ncpus_online = 64 708562eee46Srbourbon * ------------------ ----------------------- ----------------------- 709562eee46Srbourbon * 1G [256K; 25M] [32K; 1.5M] 710562eee46Srbourbon * 10G [2.5M; 250M] [156K; 15M] 711562eee46Srbourbon * 100G [25M; 2.5G] [1.5M; 150M] 712562eee46Srbourbon * 713562eee46Srbourbon */ 714562eee46Srbourbon 715562eee46Srbourbon #define SMALLFILE1_D 1000 716562eee46Srbourbon #define SMALLFILE2_D 10 717562eee46Srbourbon static u_offset_t smallfile1 = 32 * 1024; 718562eee46Srbourbon static u_offset_t smallfile2 = 32 * 1024; 719562eee46Srbourbon static clock_t smallfile_update = 0; /* lbolt value of when to recompute */ 720562eee46Srbourbon uint_t smallfile1_d = SMALLFILE1_D; 721562eee46Srbourbon uint_t smallfile2_d = SMALLFILE2_D; 722562eee46Srbourbon 7237c478bd9Sstevel@tonic-gate /* 7247c478bd9Sstevel@tonic-gate * wrip does the real work of write requests for ufs. 7257c478bd9Sstevel@tonic-gate */ 7267c478bd9Sstevel@tonic-gate int 7277c478bd9Sstevel@tonic-gate wrip(struct inode *ip, struct uio *uio, int ioflag, struct cred *cr) 7287c478bd9Sstevel@tonic-gate { 7297c478bd9Sstevel@tonic-gate rlim64_t limit = uio->uio_llimit; 7307c478bd9Sstevel@tonic-gate u_offset_t off; 7317c478bd9Sstevel@tonic-gate u_offset_t old_i_size; 7327c478bd9Sstevel@tonic-gate struct fs *fs; 7337c478bd9Sstevel@tonic-gate struct vnode *vp; 7347c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp; 7357c478bd9Sstevel@tonic-gate caddr_t base; 7367c478bd9Sstevel@tonic-gate long start_resid = uio->uio_resid; /* save starting resid */ 7377c478bd9Sstevel@tonic-gate long premove_resid; /* resid before uiomove() */ 7387c478bd9Sstevel@tonic-gate uint_t flags; 7397c478bd9Sstevel@tonic-gate int newpage; 7407c478bd9Sstevel@tonic-gate int iupdat_flag, directio_status; 7417c478bd9Sstevel@tonic-gate int n, on, mapon; 7427c478bd9Sstevel@tonic-gate int error, pagecreate; 7437c478bd9Sstevel@tonic-gate int do_dqrwlock; /* drop/reacquire vfs_dqrwlock */ 7447c478bd9Sstevel@tonic-gate int32_t iblocks; 7457c478bd9Sstevel@tonic-gate int new_iblocks; 7467c478bd9Sstevel@tonic-gate 7477c478bd9Sstevel@tonic-gate /* 7487c478bd9Sstevel@tonic-gate * ip->i_size is incremented before the uiomove 7497c478bd9Sstevel@tonic-gate * is done on a write. If the move fails (bad user 7507c478bd9Sstevel@tonic-gate * address) reset ip->i_size. 7517c478bd9Sstevel@tonic-gate * The better way would be to increment ip->i_size 7527c478bd9Sstevel@tonic-gate * only if the uiomove succeeds. 7537c478bd9Sstevel@tonic-gate */ 7547c478bd9Sstevel@tonic-gate int i_size_changed = 0; 7557c478bd9Sstevel@tonic-gate o_mode_t type; 7567c478bd9Sstevel@tonic-gate int i_seq_needed = 0; 7577c478bd9Sstevel@tonic-gate 7587c478bd9Sstevel@tonic-gate vp = ITOV(ip); 7597c478bd9Sstevel@tonic-gate 7607c478bd9Sstevel@tonic-gate /* 7617c478bd9Sstevel@tonic-gate * check for forced unmount - should not happen as 7627c478bd9Sstevel@tonic-gate * the request passed the lockfs checks. 7637c478bd9Sstevel@tonic-gate */ 7647c478bd9Sstevel@tonic-gate if ((ufsvfsp = ip->i_ufsvfs) == NULL) 7657c478bd9Sstevel@tonic-gate return (EIO); 7667c478bd9Sstevel@tonic-gate 7677c478bd9Sstevel@tonic-gate fs = ip->i_fs; 7687c478bd9Sstevel@tonic-gate 7697c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_RWIP_START, 7707c478bd9Sstevel@tonic-gate "ufs_wrip_start:vp %p", vp); 7717c478bd9Sstevel@tonic-gate 7727c478bd9Sstevel@tonic-gate ASSERT(RW_WRITE_HELD(&ip->i_contents)); 7737c478bd9Sstevel@tonic-gate 7747c478bd9Sstevel@tonic-gate /* check for valid filetype */ 7757c478bd9Sstevel@tonic-gate type = ip->i_mode & IFMT; 7767c478bd9Sstevel@tonic-gate if ((type != IFREG) && (type != IFDIR) && (type != IFATTRDIR) && 7777c478bd9Sstevel@tonic-gate (type != IFLNK) && (type != IFSHAD)) { 7787c478bd9Sstevel@tonic-gate return (EIO); 7797c478bd9Sstevel@tonic-gate } 7807c478bd9Sstevel@tonic-gate 7817c478bd9Sstevel@tonic-gate /* 7827c478bd9Sstevel@tonic-gate * the actual limit of UFS file size 7837c478bd9Sstevel@tonic-gate * is UFS_MAXOFFSET_T 7847c478bd9Sstevel@tonic-gate */ 7857c478bd9Sstevel@tonic-gate if (limit == RLIM64_INFINITY || limit > MAXOFFSET_T) 7867c478bd9Sstevel@tonic-gate limit = MAXOFFSET_T; 7877c478bd9Sstevel@tonic-gate 7887c478bd9Sstevel@tonic-gate if (uio->uio_loffset >= limit) { 7897c478bd9Sstevel@tonic-gate proc_t *p = ttoproc(curthread); 7907c478bd9Sstevel@tonic-gate 7917c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_RWIP_END, 7927c478bd9Sstevel@tonic-gate "ufs_wrip_end:vp %p error %d", vp, EINVAL); 7937c478bd9Sstevel@tonic-gate 7947c478bd9Sstevel@tonic-gate mutex_enter(&p->p_lock); 7957c478bd9Sstevel@tonic-gate (void) rctl_action(rctlproc_legacy[RLIMIT_FSIZE], p->p_rctls, 7967c478bd9Sstevel@tonic-gate p, RCA_UNSAFE_SIGINFO); 7977c478bd9Sstevel@tonic-gate mutex_exit(&p->p_lock); 7987c478bd9Sstevel@tonic-gate return (EFBIG); 7997c478bd9Sstevel@tonic-gate } 8007c478bd9Sstevel@tonic-gate 8017c478bd9Sstevel@tonic-gate /* 8027c478bd9Sstevel@tonic-gate * if largefiles are disallowed, the limit is 8037c478bd9Sstevel@tonic-gate * the pre-largefiles value of 2GB 8047c478bd9Sstevel@tonic-gate */ 8057c478bd9Sstevel@tonic-gate if (ufsvfsp->vfs_lfflags & UFS_LARGEFILES) 8067c478bd9Sstevel@tonic-gate limit = MIN(UFS_MAXOFFSET_T, limit); 8077c478bd9Sstevel@tonic-gate else 8087c478bd9Sstevel@tonic-gate limit = MIN(MAXOFF32_T, limit); 8097c478bd9Sstevel@tonic-gate 8107c478bd9Sstevel@tonic-gate if (uio->uio_loffset < (offset_t)0) { 8117c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_RWIP_END, 8127c478bd9Sstevel@tonic-gate "ufs_wrip_end:vp %p error %d", vp, EINVAL); 8137c478bd9Sstevel@tonic-gate return (EINVAL); 8147c478bd9Sstevel@tonic-gate } 8157c478bd9Sstevel@tonic-gate if (uio->uio_resid == 0) { 8167c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_RWIP_END, 8177c478bd9Sstevel@tonic-gate "ufs_wrip_end:vp %p error %d", vp, 0); 8187c478bd9Sstevel@tonic-gate return (0); 8197c478bd9Sstevel@tonic-gate } 8207c478bd9Sstevel@tonic-gate 8217c478bd9Sstevel@tonic-gate if (uio->uio_loffset >= limit) 8227c478bd9Sstevel@tonic-gate return (EFBIG); 8237c478bd9Sstevel@tonic-gate 8247c478bd9Sstevel@tonic-gate ip->i_flag |= INOACC; /* don't update ref time in getpage */ 8257c478bd9Sstevel@tonic-gate 8267c478bd9Sstevel@tonic-gate if (ioflag & (FSYNC|FDSYNC)) { 8277c478bd9Sstevel@tonic-gate ip->i_flag |= ISYNC; 8287c478bd9Sstevel@tonic-gate iupdat_flag = 1; 8297c478bd9Sstevel@tonic-gate } 8307c478bd9Sstevel@tonic-gate /* 8317c478bd9Sstevel@tonic-gate * Try to go direct 8327c478bd9Sstevel@tonic-gate */ 8337c478bd9Sstevel@tonic-gate if (ip->i_flag & IDIRECTIO || ufsvfsp->vfs_forcedirectio) { 8347c478bd9Sstevel@tonic-gate uio->uio_llimit = limit; 8357c478bd9Sstevel@tonic-gate error = ufs_directio_write(ip, uio, ioflag, 0, cr, 8367c478bd9Sstevel@tonic-gate &directio_status); 8377c478bd9Sstevel@tonic-gate /* 8387c478bd9Sstevel@tonic-gate * If ufs_directio wrote to the file or set the flags, 8397c478bd9Sstevel@tonic-gate * we need to update i_seq, but it may be deferred. 8407c478bd9Sstevel@tonic-gate */ 8417c478bd9Sstevel@tonic-gate if (start_resid != uio->uio_resid || 8427c478bd9Sstevel@tonic-gate (ip->i_flag & (ICHG|IUPD))) { 8437c478bd9Sstevel@tonic-gate i_seq_needed = 1; 8447c478bd9Sstevel@tonic-gate ip->i_flag |= ISEQ; 8457c478bd9Sstevel@tonic-gate } 8467c478bd9Sstevel@tonic-gate if (directio_status == DIRECTIO_SUCCESS) 8477c478bd9Sstevel@tonic-gate goto out; 8487c478bd9Sstevel@tonic-gate } 8497c478bd9Sstevel@tonic-gate 8507c478bd9Sstevel@tonic-gate /* 8517c478bd9Sstevel@tonic-gate * Behavior with respect to dropping/reacquiring vfs_dqrwlock: 8527c478bd9Sstevel@tonic-gate * 8537c478bd9Sstevel@tonic-gate * o shadow inodes: vfs_dqrwlock is not held at all 8547c478bd9Sstevel@tonic-gate * o quota updates: vfs_dqrwlock is read or write held 8557c478bd9Sstevel@tonic-gate * o other updates: vfs_dqrwlock is read held 8567c478bd9Sstevel@tonic-gate * 8577c478bd9Sstevel@tonic-gate * The first case is the only one where we do not hold 8587c478bd9Sstevel@tonic-gate * vfs_dqrwlock at all while entering wrip(). 8597c478bd9Sstevel@tonic-gate * We must make sure not to downgrade/drop vfs_dqrwlock if we 8607c478bd9Sstevel@tonic-gate * have it as writer, i.e. if we are updating the quota inode. 8617c478bd9Sstevel@tonic-gate * There is no potential deadlock scenario in this case as 8627c478bd9Sstevel@tonic-gate * ufs_getpage() takes care of this and avoids reacquiring 8637c478bd9Sstevel@tonic-gate * vfs_dqrwlock in that case. 8647c478bd9Sstevel@tonic-gate * 8657c478bd9Sstevel@tonic-gate * This check is done here since the above conditions do not change 8667c478bd9Sstevel@tonic-gate * and we possibly loop below, so save a few cycles. 8677c478bd9Sstevel@tonic-gate */ 8687c478bd9Sstevel@tonic-gate if ((type == IFSHAD) || 8697c478bd9Sstevel@tonic-gate (rw_owner(&ufsvfsp->vfs_dqrwlock) == curthread)) { 8707c478bd9Sstevel@tonic-gate do_dqrwlock = 0; 8717c478bd9Sstevel@tonic-gate } else { 8727c478bd9Sstevel@tonic-gate do_dqrwlock = 1; 8737c478bd9Sstevel@tonic-gate } 8747c478bd9Sstevel@tonic-gate 8757c478bd9Sstevel@tonic-gate /* 8767c478bd9Sstevel@tonic-gate * Large Files: We cast MAXBMASK to offset_t 8777c478bd9Sstevel@tonic-gate * inorder to mask out the higher bits. Since offset_t 8787c478bd9Sstevel@tonic-gate * is a signed value, the high order bit set in MAXBMASK 8797c478bd9Sstevel@tonic-gate * value makes it do the right thing by having all bits 1 8807c478bd9Sstevel@tonic-gate * in the higher word. May be removed for _SOLARIS64_. 8817c478bd9Sstevel@tonic-gate */ 8827c478bd9Sstevel@tonic-gate 8837c478bd9Sstevel@tonic-gate fs = ip->i_fs; 8847c478bd9Sstevel@tonic-gate do { 8857c478bd9Sstevel@tonic-gate u_offset_t uoff = uio->uio_loffset; 8867c478bd9Sstevel@tonic-gate off = uoff & (offset_t)MAXBMASK; 8877c478bd9Sstevel@tonic-gate mapon = (int)(uoff & (offset_t)MAXBOFFSET); 8887c478bd9Sstevel@tonic-gate on = (int)blkoff(fs, uoff); 8897c478bd9Sstevel@tonic-gate n = (int)MIN(fs->fs_bsize - on, uio->uio_resid); 8907c478bd9Sstevel@tonic-gate new_iblocks = 1; 8917c478bd9Sstevel@tonic-gate 8927c478bd9Sstevel@tonic-gate if (type == IFREG && uoff + n >= limit) { 8937c478bd9Sstevel@tonic-gate if (uoff >= limit) { 8947c478bd9Sstevel@tonic-gate error = EFBIG; 8957c478bd9Sstevel@tonic-gate goto out; 8967c478bd9Sstevel@tonic-gate } 8977c478bd9Sstevel@tonic-gate /* 8987c478bd9Sstevel@tonic-gate * since uoff + n >= limit, 8997c478bd9Sstevel@tonic-gate * therefore n >= limit - uoff, and n is an int 9007c478bd9Sstevel@tonic-gate * so it is safe to cast it to an int 9017c478bd9Sstevel@tonic-gate */ 9027c478bd9Sstevel@tonic-gate n = (int)(limit - (rlim64_t)uoff); 9037c478bd9Sstevel@tonic-gate } 9047c478bd9Sstevel@tonic-gate if (uoff + n > ip->i_size) { 9057c478bd9Sstevel@tonic-gate /* 9067c478bd9Sstevel@tonic-gate * We are extending the length of the file. 9077c478bd9Sstevel@tonic-gate * bmap is used so that we are sure that 9087c478bd9Sstevel@tonic-gate * if we need to allocate new blocks, that it 9097c478bd9Sstevel@tonic-gate * is done here before we up the file size. 9107c478bd9Sstevel@tonic-gate */ 9117c478bd9Sstevel@tonic-gate error = bmap_write(ip, uoff, (int)(on + n), 912303bf60bSsdebnath mapon == 0, NULL, cr); 9137c478bd9Sstevel@tonic-gate /* 9147c478bd9Sstevel@tonic-gate * bmap_write never drops i_contents so if 9157c478bd9Sstevel@tonic-gate * the flags are set it changed the file. 9167c478bd9Sstevel@tonic-gate */ 9177c478bd9Sstevel@tonic-gate if (ip->i_flag & (ICHG|IUPD)) { 9187c478bd9Sstevel@tonic-gate i_seq_needed = 1; 9197c478bd9Sstevel@tonic-gate ip->i_flag |= ISEQ; 9207c478bd9Sstevel@tonic-gate } 9217c478bd9Sstevel@tonic-gate if (error) 9227c478bd9Sstevel@tonic-gate break; 9237c478bd9Sstevel@tonic-gate /* 9247c478bd9Sstevel@tonic-gate * There is a window of vulnerability here. 9257c478bd9Sstevel@tonic-gate * The sequence of operations: allocate file 9267c478bd9Sstevel@tonic-gate * system blocks, uiomove the data into pages, 9277c478bd9Sstevel@tonic-gate * and then update the size of the file in the 9287c478bd9Sstevel@tonic-gate * inode, must happen atomically. However, due 9297c478bd9Sstevel@tonic-gate * to current locking constraints, this can not 9307c478bd9Sstevel@tonic-gate * be done. 9317c478bd9Sstevel@tonic-gate */ 9327c478bd9Sstevel@tonic-gate ASSERT(ip->i_writer == NULL); 9337c478bd9Sstevel@tonic-gate ip->i_writer = curthread; 9347c478bd9Sstevel@tonic-gate i_size_changed = 1; 9357c478bd9Sstevel@tonic-gate /* 9367c478bd9Sstevel@tonic-gate * If we are writing from the beginning of 9377c478bd9Sstevel@tonic-gate * the mapping, we can just create the 9387c478bd9Sstevel@tonic-gate * pages without having to read them. 9397c478bd9Sstevel@tonic-gate */ 9407c478bd9Sstevel@tonic-gate pagecreate = (mapon == 0); 9417c478bd9Sstevel@tonic-gate } else if (n == MAXBSIZE) { 9427c478bd9Sstevel@tonic-gate /* 9437c478bd9Sstevel@tonic-gate * Going to do a whole mappings worth, 9447c478bd9Sstevel@tonic-gate * so we can just create the pages w/o 9457c478bd9Sstevel@tonic-gate * having to read them in. But before 9467c478bd9Sstevel@tonic-gate * we do that, we need to make sure any 9477c478bd9Sstevel@tonic-gate * needed blocks are allocated first. 9487c478bd9Sstevel@tonic-gate */ 9497c478bd9Sstevel@tonic-gate iblocks = ip->i_blocks; 950303bf60bSsdebnath error = bmap_write(ip, uoff, (int)(on + n), 951303bf60bSsdebnath BI_ALLOC_ONLY, NULL, cr); 9527c478bd9Sstevel@tonic-gate /* 9537c478bd9Sstevel@tonic-gate * bmap_write never drops i_contents so if 9547c478bd9Sstevel@tonic-gate * the flags are set it changed the file. 9557c478bd9Sstevel@tonic-gate */ 9567c478bd9Sstevel@tonic-gate if (ip->i_flag & (ICHG|IUPD)) { 9577c478bd9Sstevel@tonic-gate i_seq_needed = 1; 9587c478bd9Sstevel@tonic-gate ip->i_flag |= ISEQ; 9597c478bd9Sstevel@tonic-gate } 9607c478bd9Sstevel@tonic-gate if (error) 9617c478bd9Sstevel@tonic-gate break; 9627c478bd9Sstevel@tonic-gate pagecreate = 1; 9637c478bd9Sstevel@tonic-gate /* 9647c478bd9Sstevel@tonic-gate * check if the new created page needed the 9657c478bd9Sstevel@tonic-gate * allocation of new disk blocks. 9667c478bd9Sstevel@tonic-gate */ 9677c478bd9Sstevel@tonic-gate if (iblocks == ip->i_blocks) 9687c478bd9Sstevel@tonic-gate new_iblocks = 0; /* no new blocks allocated */ 9697c478bd9Sstevel@tonic-gate } else { 9707c478bd9Sstevel@tonic-gate pagecreate = 0; 9717c478bd9Sstevel@tonic-gate /* 9727c478bd9Sstevel@tonic-gate * In sync mode flush the indirect blocks which 9737c478bd9Sstevel@tonic-gate * may have been allocated and not written on 9747c478bd9Sstevel@tonic-gate * disk. In above cases bmap_write will allocate 9757c478bd9Sstevel@tonic-gate * in sync mode. 9767c478bd9Sstevel@tonic-gate */ 9777c478bd9Sstevel@tonic-gate if (ioflag & (FSYNC|FDSYNC)) { 9787c478bd9Sstevel@tonic-gate error = ufs_indirblk_sync(ip, uoff); 9797c478bd9Sstevel@tonic-gate if (error) 9807c478bd9Sstevel@tonic-gate break; 9817c478bd9Sstevel@tonic-gate } 9827c478bd9Sstevel@tonic-gate } 9837c478bd9Sstevel@tonic-gate 9847c478bd9Sstevel@tonic-gate /* 9857c478bd9Sstevel@tonic-gate * At this point we can enter ufs_getpage() in one 9867c478bd9Sstevel@tonic-gate * of two ways: 9877c478bd9Sstevel@tonic-gate * 1) segmap_getmapflt() calls ufs_getpage() when the 9887c478bd9Sstevel@tonic-gate * forcefault parameter is true (pagecreate == 0) 9897c478bd9Sstevel@tonic-gate * 2) uiomove() causes a page fault. 9907c478bd9Sstevel@tonic-gate * 9917c478bd9Sstevel@tonic-gate * We have to drop the contents lock to prevent the VM 9927c478bd9Sstevel@tonic-gate * system from trying to reaquire it in ufs_getpage() 9937c478bd9Sstevel@tonic-gate * should the uiomove cause a pagefault. 9947c478bd9Sstevel@tonic-gate * 9957c478bd9Sstevel@tonic-gate * We have to drop the reader vfs_dqrwlock here as well. 9967c478bd9Sstevel@tonic-gate */ 9977c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 9987c478bd9Sstevel@tonic-gate if (do_dqrwlock) { 9997c478bd9Sstevel@tonic-gate ASSERT(RW_LOCK_HELD(&ufsvfsp->vfs_dqrwlock)); 10007c478bd9Sstevel@tonic-gate ASSERT(!(RW_WRITE_HELD(&ufsvfsp->vfs_dqrwlock))); 10017c478bd9Sstevel@tonic-gate rw_exit(&ufsvfsp->vfs_dqrwlock); 10027c478bd9Sstevel@tonic-gate } 10037c478bd9Sstevel@tonic-gate 1004a5652762Spraks newpage = 0; 1005a5652762Spraks premove_resid = uio->uio_resid; 1006a5652762Spraks if (vpm_enable) { 1007a5652762Spraks /* 1008a5652762Spraks * Copy data. If new pages are created, part of 1009a5652762Spraks * the page that is not written will be initizliazed 1010a5652762Spraks * with zeros. 1011a5652762Spraks */ 1012a5652762Spraks error = vpm_data_copy(vp, (off + mapon), (uint_t)n, 1013a5652762Spraks uio, !pagecreate, &newpage, 0, S_WRITE); 1014a5652762Spraks } else { 1015a5652762Spraks 1016a5652762Spraks base = segmap_getmapflt(segkmap, vp, (off + mapon), 10177c478bd9Sstevel@tonic-gate (uint_t)n, !pagecreate, S_WRITE); 10187c478bd9Sstevel@tonic-gate 1019a5652762Spraks /* 1020a5652762Spraks * segmap_pagecreate() returns 1 if it calls 1021a5652762Spraks * page_create_va() to allocate any pages. 1022a5652762Spraks */ 10237c478bd9Sstevel@tonic-gate 1024a5652762Spraks if (pagecreate) 1025a5652762Spraks newpage = segmap_pagecreate(segkmap, base, 1026a5652762Spraks (size_t)n, 0); 10277c478bd9Sstevel@tonic-gate 1028a5652762Spraks error = uiomove(base + mapon, (long)n, UIO_WRITE, uio); 1029a5652762Spraks } 10307c478bd9Sstevel@tonic-gate 10317c478bd9Sstevel@tonic-gate /* 10327c478bd9Sstevel@tonic-gate * If "newpage" is set, then a new page was created and it 10337c478bd9Sstevel@tonic-gate * does not contain valid data, so it needs to be initialized 10347c478bd9Sstevel@tonic-gate * at this point. 10357c478bd9Sstevel@tonic-gate * Otherwise the page contains old data, which was overwritten 10367c478bd9Sstevel@tonic-gate * partially or as a whole in uiomove. 10377c478bd9Sstevel@tonic-gate * If there is only one iovec structure within uio, then 10387c478bd9Sstevel@tonic-gate * on error uiomove will not be able to update uio->uio_loffset 10397c478bd9Sstevel@tonic-gate * and we would zero the whole page here! 10407c478bd9Sstevel@tonic-gate * 10417c478bd9Sstevel@tonic-gate * If uiomove fails because of an error, the old valid data 10427c478bd9Sstevel@tonic-gate * is kept instead of filling the rest of the page with zero's. 10437c478bd9Sstevel@tonic-gate */ 1044a5652762Spraks if (!vpm_enable && newpage && 10457c478bd9Sstevel@tonic-gate uio->uio_loffset < roundup(off + mapon + n, PAGESIZE)) { 10467c478bd9Sstevel@tonic-gate /* 10477c478bd9Sstevel@tonic-gate * We created pages w/o initializing them completely, 10487c478bd9Sstevel@tonic-gate * thus we need to zero the part that wasn't set up. 10497c478bd9Sstevel@tonic-gate * This happens on most EOF write cases and if 10507c478bd9Sstevel@tonic-gate * we had some sort of error during the uiomove. 10517c478bd9Sstevel@tonic-gate */ 10527c478bd9Sstevel@tonic-gate int nzero, nmoved; 10537c478bd9Sstevel@tonic-gate 10547c478bd9Sstevel@tonic-gate nmoved = (int)(uio->uio_loffset - (off + mapon)); 10557c478bd9Sstevel@tonic-gate ASSERT(nmoved >= 0 && nmoved <= n); 10567c478bd9Sstevel@tonic-gate nzero = roundup(on + n, PAGESIZE) - nmoved; 10577c478bd9Sstevel@tonic-gate ASSERT(nzero > 0 && mapon + nmoved + nzero <= MAXBSIZE); 10587c478bd9Sstevel@tonic-gate (void) kzero(base + mapon + nmoved, (uint_t)nzero); 10597c478bd9Sstevel@tonic-gate } 10607c478bd9Sstevel@tonic-gate 10617c478bd9Sstevel@tonic-gate /* 10627c478bd9Sstevel@tonic-gate * Unlock the pages allocated by page_create_va() 10637c478bd9Sstevel@tonic-gate * in segmap_pagecreate() 10647c478bd9Sstevel@tonic-gate */ 1065a5652762Spraks if (!vpm_enable && newpage) 10667c478bd9Sstevel@tonic-gate segmap_pageunlock(segkmap, base, (size_t)n, S_WRITE); 10677c478bd9Sstevel@tonic-gate 10687c478bd9Sstevel@tonic-gate /* 10697c478bd9Sstevel@tonic-gate * If the size of the file changed, then update the 10707c478bd9Sstevel@tonic-gate * size field in the inode now. This can't be done 10717c478bd9Sstevel@tonic-gate * before the call to segmap_pageunlock or there is 10727c478bd9Sstevel@tonic-gate * a potential deadlock with callers to ufs_putpage(). 10737c478bd9Sstevel@tonic-gate * They will be holding i_contents and trying to lock 10747c478bd9Sstevel@tonic-gate * a page, while this thread is holding a page locked 10757c478bd9Sstevel@tonic-gate * and trying to acquire i_contents. 10767c478bd9Sstevel@tonic-gate */ 10777c478bd9Sstevel@tonic-gate if (i_size_changed) { 10787c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_WRITER); 10797c478bd9Sstevel@tonic-gate old_i_size = ip->i_size; 10807c478bd9Sstevel@tonic-gate UFS_SET_ISIZE(uoff + n, ip); 10817c478bd9Sstevel@tonic-gate TRANS_INODE(ufsvfsp, ip); 10827c478bd9Sstevel@tonic-gate /* 10837c478bd9Sstevel@tonic-gate * file has grown larger than 2GB. Set flag 10847c478bd9Sstevel@tonic-gate * in superblock to indicate this, if it 10857c478bd9Sstevel@tonic-gate * is not already set. 10867c478bd9Sstevel@tonic-gate */ 10877c478bd9Sstevel@tonic-gate if ((ip->i_size > MAXOFF32_T) && 10887c478bd9Sstevel@tonic-gate !(fs->fs_flags & FSLARGEFILES)) { 10897c478bd9Sstevel@tonic-gate ASSERT(ufsvfsp->vfs_lfflags & UFS_LARGEFILES); 10907c478bd9Sstevel@tonic-gate mutex_enter(&ufsvfsp->vfs_lock); 10917c478bd9Sstevel@tonic-gate fs->fs_flags |= FSLARGEFILES; 10927c478bd9Sstevel@tonic-gate ufs_sbwrite(ufsvfsp); 10937c478bd9Sstevel@tonic-gate mutex_exit(&ufsvfsp->vfs_lock); 10947c478bd9Sstevel@tonic-gate } 10957c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 10967c478bd9Sstevel@tonic-gate ip->i_writer = NULL; 10977c478bd9Sstevel@tonic-gate cv_broadcast(&ip->i_wrcv); 10987c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 10997c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 11007c478bd9Sstevel@tonic-gate } 11017c478bd9Sstevel@tonic-gate 11027c478bd9Sstevel@tonic-gate if (error) { 11037c478bd9Sstevel@tonic-gate /* 11047c478bd9Sstevel@tonic-gate * If we failed on a write, we may have already 11057c478bd9Sstevel@tonic-gate * allocated file blocks as well as pages. It's 11067c478bd9Sstevel@tonic-gate * hard to undo the block allocation, but we must 11077c478bd9Sstevel@tonic-gate * be sure to invalidate any pages that may have 11087c478bd9Sstevel@tonic-gate * been allocated. 11097c478bd9Sstevel@tonic-gate * 11107c478bd9Sstevel@tonic-gate * If the page was created without initialization 11117c478bd9Sstevel@tonic-gate * then we must check if it should be possible 11127c478bd9Sstevel@tonic-gate * to destroy the new page and to keep the old data 11137c478bd9Sstevel@tonic-gate * on the disk. 11147c478bd9Sstevel@tonic-gate * 11157c478bd9Sstevel@tonic-gate * It is possible to destroy the page without 11167c478bd9Sstevel@tonic-gate * having to write back its contents only when 11177c478bd9Sstevel@tonic-gate * - the size of the file keeps unchanged 11187c478bd9Sstevel@tonic-gate * - bmap_write() did not allocate new disk blocks 11197c478bd9Sstevel@tonic-gate * it is possible to create big files using "seek" and 11207c478bd9Sstevel@tonic-gate * write to the end of the file. A "write" to a 11217c478bd9Sstevel@tonic-gate * position before the end of the file would not 11227c478bd9Sstevel@tonic-gate * change the size of the file but it would allocate 11237c478bd9Sstevel@tonic-gate * new disk blocks. 11247c478bd9Sstevel@tonic-gate * - uiomove intended to overwrite the whole page. 11257c478bd9Sstevel@tonic-gate * - a new page was created (newpage == 1). 11267c478bd9Sstevel@tonic-gate */ 11277c478bd9Sstevel@tonic-gate 11287c478bd9Sstevel@tonic-gate if (i_size_changed == 0 && new_iblocks == 0 && 11297c478bd9Sstevel@tonic-gate newpage) { 11307c478bd9Sstevel@tonic-gate 11317c478bd9Sstevel@tonic-gate /* unwind what uiomove eventually last did */ 11327c478bd9Sstevel@tonic-gate uio->uio_resid = premove_resid; 11337c478bd9Sstevel@tonic-gate 11347c478bd9Sstevel@tonic-gate /* 11357c478bd9Sstevel@tonic-gate * destroy the page, do not write ambiguous 11367c478bd9Sstevel@tonic-gate * data to the disk. 11377c478bd9Sstevel@tonic-gate */ 11387c478bd9Sstevel@tonic-gate flags = SM_DESTROY; 11397c478bd9Sstevel@tonic-gate } else { 11407c478bd9Sstevel@tonic-gate /* 11417c478bd9Sstevel@tonic-gate * write the page back to the disk, if dirty, 11427c478bd9Sstevel@tonic-gate * and remove the page from the cache. 11437c478bd9Sstevel@tonic-gate */ 11447c478bd9Sstevel@tonic-gate flags = SM_INVAL; 11457c478bd9Sstevel@tonic-gate } 1146a5652762Spraks 1147a5652762Spraks if (vpm_enable) { 1148a5652762Spraks /* 1149a5652762Spraks * Flush pages. 1150a5652762Spraks */ 1151a5652762Spraks (void) vpm_sync_pages(vp, off, n, flags); 1152a5652762Spraks } else { 1153a5652762Spraks (void) segmap_release(segkmap, base, flags); 1154a5652762Spraks } 11557c478bd9Sstevel@tonic-gate } else { 11567c478bd9Sstevel@tonic-gate flags = 0; 11577c478bd9Sstevel@tonic-gate /* 11587c478bd9Sstevel@tonic-gate * Force write back for synchronous write cases. 11597c478bd9Sstevel@tonic-gate */ 11607c478bd9Sstevel@tonic-gate if ((ioflag & (FSYNC|FDSYNC)) || type == IFDIR) { 11617c478bd9Sstevel@tonic-gate /* 11627c478bd9Sstevel@tonic-gate * If the sticky bit is set but the 11637c478bd9Sstevel@tonic-gate * execute bit is not set, we do a 11647c478bd9Sstevel@tonic-gate * synchronous write back and free 11657c478bd9Sstevel@tonic-gate * the page when done. We set up swap 11667c478bd9Sstevel@tonic-gate * files to be handled this way to 11677c478bd9Sstevel@tonic-gate * prevent servers from keeping around 11687c478bd9Sstevel@tonic-gate * the client's swap pages too long. 11697c478bd9Sstevel@tonic-gate * XXX - there ought to be a better way. 11707c478bd9Sstevel@tonic-gate */ 11717c478bd9Sstevel@tonic-gate if (IS_SWAPVP(vp)) { 11727c478bd9Sstevel@tonic-gate flags = SM_WRITE | SM_FREE | 11737c478bd9Sstevel@tonic-gate SM_DONTNEED; 11747c478bd9Sstevel@tonic-gate iupdat_flag = 0; 11757c478bd9Sstevel@tonic-gate } else { 11767c478bd9Sstevel@tonic-gate flags = SM_WRITE; 11777c478bd9Sstevel@tonic-gate } 11787c478bd9Sstevel@tonic-gate } else if (n + on == MAXBSIZE || IS_SWAPVP(vp)) { 11797c478bd9Sstevel@tonic-gate /* 11807c478bd9Sstevel@tonic-gate * Have written a whole block. 11817c478bd9Sstevel@tonic-gate * Start an asynchronous write and 11827c478bd9Sstevel@tonic-gate * mark the buffer to indicate that 11837c478bd9Sstevel@tonic-gate * it won't be needed again soon. 11847c478bd9Sstevel@tonic-gate */ 11857c478bd9Sstevel@tonic-gate flags = SM_WRITE | SM_ASYNC | SM_DONTNEED; 11867c478bd9Sstevel@tonic-gate } 1187a5652762Spraks if (vpm_enable) { 1188a5652762Spraks /* 1189a5652762Spraks * Flush pages. 1190a5652762Spraks */ 1191a5652762Spraks (void) vpm_sync_pages(vp, off, n, flags); 1192a5652762Spraks } else { 1193a5652762Spraks (void) segmap_release(segkmap, base, flags); 1194a5652762Spraks } 11957c478bd9Sstevel@tonic-gate /* 11967c478bd9Sstevel@tonic-gate * If the operation failed and is synchronous, 11977c478bd9Sstevel@tonic-gate * then we need to unwind what uiomove() last 11987c478bd9Sstevel@tonic-gate * did so we can potentially return an error to 11997c478bd9Sstevel@tonic-gate * the caller. If this write operation was 12007c478bd9Sstevel@tonic-gate * done in two pieces and the first succeeded, 12017c478bd9Sstevel@tonic-gate * then we won't return an error for the second 12027c478bd9Sstevel@tonic-gate * piece that failed. However, we only want to 12037c478bd9Sstevel@tonic-gate * return a resid value that reflects what was 12047c478bd9Sstevel@tonic-gate * really done. 12057c478bd9Sstevel@tonic-gate * 12067c478bd9Sstevel@tonic-gate * Failures for non-synchronous operations can 12077c478bd9Sstevel@tonic-gate * be ignored since the page subsystem will 12087c478bd9Sstevel@tonic-gate * retry the operation until it succeeds or the 12097c478bd9Sstevel@tonic-gate * file system is unmounted. 12107c478bd9Sstevel@tonic-gate */ 12117c478bd9Sstevel@tonic-gate if (error) { 12127c478bd9Sstevel@tonic-gate if ((ioflag & (FSYNC | FDSYNC)) || 12137c478bd9Sstevel@tonic-gate type == IFDIR) { 12147c478bd9Sstevel@tonic-gate uio->uio_resid = premove_resid; 12157c478bd9Sstevel@tonic-gate } else { 12167c478bd9Sstevel@tonic-gate error = 0; 12177c478bd9Sstevel@tonic-gate } 12187c478bd9Sstevel@tonic-gate } 12197c478bd9Sstevel@tonic-gate } 12207c478bd9Sstevel@tonic-gate 12217c478bd9Sstevel@tonic-gate /* 12227c478bd9Sstevel@tonic-gate * Re-acquire contents lock. 12237c478bd9Sstevel@tonic-gate * If it was dropped, reacquire reader vfs_dqrwlock as well. 12247c478bd9Sstevel@tonic-gate */ 12257c478bd9Sstevel@tonic-gate if (do_dqrwlock) 12267c478bd9Sstevel@tonic-gate rw_enter(&ufsvfsp->vfs_dqrwlock, RW_READER); 12277c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_WRITER); 12287c478bd9Sstevel@tonic-gate 12297c478bd9Sstevel@tonic-gate /* 12307c478bd9Sstevel@tonic-gate * If the uiomove() failed or if a synchronous 12317c478bd9Sstevel@tonic-gate * page push failed, fix up i_size. 12327c478bd9Sstevel@tonic-gate */ 12337c478bd9Sstevel@tonic-gate if (error) { 12347c478bd9Sstevel@tonic-gate if (i_size_changed) { 12357c478bd9Sstevel@tonic-gate /* 12367c478bd9Sstevel@tonic-gate * The uiomove failed, and we 12377c478bd9Sstevel@tonic-gate * allocated blocks,so get rid 12387c478bd9Sstevel@tonic-gate * of them. 12397c478bd9Sstevel@tonic-gate */ 12407c478bd9Sstevel@tonic-gate (void) ufs_itrunc(ip, old_i_size, 0, cr); 12417c478bd9Sstevel@tonic-gate } 12427c478bd9Sstevel@tonic-gate } else { 12437c478bd9Sstevel@tonic-gate /* 12447c478bd9Sstevel@tonic-gate * XXX - Can this be out of the loop? 12457c478bd9Sstevel@tonic-gate */ 12467c478bd9Sstevel@tonic-gate ip->i_flag |= IUPD | ICHG; 12477c478bd9Sstevel@tonic-gate /* 12487c478bd9Sstevel@tonic-gate * Only do one increase of i_seq for multiple 12497c478bd9Sstevel@tonic-gate * pieces. Because we drop locks, record 12507c478bd9Sstevel@tonic-gate * the fact that we changed the timestamp and 12517c478bd9Sstevel@tonic-gate * are deferring the increase in case another thread 12527c478bd9Sstevel@tonic-gate * pushes our timestamp update. 12537c478bd9Sstevel@tonic-gate */ 12547c478bd9Sstevel@tonic-gate i_seq_needed = 1; 12557c478bd9Sstevel@tonic-gate ip->i_flag |= ISEQ; 12567c478bd9Sstevel@tonic-gate if (i_size_changed) 12577c478bd9Sstevel@tonic-gate ip->i_flag |= IATTCHG; 12587c478bd9Sstevel@tonic-gate if ((ip->i_mode & (IEXEC | (IEXEC >> 3) | 12597c478bd9Sstevel@tonic-gate (IEXEC >> 6))) != 0 && 12607c478bd9Sstevel@tonic-gate (ip->i_mode & (ISUID | ISGID)) != 0 && 12617c478bd9Sstevel@tonic-gate secpolicy_vnode_setid_retain(cr, 12627c478bd9Sstevel@tonic-gate (ip->i_mode & ISUID) != 0 && ip->i_uid == 0) != 0) { 12637c478bd9Sstevel@tonic-gate /* 12647c478bd9Sstevel@tonic-gate * Clear Set-UID & Set-GID bits on 12657c478bd9Sstevel@tonic-gate * successful write if not privileged 12667c478bd9Sstevel@tonic-gate * and at least one of the execute bits 12677c478bd9Sstevel@tonic-gate * is set. If we always clear Set-GID, 12687c478bd9Sstevel@tonic-gate * mandatory file and record locking is 12697c478bd9Sstevel@tonic-gate * unuseable. 12707c478bd9Sstevel@tonic-gate */ 12717c478bd9Sstevel@tonic-gate ip->i_mode &= ~(ISUID | ISGID); 12727c478bd9Sstevel@tonic-gate } 12737c478bd9Sstevel@tonic-gate } 1274f90bab20Sswilcox /* 1275f90bab20Sswilcox * In the case the FDSYNC flag is set and this is a 1276f90bab20Sswilcox * "rewrite" we won't log a delta. 1277f90bab20Sswilcox * The FSYNC flag overrides all cases. 1278f90bab20Sswilcox */ 1279f90bab20Sswilcox if (!ufs_check_rewrite(ip, uio, ioflag) || !(ioflag & FDSYNC)) { 1280f90bab20Sswilcox TRANS_INODE(ufsvfsp, ip); 1281f90bab20Sswilcox } 12827c478bd9Sstevel@tonic-gate } while (error == 0 && uio->uio_resid > 0 && n != 0); 12837c478bd9Sstevel@tonic-gate 12847c478bd9Sstevel@tonic-gate out: 12857c478bd9Sstevel@tonic-gate /* 12867c478bd9Sstevel@tonic-gate * Make sure i_seq is increased at least once per write 12877c478bd9Sstevel@tonic-gate */ 12887c478bd9Sstevel@tonic-gate if (i_seq_needed) { 12897c478bd9Sstevel@tonic-gate ip->i_seq++; 12907c478bd9Sstevel@tonic-gate ip->i_flag &= ~ISEQ; /* no longer deferred */ 12917c478bd9Sstevel@tonic-gate } 12927c478bd9Sstevel@tonic-gate 12937c478bd9Sstevel@tonic-gate /* 12947c478bd9Sstevel@tonic-gate * Inode is updated according to this table - 12957c478bd9Sstevel@tonic-gate * 12967c478bd9Sstevel@tonic-gate * FSYNC FDSYNC(posix.4) 12977c478bd9Sstevel@tonic-gate * -------------------------- 12987c478bd9Sstevel@tonic-gate * always@ IATTCHG|IBDWRITE 12997c478bd9Sstevel@tonic-gate * 13007c478bd9Sstevel@tonic-gate * @ - If we are doing synchronous write the only time we should 13017c478bd9Sstevel@tonic-gate * not be sync'ing the ip here is if we have the stickyhack 13027c478bd9Sstevel@tonic-gate * activated, the file is marked with the sticky bit and 13037c478bd9Sstevel@tonic-gate * no exec bit, the file length has not been changed and 13047c478bd9Sstevel@tonic-gate * no new blocks have been allocated during this write. 13057c478bd9Sstevel@tonic-gate */ 13067c478bd9Sstevel@tonic-gate 13077c478bd9Sstevel@tonic-gate if ((ip->i_flag & ISYNC) != 0) { 13087c478bd9Sstevel@tonic-gate /* 13097c478bd9Sstevel@tonic-gate * we have eliminated nosync 13107c478bd9Sstevel@tonic-gate */ 13117c478bd9Sstevel@tonic-gate if ((ip->i_flag & (IATTCHG|IBDWRITE)) || 13127c478bd9Sstevel@tonic-gate ((ioflag & FSYNC) && iupdat_flag)) { 13137c478bd9Sstevel@tonic-gate ufs_iupdat(ip, 1); 13147c478bd9Sstevel@tonic-gate } 13157c478bd9Sstevel@tonic-gate } 13167c478bd9Sstevel@tonic-gate 13177c478bd9Sstevel@tonic-gate /* 13187c478bd9Sstevel@tonic-gate * If we've already done a partial-write, terminate 13197c478bd9Sstevel@tonic-gate * the write but return no error unless the error is ENOSPC 13207c478bd9Sstevel@tonic-gate * because the caller can detect this and free resources and 13217c478bd9Sstevel@tonic-gate * try again. 13227c478bd9Sstevel@tonic-gate */ 13237c478bd9Sstevel@tonic-gate if ((start_resid != uio->uio_resid) && (error != ENOSPC)) 13247c478bd9Sstevel@tonic-gate error = 0; 13257c478bd9Sstevel@tonic-gate 13267c478bd9Sstevel@tonic-gate ip->i_flag &= ~(INOACC | ISYNC); 13277c478bd9Sstevel@tonic-gate ITIMES_NOLOCK(ip); 13287c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_RWIP_END, 13297c478bd9Sstevel@tonic-gate "ufs_wrip_end:vp %p error %d", vp, error); 13307c478bd9Sstevel@tonic-gate return (error); 13317c478bd9Sstevel@tonic-gate } 13327c478bd9Sstevel@tonic-gate 13337c478bd9Sstevel@tonic-gate /* 13347c478bd9Sstevel@tonic-gate * rdip does the real work of read requests for ufs. 13357c478bd9Sstevel@tonic-gate */ 13367c478bd9Sstevel@tonic-gate int 13377c478bd9Sstevel@tonic-gate rdip(struct inode *ip, struct uio *uio, int ioflag, cred_t *cr) 13387c478bd9Sstevel@tonic-gate { 13397c478bd9Sstevel@tonic-gate u_offset_t off; 13407c478bd9Sstevel@tonic-gate caddr_t base; 13417c478bd9Sstevel@tonic-gate struct fs *fs; 13427c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp; 13437c478bd9Sstevel@tonic-gate struct vnode *vp; 13447c478bd9Sstevel@tonic-gate long oresid = uio->uio_resid; 13457c478bd9Sstevel@tonic-gate u_offset_t n, on, mapon; 13467c478bd9Sstevel@tonic-gate int error = 0; 13477c478bd9Sstevel@tonic-gate int doupdate = 1; 1348562eee46Srbourbon uint_t flags; 13497c478bd9Sstevel@tonic-gate int dofree, directio_status; 13507c478bd9Sstevel@tonic-gate krw_t rwtype; 13517c478bd9Sstevel@tonic-gate o_mode_t type; 13527c478bd9Sstevel@tonic-gate 13537c478bd9Sstevel@tonic-gate vp = ITOV(ip); 13547c478bd9Sstevel@tonic-gate 13557c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_RWIP_START, 13567c478bd9Sstevel@tonic-gate "ufs_rdip_start:vp %p", vp); 13577c478bd9Sstevel@tonic-gate 13587c478bd9Sstevel@tonic-gate ASSERT(RW_LOCK_HELD(&ip->i_contents)); 13597c478bd9Sstevel@tonic-gate 13607c478bd9Sstevel@tonic-gate ufsvfsp = ip->i_ufsvfs; 13617c478bd9Sstevel@tonic-gate 13627c478bd9Sstevel@tonic-gate if (ufsvfsp == NULL) 13637c478bd9Sstevel@tonic-gate return (EIO); 13647c478bd9Sstevel@tonic-gate 13657c478bd9Sstevel@tonic-gate fs = ufsvfsp->vfs_fs; 13667c478bd9Sstevel@tonic-gate 13677c478bd9Sstevel@tonic-gate /* check for valid filetype */ 13687c478bd9Sstevel@tonic-gate type = ip->i_mode & IFMT; 13697c478bd9Sstevel@tonic-gate if ((type != IFREG) && (type != IFDIR) && (type != IFATTRDIR) && 13707c478bd9Sstevel@tonic-gate (type != IFLNK) && (type != IFSHAD)) { 13717c478bd9Sstevel@tonic-gate return (EIO); 13727c478bd9Sstevel@tonic-gate } 13737c478bd9Sstevel@tonic-gate 13747c478bd9Sstevel@tonic-gate if (uio->uio_loffset > UFS_MAXOFFSET_T) { 13757c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_RWIP_END, 13767c478bd9Sstevel@tonic-gate "ufs_rdip_end:vp %p error %d", vp, EINVAL); 13777c478bd9Sstevel@tonic-gate error = 0; 13787c478bd9Sstevel@tonic-gate goto out; 13797c478bd9Sstevel@tonic-gate } 13807c478bd9Sstevel@tonic-gate if (uio->uio_loffset < (offset_t)0) { 13817c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_RWIP_END, 13827c478bd9Sstevel@tonic-gate "ufs_rdip_end:vp %p error %d", vp, EINVAL); 13837c478bd9Sstevel@tonic-gate return (EINVAL); 13847c478bd9Sstevel@tonic-gate } 13857c478bd9Sstevel@tonic-gate if (uio->uio_resid == 0) { 13867c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_RWIP_END, 13877c478bd9Sstevel@tonic-gate "ufs_rdip_end:vp %p error %d", vp, 0); 13887c478bd9Sstevel@tonic-gate return (0); 13897c478bd9Sstevel@tonic-gate } 13907c478bd9Sstevel@tonic-gate 13917c478bd9Sstevel@tonic-gate if (!ULOCKFS_IS_NOIACC(ITOUL(ip)) && (fs->fs_ronly == 0) && 13927c478bd9Sstevel@tonic-gate (!ufsvfsp->vfs_noatime)) { 13937c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 13947c478bd9Sstevel@tonic-gate ip->i_flag |= IACC; 13957c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 13967c478bd9Sstevel@tonic-gate } 13977c478bd9Sstevel@tonic-gate /* 13987c478bd9Sstevel@tonic-gate * Try to go direct 13997c478bd9Sstevel@tonic-gate */ 14007c478bd9Sstevel@tonic-gate if (ip->i_flag & IDIRECTIO || ufsvfsp->vfs_forcedirectio) { 14017c478bd9Sstevel@tonic-gate error = ufs_directio_read(ip, uio, cr, &directio_status); 14027c478bd9Sstevel@tonic-gate if (directio_status == DIRECTIO_SUCCESS) 14037c478bd9Sstevel@tonic-gate goto out; 14047c478bd9Sstevel@tonic-gate } 14057c478bd9Sstevel@tonic-gate 14067c478bd9Sstevel@tonic-gate rwtype = (rw_write_held(&ip->i_contents)?RW_WRITER:RW_READER); 14077c478bd9Sstevel@tonic-gate 14087c478bd9Sstevel@tonic-gate do { 14097c478bd9Sstevel@tonic-gate offset_t diff; 14107c478bd9Sstevel@tonic-gate u_offset_t uoff = uio->uio_loffset; 14117c478bd9Sstevel@tonic-gate off = uoff & (offset_t)MAXBMASK; 14127c478bd9Sstevel@tonic-gate mapon = (u_offset_t)(uoff & (offset_t)MAXBOFFSET); 14137c478bd9Sstevel@tonic-gate on = (u_offset_t)blkoff(fs, uoff); 14147c478bd9Sstevel@tonic-gate n = MIN((u_offset_t)fs->fs_bsize - on, 14157c478bd9Sstevel@tonic-gate (u_offset_t)uio->uio_resid); 14167c478bd9Sstevel@tonic-gate 14177c478bd9Sstevel@tonic-gate diff = ip->i_size - uoff; 14187c478bd9Sstevel@tonic-gate 14197c478bd9Sstevel@tonic-gate if (diff <= (offset_t)0) { 14207c478bd9Sstevel@tonic-gate error = 0; 14217c478bd9Sstevel@tonic-gate goto out; 14227c478bd9Sstevel@tonic-gate } 14237c478bd9Sstevel@tonic-gate if (diff < (offset_t)n) 14247c478bd9Sstevel@tonic-gate n = (int)diff; 1425562eee46Srbourbon 1426562eee46Srbourbon /* 1427562eee46Srbourbon * We update smallfile2 and smallfile1 at most every second. 1428562eee46Srbourbon */ 1429562eee46Srbourbon if (lbolt >= smallfile_update) { 1430562eee46Srbourbon uint64_t percpufreeb; 1431562eee46Srbourbon if (smallfile1_d == 0) smallfile1_d = SMALLFILE1_D; 1432562eee46Srbourbon if (smallfile2_d == 0) smallfile2_d = SMALLFILE2_D; 1433562eee46Srbourbon percpufreeb = ptob((uint64_t)freemem) / ncpus_online; 1434562eee46Srbourbon smallfile1 = percpufreeb / smallfile1_d; 1435562eee46Srbourbon smallfile2 = percpufreeb / smallfile2_d; 1436562eee46Srbourbon smallfile1 = MAX(smallfile1, smallfile); 1437562eee46Srbourbon smallfile1 = MAX(smallfile1, smallfile64); 1438562eee46Srbourbon smallfile2 = MAX(smallfile1, smallfile2); 1439562eee46Srbourbon smallfile_update = lbolt + hz; 1440562eee46Srbourbon } 1441562eee46Srbourbon 14427c478bd9Sstevel@tonic-gate dofree = freebehind && 1443562eee46Srbourbon ip->i_nextr == (off & PAGEMASK) && off > smallfile1; 14447c478bd9Sstevel@tonic-gate 14457c478bd9Sstevel@tonic-gate /* 14467c478bd9Sstevel@tonic-gate * At this point we can enter ufs_getpage() in one of two 14477c478bd9Sstevel@tonic-gate * ways: 14487c478bd9Sstevel@tonic-gate * 1) segmap_getmapflt() calls ufs_getpage() when the 14497c478bd9Sstevel@tonic-gate * forcefault parameter is true (value of 1 is passed) 14507c478bd9Sstevel@tonic-gate * 2) uiomove() causes a page fault. 14517c478bd9Sstevel@tonic-gate * 14527c478bd9Sstevel@tonic-gate * We cannot hold onto an i_contents reader lock without 14537c478bd9Sstevel@tonic-gate * risking deadlock in ufs_getpage() so drop a reader lock. 14547c478bd9Sstevel@tonic-gate * The ufs_getpage() dolock logic already allows for a 14557c478bd9Sstevel@tonic-gate * thread holding i_contents as writer to work properly 14567c478bd9Sstevel@tonic-gate * so we keep a writer lock. 14577c478bd9Sstevel@tonic-gate */ 14587c478bd9Sstevel@tonic-gate if (rwtype == RW_READER) 14597c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 14607c478bd9Sstevel@tonic-gate 1461a5652762Spraks if (vpm_enable) { 1462a5652762Spraks /* 1463a5652762Spraks * Copy data. 1464a5652762Spraks */ 1465a5652762Spraks error = vpm_data_copy(vp, (off + mapon), (uint_t)n, 1466a5652762Spraks uio, 1, NULL, 0, S_READ); 1467a5652762Spraks } else { 1468a5652762Spraks base = segmap_getmapflt(segkmap, vp, (off + mapon), 1469a5652762Spraks (uint_t)n, 1, S_READ); 1470a5652762Spraks error = uiomove(base + mapon, (long)n, UIO_READ, uio); 1471a5652762Spraks } 14727c478bd9Sstevel@tonic-gate 14737c478bd9Sstevel@tonic-gate flags = 0; 14747c478bd9Sstevel@tonic-gate if (!error) { 14757c478bd9Sstevel@tonic-gate /* 1476562eee46Srbourbon * If reading sequential we won't need this 1477562eee46Srbourbon * buffer again soon. For offsets in range 1478562eee46Srbourbon * [smallfile1, smallfile2] release the pages 1479562eee46Srbourbon * at the tail of the cache list, larger 1480562eee46Srbourbon * offsets are released at the head. 14817c478bd9Sstevel@tonic-gate */ 14827c478bd9Sstevel@tonic-gate if (dofree) { 1483562eee46Srbourbon flags = SM_FREE | SM_ASYNC; 1484562eee46Srbourbon if ((cache_read_ahead == 0) && 1485562eee46Srbourbon (off > smallfile2)) 1486562eee46Srbourbon flags |= SM_DONTNEED; 14877c478bd9Sstevel@tonic-gate } 14887c478bd9Sstevel@tonic-gate /* 14897c478bd9Sstevel@tonic-gate * In POSIX SYNC (FSYNC and FDSYNC) read mode, 14907c478bd9Sstevel@tonic-gate * we want to make sure that the page which has 14917c478bd9Sstevel@tonic-gate * been read, is written on disk if it is dirty. 14927c478bd9Sstevel@tonic-gate * And corresponding indirect blocks should also 14937c478bd9Sstevel@tonic-gate * be flushed out. 14947c478bd9Sstevel@tonic-gate */ 14957c478bd9Sstevel@tonic-gate if ((ioflag & FRSYNC) && (ioflag & (FSYNC|FDSYNC))) { 14967c478bd9Sstevel@tonic-gate flags &= ~SM_ASYNC; 14977c478bd9Sstevel@tonic-gate flags |= SM_WRITE; 14987c478bd9Sstevel@tonic-gate } 1499a5652762Spraks if (vpm_enable) { 1500a5652762Spraks error = vpm_sync_pages(vp, off, n, flags); 1501a5652762Spraks } else { 1502a5652762Spraks error = segmap_release(segkmap, base, flags); 1503a5652762Spraks } 1504a5652762Spraks } else { 1505a5652762Spraks if (vpm_enable) { 1506a5652762Spraks (void) vpm_sync_pages(vp, off, n, flags); 1507a5652762Spraks } else { 1508a5652762Spraks (void) segmap_release(segkmap, base, flags); 1509a5652762Spraks } 1510a5652762Spraks } 15117c478bd9Sstevel@tonic-gate 15127c478bd9Sstevel@tonic-gate if (rwtype == RW_READER) 15137c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, rwtype); 15147c478bd9Sstevel@tonic-gate } while (error == 0 && uio->uio_resid > 0 && n != 0); 15157c478bd9Sstevel@tonic-gate out: 15167c478bd9Sstevel@tonic-gate /* 15177c478bd9Sstevel@tonic-gate * Inode is updated according to this table if FRSYNC is set. 15187c478bd9Sstevel@tonic-gate * 15197c478bd9Sstevel@tonic-gate * FSYNC FDSYNC(posix.4) 15207c478bd9Sstevel@tonic-gate * -------------------------- 15217c478bd9Sstevel@tonic-gate * always IATTCHG|IBDWRITE 15227c478bd9Sstevel@tonic-gate */ 15237c478bd9Sstevel@tonic-gate /* 15247c478bd9Sstevel@tonic-gate * The inode is not updated if we're logging and the inode is a 15257c478bd9Sstevel@tonic-gate * directory with FRSYNC, FSYNC and FDSYNC flags set. 15267c478bd9Sstevel@tonic-gate */ 15277c478bd9Sstevel@tonic-gate if (ioflag & FRSYNC) { 15287c478bd9Sstevel@tonic-gate if (TRANS_ISTRANS(ufsvfsp) && ((ip->i_mode & IFMT) == IFDIR)) { 15297c478bd9Sstevel@tonic-gate doupdate = 0; 15307c478bd9Sstevel@tonic-gate } 15317c478bd9Sstevel@tonic-gate if (doupdate) { 15327c478bd9Sstevel@tonic-gate if ((ioflag & FSYNC) || 15337c478bd9Sstevel@tonic-gate ((ioflag & FDSYNC) && 15347c478bd9Sstevel@tonic-gate (ip->i_flag & (IATTCHG|IBDWRITE)))) { 15357c478bd9Sstevel@tonic-gate ufs_iupdat(ip, 1); 15367c478bd9Sstevel@tonic-gate } 15377c478bd9Sstevel@tonic-gate } 15387c478bd9Sstevel@tonic-gate } 15397c478bd9Sstevel@tonic-gate /* 15407c478bd9Sstevel@tonic-gate * If we've already done a partial read, terminate 15417c478bd9Sstevel@tonic-gate * the read but return no error. 15427c478bd9Sstevel@tonic-gate */ 15437c478bd9Sstevel@tonic-gate if (oresid != uio->uio_resid) 15447c478bd9Sstevel@tonic-gate error = 0; 15457c478bd9Sstevel@tonic-gate ITIMES(ip); 15467c478bd9Sstevel@tonic-gate 15477c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_RWIP_END, 15487c478bd9Sstevel@tonic-gate "ufs_rdip_end:vp %p error %d", vp, error); 15497c478bd9Sstevel@tonic-gate return (error); 15507c478bd9Sstevel@tonic-gate } 15517c478bd9Sstevel@tonic-gate 15527c478bd9Sstevel@tonic-gate /* ARGSUSED */ 15537c478bd9Sstevel@tonic-gate static int 15547c478bd9Sstevel@tonic-gate ufs_ioctl( 15557c478bd9Sstevel@tonic-gate struct vnode *vp, 15567c478bd9Sstevel@tonic-gate int cmd, 15577c478bd9Sstevel@tonic-gate intptr_t arg, 15587c478bd9Sstevel@tonic-gate int flag, 15597c478bd9Sstevel@tonic-gate struct cred *cr, 15607c478bd9Sstevel@tonic-gate int *rvalp) 15617c478bd9Sstevel@tonic-gate { 15627c478bd9Sstevel@tonic-gate struct lockfs lockfs, lockfs_out; 15637c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp = VTOI(vp)->i_ufsvfs; 15647c478bd9Sstevel@tonic-gate char *comment, *original_comment; 15657c478bd9Sstevel@tonic-gate struct fs *fs; 15667c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 15677c478bd9Sstevel@tonic-gate offset_t off; 15687c478bd9Sstevel@tonic-gate extern int maxphys; 15697c478bd9Sstevel@tonic-gate int error; 15707c478bd9Sstevel@tonic-gate int issync; 15717c478bd9Sstevel@tonic-gate int trans_size; 15727c478bd9Sstevel@tonic-gate 15737c478bd9Sstevel@tonic-gate 15747c478bd9Sstevel@tonic-gate /* 15757c478bd9Sstevel@tonic-gate * forcibly unmounted 15767c478bd9Sstevel@tonic-gate */ 15777c478bd9Sstevel@tonic-gate if (ufsvfsp == NULL) { 15787c478bd9Sstevel@tonic-gate return (EIO); 15797c478bd9Sstevel@tonic-gate } 15807c478bd9Sstevel@tonic-gate 15817c478bd9Sstevel@tonic-gate fs = ufsvfsp->vfs_fs; 15827c478bd9Sstevel@tonic-gate 15837c478bd9Sstevel@tonic-gate if (cmd == Q_QUOTACTL) { 15847c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_QUOTA_MASK); 15857c478bd9Sstevel@tonic-gate if (error) 15867c478bd9Sstevel@tonic-gate return (error); 15877c478bd9Sstevel@tonic-gate 15887c478bd9Sstevel@tonic-gate if (ulp) { 15897c478bd9Sstevel@tonic-gate TRANS_BEGIN_ASYNC(ufsvfsp, TOP_QUOTA, 15907c478bd9Sstevel@tonic-gate TOP_SETQUOTA_SIZE(fs)); 15917c478bd9Sstevel@tonic-gate } 15927c478bd9Sstevel@tonic-gate 15937c478bd9Sstevel@tonic-gate error = quotactl(vp, arg, flag, cr); 15947c478bd9Sstevel@tonic-gate 15957c478bd9Sstevel@tonic-gate if (ulp) { 15967c478bd9Sstevel@tonic-gate TRANS_END_ASYNC(ufsvfsp, TOP_QUOTA, 15977c478bd9Sstevel@tonic-gate TOP_SETQUOTA_SIZE(fs)); 15987c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 15997c478bd9Sstevel@tonic-gate } 16007c478bd9Sstevel@tonic-gate return (error); 16017c478bd9Sstevel@tonic-gate } 16027c478bd9Sstevel@tonic-gate 16037c478bd9Sstevel@tonic-gate switch (cmd) { 16047c478bd9Sstevel@tonic-gate case _FIOLFS: 16057c478bd9Sstevel@tonic-gate /* 16067c478bd9Sstevel@tonic-gate * file system locking 16077c478bd9Sstevel@tonic-gate */ 16087c478bd9Sstevel@tonic-gate if (secpolicy_fs_config(cr, ufsvfsp->vfs_vfs) != 0) 16097c478bd9Sstevel@tonic-gate return (EPERM); 16107c478bd9Sstevel@tonic-gate 16117c478bd9Sstevel@tonic-gate if ((flag & DATAMODEL_MASK) == DATAMODEL_NATIVE) { 16127c478bd9Sstevel@tonic-gate if (copyin((caddr_t)arg, &lockfs, 16137c478bd9Sstevel@tonic-gate sizeof (struct lockfs))) 16147c478bd9Sstevel@tonic-gate return (EFAULT); 16157c478bd9Sstevel@tonic-gate } 16167c478bd9Sstevel@tonic-gate #ifdef _SYSCALL32_IMPL 16177c478bd9Sstevel@tonic-gate else { 16187c478bd9Sstevel@tonic-gate struct lockfs32 lockfs32; 16197c478bd9Sstevel@tonic-gate /* Translate ILP32 lockfs to LP64 lockfs */ 16207c478bd9Sstevel@tonic-gate if (copyin((caddr_t)arg, &lockfs32, 16217c478bd9Sstevel@tonic-gate sizeof (struct lockfs32))) 16227c478bd9Sstevel@tonic-gate return (EFAULT); 16237c478bd9Sstevel@tonic-gate lockfs.lf_lock = (ulong_t)lockfs32.lf_lock; 16247c478bd9Sstevel@tonic-gate lockfs.lf_flags = (ulong_t)lockfs32.lf_flags; 16257c478bd9Sstevel@tonic-gate lockfs.lf_key = (ulong_t)lockfs32.lf_key; 16267c478bd9Sstevel@tonic-gate lockfs.lf_comlen = (ulong_t)lockfs32.lf_comlen; 16277c478bd9Sstevel@tonic-gate lockfs.lf_comment = 16287c478bd9Sstevel@tonic-gate (caddr_t)(uintptr_t)lockfs32.lf_comment; 16297c478bd9Sstevel@tonic-gate } 16307c478bd9Sstevel@tonic-gate #endif /* _SYSCALL32_IMPL */ 16317c478bd9Sstevel@tonic-gate 16327c478bd9Sstevel@tonic-gate if (lockfs.lf_comlen) { 16337c478bd9Sstevel@tonic-gate if (lockfs.lf_comlen > LOCKFS_MAXCOMMENTLEN) 16347c478bd9Sstevel@tonic-gate return (ENAMETOOLONG); 16357c478bd9Sstevel@tonic-gate comment = kmem_alloc(lockfs.lf_comlen, 16367c478bd9Sstevel@tonic-gate KM_SLEEP); 16377c478bd9Sstevel@tonic-gate if (copyin(lockfs.lf_comment, comment, 16387c478bd9Sstevel@tonic-gate lockfs.lf_comlen)) { 16397c478bd9Sstevel@tonic-gate kmem_free(comment, lockfs.lf_comlen); 16407c478bd9Sstevel@tonic-gate return (EFAULT); 16417c478bd9Sstevel@tonic-gate } 16427c478bd9Sstevel@tonic-gate original_comment = lockfs.lf_comment; 16437c478bd9Sstevel@tonic-gate lockfs.lf_comment = comment; 16447c478bd9Sstevel@tonic-gate } 16457c478bd9Sstevel@tonic-gate if ((error = ufs_fiolfs(vp, &lockfs, 0)) == 0) { 16467c478bd9Sstevel@tonic-gate lockfs.lf_comment = original_comment; 16477c478bd9Sstevel@tonic-gate 16487c478bd9Sstevel@tonic-gate if ((flag & DATAMODEL_MASK) == 16497c478bd9Sstevel@tonic-gate DATAMODEL_NATIVE) { 16507c478bd9Sstevel@tonic-gate (void) copyout(&lockfs, (caddr_t)arg, 16517c478bd9Sstevel@tonic-gate sizeof (struct lockfs)); 16527c478bd9Sstevel@tonic-gate } 16537c478bd9Sstevel@tonic-gate #ifdef _SYSCALL32_IMPL 16547c478bd9Sstevel@tonic-gate else { 16557c478bd9Sstevel@tonic-gate struct lockfs32 lockfs32; 16567c478bd9Sstevel@tonic-gate /* Translate LP64 to ILP32 lockfs */ 16577c478bd9Sstevel@tonic-gate lockfs32.lf_lock = 16587c478bd9Sstevel@tonic-gate (uint32_t)lockfs.lf_lock; 16597c478bd9Sstevel@tonic-gate lockfs32.lf_flags = 16607c478bd9Sstevel@tonic-gate (uint32_t)lockfs.lf_flags; 16617c478bd9Sstevel@tonic-gate lockfs32.lf_key = 16627c478bd9Sstevel@tonic-gate (uint32_t)lockfs.lf_key; 16637c478bd9Sstevel@tonic-gate lockfs32.lf_comlen = 16647c478bd9Sstevel@tonic-gate (uint32_t)lockfs.lf_comlen; 16657c478bd9Sstevel@tonic-gate lockfs32.lf_comment = 16667c478bd9Sstevel@tonic-gate (uint32_t)(uintptr_t)lockfs.lf_comment; 16677c478bd9Sstevel@tonic-gate (void) copyout(&lockfs32, (caddr_t)arg, 16687c478bd9Sstevel@tonic-gate sizeof (struct lockfs32)); 16697c478bd9Sstevel@tonic-gate } 16707c478bd9Sstevel@tonic-gate #endif /* _SYSCALL32_IMPL */ 16717c478bd9Sstevel@tonic-gate 16727c478bd9Sstevel@tonic-gate } else { 16737c478bd9Sstevel@tonic-gate if (lockfs.lf_comlen) 16747c478bd9Sstevel@tonic-gate kmem_free(comment, lockfs.lf_comlen); 16757c478bd9Sstevel@tonic-gate } 16767c478bd9Sstevel@tonic-gate return (error); 16777c478bd9Sstevel@tonic-gate 16787c478bd9Sstevel@tonic-gate case _FIOLFSS: 16797c478bd9Sstevel@tonic-gate /* 16807c478bd9Sstevel@tonic-gate * get file system locking status 16817c478bd9Sstevel@tonic-gate */ 16827c478bd9Sstevel@tonic-gate 16837c478bd9Sstevel@tonic-gate if ((flag & DATAMODEL_MASK) == DATAMODEL_NATIVE) { 16847c478bd9Sstevel@tonic-gate if (copyin((caddr_t)arg, &lockfs, 16857c478bd9Sstevel@tonic-gate sizeof (struct lockfs))) 16867c478bd9Sstevel@tonic-gate return (EFAULT); 16877c478bd9Sstevel@tonic-gate } 16887c478bd9Sstevel@tonic-gate #ifdef _SYSCALL32_IMPL 16897c478bd9Sstevel@tonic-gate else { 16907c478bd9Sstevel@tonic-gate struct lockfs32 lockfs32; 16917c478bd9Sstevel@tonic-gate /* Translate ILP32 lockfs to LP64 lockfs */ 16927c478bd9Sstevel@tonic-gate if (copyin((caddr_t)arg, &lockfs32, 16937c478bd9Sstevel@tonic-gate sizeof (struct lockfs32))) 16947c478bd9Sstevel@tonic-gate return (EFAULT); 16957c478bd9Sstevel@tonic-gate lockfs.lf_lock = (ulong_t)lockfs32.lf_lock; 16967c478bd9Sstevel@tonic-gate lockfs.lf_flags = (ulong_t)lockfs32.lf_flags; 16977c478bd9Sstevel@tonic-gate lockfs.lf_key = (ulong_t)lockfs32.lf_key; 16987c478bd9Sstevel@tonic-gate lockfs.lf_comlen = (ulong_t)lockfs32.lf_comlen; 16997c478bd9Sstevel@tonic-gate lockfs.lf_comment = 17007c478bd9Sstevel@tonic-gate (caddr_t)(uintptr_t)lockfs32.lf_comment; 17017c478bd9Sstevel@tonic-gate } 17027c478bd9Sstevel@tonic-gate #endif /* _SYSCALL32_IMPL */ 17037c478bd9Sstevel@tonic-gate 17047c478bd9Sstevel@tonic-gate if (error = ufs_fiolfss(vp, &lockfs_out)) 17057c478bd9Sstevel@tonic-gate return (error); 17067c478bd9Sstevel@tonic-gate lockfs.lf_lock = lockfs_out.lf_lock; 17077c478bd9Sstevel@tonic-gate lockfs.lf_key = lockfs_out.lf_key; 17087c478bd9Sstevel@tonic-gate lockfs.lf_flags = lockfs_out.lf_flags; 17097c478bd9Sstevel@tonic-gate lockfs.lf_comlen = MIN(lockfs.lf_comlen, 17107c478bd9Sstevel@tonic-gate lockfs_out.lf_comlen); 17117c478bd9Sstevel@tonic-gate 17127c478bd9Sstevel@tonic-gate if ((flag & DATAMODEL_MASK) == DATAMODEL_NATIVE) { 17137c478bd9Sstevel@tonic-gate if (copyout(&lockfs, (caddr_t)arg, 17147c478bd9Sstevel@tonic-gate sizeof (struct lockfs))) 17157c478bd9Sstevel@tonic-gate return (EFAULT); 17167c478bd9Sstevel@tonic-gate } 17177c478bd9Sstevel@tonic-gate #ifdef _SYSCALL32_IMPL 17187c478bd9Sstevel@tonic-gate else { 17197c478bd9Sstevel@tonic-gate /* Translate LP64 to ILP32 lockfs */ 17207c478bd9Sstevel@tonic-gate struct lockfs32 lockfs32; 17217c478bd9Sstevel@tonic-gate lockfs32.lf_lock = (uint32_t)lockfs.lf_lock; 17227c478bd9Sstevel@tonic-gate lockfs32.lf_flags = (uint32_t)lockfs.lf_flags; 17237c478bd9Sstevel@tonic-gate lockfs32.lf_key = (uint32_t)lockfs.lf_key; 17247c478bd9Sstevel@tonic-gate lockfs32.lf_comlen = (uint32_t)lockfs.lf_comlen; 17257c478bd9Sstevel@tonic-gate lockfs32.lf_comment = 17267c478bd9Sstevel@tonic-gate (uint32_t)(uintptr_t)lockfs.lf_comment; 17277c478bd9Sstevel@tonic-gate if (copyout(&lockfs32, (caddr_t)arg, 17287c478bd9Sstevel@tonic-gate sizeof (struct lockfs32))) 17297c478bd9Sstevel@tonic-gate return (EFAULT); 17307c478bd9Sstevel@tonic-gate } 17317c478bd9Sstevel@tonic-gate #endif /* _SYSCALL32_IMPL */ 17327c478bd9Sstevel@tonic-gate 17337c478bd9Sstevel@tonic-gate if (lockfs.lf_comlen && 17347c478bd9Sstevel@tonic-gate lockfs.lf_comment && lockfs_out.lf_comment) 17357c478bd9Sstevel@tonic-gate if (copyout(lockfs_out.lf_comment, 17367c478bd9Sstevel@tonic-gate lockfs.lf_comment, 17377c478bd9Sstevel@tonic-gate lockfs.lf_comlen)) 17387c478bd9Sstevel@tonic-gate return (EFAULT); 17397c478bd9Sstevel@tonic-gate return (0); 17407c478bd9Sstevel@tonic-gate 17417c478bd9Sstevel@tonic-gate case _FIOSATIME: 17427c478bd9Sstevel@tonic-gate /* 17437c478bd9Sstevel@tonic-gate * set access time 17447c478bd9Sstevel@tonic-gate */ 17457c478bd9Sstevel@tonic-gate 17467c478bd9Sstevel@tonic-gate /* 17477c478bd9Sstevel@tonic-gate * if mounted w/o atime, return quietly. 17487c478bd9Sstevel@tonic-gate * I briefly thought about returning ENOSYS, but 17497c478bd9Sstevel@tonic-gate * figured that most apps would consider this fatal 17507c478bd9Sstevel@tonic-gate * but the idea is to make this as seamless as poss. 17517c478bd9Sstevel@tonic-gate */ 17527c478bd9Sstevel@tonic-gate if (ufsvfsp->vfs_noatime) 17537c478bd9Sstevel@tonic-gate return (0); 17547c478bd9Sstevel@tonic-gate 17557c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, 17567c478bd9Sstevel@tonic-gate ULOCKFS_SETATTR_MASK); 17577c478bd9Sstevel@tonic-gate if (error) 17587c478bd9Sstevel@tonic-gate return (error); 17597c478bd9Sstevel@tonic-gate 17607c478bd9Sstevel@tonic-gate if (ulp) { 17617c478bd9Sstevel@tonic-gate trans_size = (int)TOP_SETATTR_SIZE(VTOI(vp)); 17627c478bd9Sstevel@tonic-gate TRANS_BEGIN_CSYNC(ufsvfsp, issync, 17637c478bd9Sstevel@tonic-gate TOP_SETATTR, trans_size); 17647c478bd9Sstevel@tonic-gate } 17657c478bd9Sstevel@tonic-gate 17667c478bd9Sstevel@tonic-gate error = ufs_fiosatime(vp, (struct timeval *)arg, 17677c478bd9Sstevel@tonic-gate flag, cr); 17687c478bd9Sstevel@tonic-gate 17697c478bd9Sstevel@tonic-gate if (ulp) { 17707c478bd9Sstevel@tonic-gate TRANS_END_CSYNC(ufsvfsp, error, issync, 17717c478bd9Sstevel@tonic-gate TOP_SETATTR, trans_size); 17727c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 17737c478bd9Sstevel@tonic-gate } 17747c478bd9Sstevel@tonic-gate return (error); 17757c478bd9Sstevel@tonic-gate 17767c478bd9Sstevel@tonic-gate case _FIOSDIO: 17777c478bd9Sstevel@tonic-gate /* 17787c478bd9Sstevel@tonic-gate * set delayed-io 17797c478bd9Sstevel@tonic-gate */ 17807c478bd9Sstevel@tonic-gate return (ufs_fiosdio(vp, (uint_t *)arg, flag, cr)); 17817c478bd9Sstevel@tonic-gate 17827c478bd9Sstevel@tonic-gate case _FIOGDIO: 17837c478bd9Sstevel@tonic-gate /* 17847c478bd9Sstevel@tonic-gate * get delayed-io 17857c478bd9Sstevel@tonic-gate */ 17867c478bd9Sstevel@tonic-gate return (ufs_fiogdio(vp, (uint_t *)arg, flag, cr)); 17877c478bd9Sstevel@tonic-gate 17887c478bd9Sstevel@tonic-gate case _FIOIO: 17897c478bd9Sstevel@tonic-gate /* 17907c478bd9Sstevel@tonic-gate * inode open 17917c478bd9Sstevel@tonic-gate */ 17927c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, 17937c478bd9Sstevel@tonic-gate ULOCKFS_VGET_MASK); 17947c478bd9Sstevel@tonic-gate if (error) 17957c478bd9Sstevel@tonic-gate return (error); 17967c478bd9Sstevel@tonic-gate 17977c478bd9Sstevel@tonic-gate error = ufs_fioio(vp, (struct fioio *)arg, flag, cr); 17987c478bd9Sstevel@tonic-gate 17997c478bd9Sstevel@tonic-gate if (ulp) { 18007c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 18017c478bd9Sstevel@tonic-gate } 18027c478bd9Sstevel@tonic-gate return (error); 18037c478bd9Sstevel@tonic-gate 18047c478bd9Sstevel@tonic-gate case _FIOFFS: 18057c478bd9Sstevel@tonic-gate /* 18067c478bd9Sstevel@tonic-gate * file system flush (push w/invalidate) 18077c478bd9Sstevel@tonic-gate */ 18087c478bd9Sstevel@tonic-gate if ((caddr_t)arg != NULL) 18097c478bd9Sstevel@tonic-gate return (EINVAL); 18107c478bd9Sstevel@tonic-gate return (ufs_fioffs(vp, NULL, cr)); 18117c478bd9Sstevel@tonic-gate 18127c478bd9Sstevel@tonic-gate case _FIOISBUSY: 18137c478bd9Sstevel@tonic-gate /* 18147c478bd9Sstevel@tonic-gate * Contract-private interface for Legato 18157c478bd9Sstevel@tonic-gate * Purge this vnode from the DNLC and decide 18167c478bd9Sstevel@tonic-gate * if this vnode is busy (*arg == 1) or not 18177c478bd9Sstevel@tonic-gate * (*arg == 0) 18187c478bd9Sstevel@tonic-gate */ 18197c478bd9Sstevel@tonic-gate if (secpolicy_fs_config(cr, ufsvfsp->vfs_vfs) != 0) 18207c478bd9Sstevel@tonic-gate return (EPERM); 18217c478bd9Sstevel@tonic-gate error = ufs_fioisbusy(vp, (int *)arg, cr); 18227c478bd9Sstevel@tonic-gate return (error); 18237c478bd9Sstevel@tonic-gate 18247c478bd9Sstevel@tonic-gate case _FIODIRECTIO: 18257c478bd9Sstevel@tonic-gate return (ufs_fiodirectio(vp, (int)arg, cr)); 18267c478bd9Sstevel@tonic-gate 18277c478bd9Sstevel@tonic-gate case _FIOTUNE: 18287c478bd9Sstevel@tonic-gate /* 18297c478bd9Sstevel@tonic-gate * Tune the file system (aka setting fs attributes) 18307c478bd9Sstevel@tonic-gate */ 18317c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, 18327c478bd9Sstevel@tonic-gate ULOCKFS_SETATTR_MASK); 18337c478bd9Sstevel@tonic-gate if (error) 18347c478bd9Sstevel@tonic-gate return (error); 18357c478bd9Sstevel@tonic-gate 18367c478bd9Sstevel@tonic-gate error = ufs_fiotune(vp, (struct fiotune *)arg, cr); 18377c478bd9Sstevel@tonic-gate 18387c478bd9Sstevel@tonic-gate if (ulp) 18397c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 18407c478bd9Sstevel@tonic-gate return (error); 18417c478bd9Sstevel@tonic-gate 18427c478bd9Sstevel@tonic-gate case _FIOLOGENABLE: 18437c478bd9Sstevel@tonic-gate if (secpolicy_fs_config(cr, ufsvfsp->vfs_vfs) != 0) 18447c478bd9Sstevel@tonic-gate return (EPERM); 18457c478bd9Sstevel@tonic-gate return (ufs_fiologenable(vp, (void *)arg, cr, flag)); 18467c478bd9Sstevel@tonic-gate 18477c478bd9Sstevel@tonic-gate case _FIOLOGDISABLE: 18487c478bd9Sstevel@tonic-gate if (secpolicy_fs_config(cr, ufsvfsp->vfs_vfs) != 0) 18497c478bd9Sstevel@tonic-gate return (EPERM); 18507c478bd9Sstevel@tonic-gate return (ufs_fiologdisable(vp, (void *)arg, cr, flag)); 18517c478bd9Sstevel@tonic-gate 18527c478bd9Sstevel@tonic-gate case _FIOISLOG: 18537c478bd9Sstevel@tonic-gate return (ufs_fioislog(vp, (void *)arg, cr, flag)); 18547c478bd9Sstevel@tonic-gate 18557c478bd9Sstevel@tonic-gate case _FIOSNAPSHOTCREATE_MULTI: 18567c478bd9Sstevel@tonic-gate { 18577c478bd9Sstevel@tonic-gate struct fiosnapcreate_multi fc, *fcp; 18587c478bd9Sstevel@tonic-gate size_t fcm_size; 18597c478bd9Sstevel@tonic-gate 18607c478bd9Sstevel@tonic-gate if (copyin((void *)arg, &fc, sizeof (fc))) 18617c478bd9Sstevel@tonic-gate return (EFAULT); 18627c478bd9Sstevel@tonic-gate if (fc.backfilecount > MAX_BACKFILE_COUNT) 18637c478bd9Sstevel@tonic-gate return (EINVAL); 18647c478bd9Sstevel@tonic-gate fcm_size = sizeof (struct fiosnapcreate_multi) + 18657c478bd9Sstevel@tonic-gate (fc.backfilecount - 1) * sizeof (int); 18667c478bd9Sstevel@tonic-gate fcp = (struct fiosnapcreate_multi *) 18677c478bd9Sstevel@tonic-gate kmem_alloc(fcm_size, KM_SLEEP); 18687c478bd9Sstevel@tonic-gate if (copyin((void *)arg, fcp, fcm_size)) { 18697c478bd9Sstevel@tonic-gate kmem_free(fcp, fcm_size); 18707c478bd9Sstevel@tonic-gate return (EFAULT); 18717c478bd9Sstevel@tonic-gate } 18727c478bd9Sstevel@tonic-gate error = ufs_snap_create(vp, fcp, cr); 18737c478bd9Sstevel@tonic-gate if (!error && copyout(fcp, (void *)arg, fcm_size)) 18747c478bd9Sstevel@tonic-gate error = EFAULT; 18757c478bd9Sstevel@tonic-gate kmem_free(fcp, fcm_size); 18767c478bd9Sstevel@tonic-gate return (error); 18777c478bd9Sstevel@tonic-gate } 18787c478bd9Sstevel@tonic-gate 18797c478bd9Sstevel@tonic-gate case _FIOSNAPSHOTDELETE: 18807c478bd9Sstevel@tonic-gate { 18817c478bd9Sstevel@tonic-gate struct fiosnapdelete fc; 18827c478bd9Sstevel@tonic-gate 18837c478bd9Sstevel@tonic-gate if (copyin((void *)arg, &fc, sizeof (fc))) 18847c478bd9Sstevel@tonic-gate return (EFAULT); 18857c478bd9Sstevel@tonic-gate error = ufs_snap_delete(vp, &fc, cr); 18867c478bd9Sstevel@tonic-gate if (!error && copyout(&fc, (void *)arg, sizeof (fc))) 18877c478bd9Sstevel@tonic-gate error = EFAULT; 18887c478bd9Sstevel@tonic-gate return (error); 18897c478bd9Sstevel@tonic-gate } 18907c478bd9Sstevel@tonic-gate 18917c478bd9Sstevel@tonic-gate case _FIOGETSUPERBLOCK: 18927c478bd9Sstevel@tonic-gate if (copyout(fs, (void *)arg, SBSIZE)) 18937c478bd9Sstevel@tonic-gate return (EFAULT); 18947c478bd9Sstevel@tonic-gate return (0); 18957c478bd9Sstevel@tonic-gate 18967c478bd9Sstevel@tonic-gate case _FIOGETMAXPHYS: 18977c478bd9Sstevel@tonic-gate if (copyout(&maxphys, (void *)arg, sizeof (maxphys))) 18987c478bd9Sstevel@tonic-gate return (EFAULT); 18997c478bd9Sstevel@tonic-gate return (0); 19007c478bd9Sstevel@tonic-gate 19017c478bd9Sstevel@tonic-gate /* 19027c478bd9Sstevel@tonic-gate * The following 3 ioctls are for TSufs support 19037c478bd9Sstevel@tonic-gate * although could potentially be used elsewhere 19047c478bd9Sstevel@tonic-gate */ 19057c478bd9Sstevel@tonic-gate case _FIO_SET_LUFS_DEBUG: 19067c478bd9Sstevel@tonic-gate if (secpolicy_fs_config(cr, ufsvfsp->vfs_vfs) != 0) 19077c478bd9Sstevel@tonic-gate return (EPERM); 19087c478bd9Sstevel@tonic-gate lufs_debug = (uint32_t)arg; 19097c478bd9Sstevel@tonic-gate return (0); 19107c478bd9Sstevel@tonic-gate 19117c478bd9Sstevel@tonic-gate case _FIO_SET_LUFS_ERROR: 19127c478bd9Sstevel@tonic-gate if (secpolicy_fs_config(cr, ufsvfsp->vfs_vfs) != 0) 19137c478bd9Sstevel@tonic-gate return (EPERM); 19147c478bd9Sstevel@tonic-gate TRANS_SETERROR(ufsvfsp); 19157c478bd9Sstevel@tonic-gate return (0); 19167c478bd9Sstevel@tonic-gate 19177c478bd9Sstevel@tonic-gate case _FIO_GET_TOP_STATS: 19187c478bd9Sstevel@tonic-gate { 19197c478bd9Sstevel@tonic-gate fio_lufs_stats_t *ls; 19207c478bd9Sstevel@tonic-gate ml_unit_t *ul = ufsvfsp->vfs_log; 19217c478bd9Sstevel@tonic-gate 19227c478bd9Sstevel@tonic-gate ls = kmem_zalloc(sizeof (*ls), KM_SLEEP); 19237c478bd9Sstevel@tonic-gate ls->ls_debug = ul->un_debug; /* return debug value */ 19247c478bd9Sstevel@tonic-gate /* Copy stucture if statistics are being kept */ 19257c478bd9Sstevel@tonic-gate if (ul->un_logmap->mtm_tops) { 19267c478bd9Sstevel@tonic-gate ls->ls_topstats = *(ul->un_logmap->mtm_tops); 19277c478bd9Sstevel@tonic-gate } 19287c478bd9Sstevel@tonic-gate error = 0; 19297c478bd9Sstevel@tonic-gate if (copyout(ls, (void *)arg, sizeof (*ls))) 19307c478bd9Sstevel@tonic-gate error = EFAULT; 19317c478bd9Sstevel@tonic-gate kmem_free(ls, sizeof (*ls)); 19327c478bd9Sstevel@tonic-gate return (error); 19337c478bd9Sstevel@tonic-gate } 19347c478bd9Sstevel@tonic-gate 19357c478bd9Sstevel@tonic-gate case _FIO_SEEK_DATA: 19367c478bd9Sstevel@tonic-gate case _FIO_SEEK_HOLE: 19377c478bd9Sstevel@tonic-gate if (ddi_copyin((void *)arg, &off, sizeof (off), flag)) 19387c478bd9Sstevel@tonic-gate return (EFAULT); 19397c478bd9Sstevel@tonic-gate /* offset paramater is in/out */ 19407c478bd9Sstevel@tonic-gate error = ufs_fio_holey(vp, cmd, &off); 19417c478bd9Sstevel@tonic-gate if (error) 19427c478bd9Sstevel@tonic-gate return (error); 19437c478bd9Sstevel@tonic-gate if (ddi_copyout(&off, (void *)arg, sizeof (off), flag)) 19447c478bd9Sstevel@tonic-gate return (EFAULT); 19457c478bd9Sstevel@tonic-gate return (0); 19467c478bd9Sstevel@tonic-gate 19477c478bd9Sstevel@tonic-gate default: 19487c478bd9Sstevel@tonic-gate return (ENOTTY); 19497c478bd9Sstevel@tonic-gate } 19507c478bd9Sstevel@tonic-gate } 19517c478bd9Sstevel@tonic-gate 19527c478bd9Sstevel@tonic-gate /* ARGSUSED */ 19537c478bd9Sstevel@tonic-gate static int 19547c478bd9Sstevel@tonic-gate ufs_getattr(struct vnode *vp, struct vattr *vap, int flags, 19557c478bd9Sstevel@tonic-gate struct cred *cr) 19567c478bd9Sstevel@tonic-gate { 19577c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 19587c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp; 19597c478bd9Sstevel@tonic-gate int err; 19607c478bd9Sstevel@tonic-gate 19617c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_GETATTR_START, 19627c478bd9Sstevel@tonic-gate "ufs_getattr_start:vp %p flags %x", vp, flags); 19637c478bd9Sstevel@tonic-gate 19647c478bd9Sstevel@tonic-gate if (vap->va_mask == AT_SIZE) { 19657c478bd9Sstevel@tonic-gate /* 19667c478bd9Sstevel@tonic-gate * for performance, if only the size is requested don't bother 19677c478bd9Sstevel@tonic-gate * with anything else. 19687c478bd9Sstevel@tonic-gate */ 19697c478bd9Sstevel@tonic-gate UFS_GET_ISIZE(&vap->va_size, ip); 19707c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_GETATTR_END, 19717c478bd9Sstevel@tonic-gate "ufs_getattr_end:vp %p", vp); 19727c478bd9Sstevel@tonic-gate return (0); 19737c478bd9Sstevel@tonic-gate } 19747c478bd9Sstevel@tonic-gate 19757c478bd9Sstevel@tonic-gate /* 19767c478bd9Sstevel@tonic-gate * inlined lockfs checks 19777c478bd9Sstevel@tonic-gate */ 19787c478bd9Sstevel@tonic-gate ufsvfsp = ip->i_ufsvfs; 19797c478bd9Sstevel@tonic-gate if ((ufsvfsp == NULL) || ULOCKFS_IS_HLOCK(&ufsvfsp->vfs_ulockfs)) { 19807c478bd9Sstevel@tonic-gate err = EIO; 19817c478bd9Sstevel@tonic-gate goto out; 19827c478bd9Sstevel@tonic-gate } 19837c478bd9Sstevel@tonic-gate 19847c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_READER); 19857c478bd9Sstevel@tonic-gate /* 19867c478bd9Sstevel@tonic-gate * Return all the attributes. This should be refined so 19877c478bd9Sstevel@tonic-gate * that it only returns what's asked for. 19887c478bd9Sstevel@tonic-gate */ 19897c478bd9Sstevel@tonic-gate 19907c478bd9Sstevel@tonic-gate /* 19917c478bd9Sstevel@tonic-gate * Copy from inode table. 19927c478bd9Sstevel@tonic-gate */ 19937c478bd9Sstevel@tonic-gate vap->va_type = vp->v_type; 19947c478bd9Sstevel@tonic-gate vap->va_mode = ip->i_mode & MODEMASK; 19957c478bd9Sstevel@tonic-gate /* 19967c478bd9Sstevel@tonic-gate * If there is an ACL and there is a mask entry, then do the 19977c478bd9Sstevel@tonic-gate * extra work that completes the equivalent of an acltomode(3) 19987c478bd9Sstevel@tonic-gate * call. According to POSIX P1003.1e, the acl mask should be 19997c478bd9Sstevel@tonic-gate * returned in the group permissions field. 20007c478bd9Sstevel@tonic-gate * 20017c478bd9Sstevel@tonic-gate * - start with the original permission and mode bits (from above) 20027c478bd9Sstevel@tonic-gate * - clear the group owner bits 20037c478bd9Sstevel@tonic-gate * - add in the mask bits. 20047c478bd9Sstevel@tonic-gate */ 20057c478bd9Sstevel@tonic-gate if (ip->i_ufs_acl && ip->i_ufs_acl->aclass.acl_ismask) { 20067c478bd9Sstevel@tonic-gate vap->va_mode &= ~((VREAD | VWRITE | VEXEC) >> 3); 20077c478bd9Sstevel@tonic-gate vap->va_mode |= 20087c478bd9Sstevel@tonic-gate (ip->i_ufs_acl->aclass.acl_maskbits & PERMMASK) << 3; 20097c478bd9Sstevel@tonic-gate } 20107c478bd9Sstevel@tonic-gate vap->va_uid = ip->i_uid; 20117c478bd9Sstevel@tonic-gate vap->va_gid = ip->i_gid; 20127c478bd9Sstevel@tonic-gate vap->va_fsid = ip->i_dev; 20137c478bd9Sstevel@tonic-gate vap->va_nodeid = (ino64_t)ip->i_number; 20147c478bd9Sstevel@tonic-gate vap->va_nlink = ip->i_nlink; 20157c478bd9Sstevel@tonic-gate vap->va_size = ip->i_size; 20167c478bd9Sstevel@tonic-gate if (vp->v_type == VCHR || vp->v_type == VBLK) 20177c478bd9Sstevel@tonic-gate vap->va_rdev = ip->i_rdev; 20187c478bd9Sstevel@tonic-gate else 20197c478bd9Sstevel@tonic-gate vap->va_rdev = 0; /* not a b/c spec. */ 20207c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 20217c478bd9Sstevel@tonic-gate ITIMES_NOLOCK(ip); /* mark correct time in inode */ 20227c478bd9Sstevel@tonic-gate vap->va_seq = ip->i_seq; 20237c478bd9Sstevel@tonic-gate vap->va_atime.tv_sec = (time_t)ip->i_atime.tv_sec; 20247c478bd9Sstevel@tonic-gate vap->va_atime.tv_nsec = ip->i_atime.tv_usec*1000; 20257c478bd9Sstevel@tonic-gate vap->va_mtime.tv_sec = (time_t)ip->i_mtime.tv_sec; 20267c478bd9Sstevel@tonic-gate vap->va_mtime.tv_nsec = ip->i_mtime.tv_usec*1000; 20277c478bd9Sstevel@tonic-gate vap->va_ctime.tv_sec = (time_t)ip->i_ctime.tv_sec; 20287c478bd9Sstevel@tonic-gate vap->va_ctime.tv_nsec = ip->i_ctime.tv_usec*1000; 20297c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 20307c478bd9Sstevel@tonic-gate 20317c478bd9Sstevel@tonic-gate switch (ip->i_mode & IFMT) { 20327c478bd9Sstevel@tonic-gate 20337c478bd9Sstevel@tonic-gate case IFBLK: 20347c478bd9Sstevel@tonic-gate vap->va_blksize = MAXBSIZE; /* was BLKDEV_IOSIZE */ 20357c478bd9Sstevel@tonic-gate break; 20367c478bd9Sstevel@tonic-gate 20377c478bd9Sstevel@tonic-gate case IFCHR: 20387c478bd9Sstevel@tonic-gate vap->va_blksize = MAXBSIZE; 20397c478bd9Sstevel@tonic-gate break; 20407c478bd9Sstevel@tonic-gate 20417c478bd9Sstevel@tonic-gate default: 20427c478bd9Sstevel@tonic-gate vap->va_blksize = ip->i_fs->fs_bsize; 20437c478bd9Sstevel@tonic-gate break; 20447c478bd9Sstevel@tonic-gate } 20457c478bd9Sstevel@tonic-gate vap->va_nblocks = (fsblkcnt64_t)ip->i_blocks; 20467c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 20477c478bd9Sstevel@tonic-gate err = 0; 20487c478bd9Sstevel@tonic-gate 20497c478bd9Sstevel@tonic-gate out: 20507c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_GETATTR_END, "ufs_getattr_end:vp %p", vp); 20517c478bd9Sstevel@tonic-gate 20527c478bd9Sstevel@tonic-gate return (err); 20537c478bd9Sstevel@tonic-gate } 20547c478bd9Sstevel@tonic-gate 20557c478bd9Sstevel@tonic-gate /*ARGSUSED4*/ 20567c478bd9Sstevel@tonic-gate static int 20577c478bd9Sstevel@tonic-gate ufs_setattr( 20587c478bd9Sstevel@tonic-gate struct vnode *vp, 20597c478bd9Sstevel@tonic-gate struct vattr *vap, 20607c478bd9Sstevel@tonic-gate int flags, 20617c478bd9Sstevel@tonic-gate struct cred *cr, 20627c478bd9Sstevel@tonic-gate caller_context_t *ct) 20637c478bd9Sstevel@tonic-gate { 20647c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 20657c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp = ip->i_ufsvfs; 20667c478bd9Sstevel@tonic-gate struct fs *fs; 20677c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 20687c478bd9Sstevel@tonic-gate char *errmsg1; 20697c478bd9Sstevel@tonic-gate char *errmsg2; 20707c478bd9Sstevel@tonic-gate long blocks; 20717c478bd9Sstevel@tonic-gate long int mask = vap->va_mask; 20727c478bd9Sstevel@tonic-gate size_t len1, len2; 20737c478bd9Sstevel@tonic-gate int issync; 20747c478bd9Sstevel@tonic-gate int trans_size; 20757c478bd9Sstevel@tonic-gate int dotrans; 20767c478bd9Sstevel@tonic-gate int dorwlock; 20777c478bd9Sstevel@tonic-gate int error; 20787c478bd9Sstevel@tonic-gate int owner_change; 20797c478bd9Sstevel@tonic-gate int dodqlock; 20807c478bd9Sstevel@tonic-gate timestruc_t now; 20817c478bd9Sstevel@tonic-gate vattr_t oldva; 20827c478bd9Sstevel@tonic-gate int retry = 1; 2083*02ff05a9Svsakar int indeadlock; 20847c478bd9Sstevel@tonic-gate 20857c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_SETATTR_START, 20867c478bd9Sstevel@tonic-gate "ufs_setattr_start:vp %p flags %x", vp, flags); 20877c478bd9Sstevel@tonic-gate 20887c478bd9Sstevel@tonic-gate /* 20897c478bd9Sstevel@tonic-gate * Cannot set these attributes. 20907c478bd9Sstevel@tonic-gate */ 20917c478bd9Sstevel@tonic-gate if (mask & AT_NOSET) { 20927c478bd9Sstevel@tonic-gate error = EINVAL; 20937c478bd9Sstevel@tonic-gate goto out; 20947c478bd9Sstevel@tonic-gate } 20957c478bd9Sstevel@tonic-gate 20967c478bd9Sstevel@tonic-gate /* 20977c478bd9Sstevel@tonic-gate * check for forced unmount 20987c478bd9Sstevel@tonic-gate */ 20997c478bd9Sstevel@tonic-gate if (ufsvfsp == NULL) 21007c478bd9Sstevel@tonic-gate return (EIO); 21017c478bd9Sstevel@tonic-gate 21027c478bd9Sstevel@tonic-gate fs = ufsvfsp->vfs_fs; 21037c478bd9Sstevel@tonic-gate if (fs->fs_ronly != 0) 21047c478bd9Sstevel@tonic-gate return (EROFS); 21057c478bd9Sstevel@tonic-gate 21067c478bd9Sstevel@tonic-gate again: 21077c478bd9Sstevel@tonic-gate errmsg1 = NULL; 21087c478bd9Sstevel@tonic-gate errmsg2 = NULL; 21097c478bd9Sstevel@tonic-gate dotrans = 0; 21107c478bd9Sstevel@tonic-gate dorwlock = 0; 21117c478bd9Sstevel@tonic-gate dodqlock = 0; 21127c478bd9Sstevel@tonic-gate 21137c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_SETATTR_MASK); 21147c478bd9Sstevel@tonic-gate if (error) 21157c478bd9Sstevel@tonic-gate goto out; 21167c478bd9Sstevel@tonic-gate 21177c478bd9Sstevel@tonic-gate /* 21187c478bd9Sstevel@tonic-gate * Acquire i_rwlock before TRANS_BEGIN_CSYNC() if this is a file. 21197c478bd9Sstevel@tonic-gate * This follows the protocol for read()/write(). 21207c478bd9Sstevel@tonic-gate */ 21217c478bd9Sstevel@tonic-gate if (vp->v_type != VDIR) { 2122*02ff05a9Svsakar /* 2123*02ff05a9Svsakar * ufs_tryirwlock uses rw_tryenter and checks for SLOCK to 2124*02ff05a9Svsakar * avoid i_rwlock, ufs_lockfs_begin deadlock. If deadlock 2125*02ff05a9Svsakar * possible, retries the operation. 2126*02ff05a9Svsakar */ 2127*02ff05a9Svsakar ufs_tryirwlock(&ip->i_rwlock, RW_WRITER, retry_file); 2128*02ff05a9Svsakar if (indeadlock) { 2129*02ff05a9Svsakar if (ulp) 2130*02ff05a9Svsakar ufs_lockfs_end(ulp); 2131*02ff05a9Svsakar goto again; 2132*02ff05a9Svsakar } 21337c478bd9Sstevel@tonic-gate dorwlock = 1; 21347c478bd9Sstevel@tonic-gate } 21357c478bd9Sstevel@tonic-gate 21367c478bd9Sstevel@tonic-gate /* 21377c478bd9Sstevel@tonic-gate * Truncate file. Must have write permission and not be a directory. 21387c478bd9Sstevel@tonic-gate */ 21397c478bd9Sstevel@tonic-gate if (mask & AT_SIZE) { 21407c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_WRITER); 21417c478bd9Sstevel@tonic-gate if (vp->v_type == VDIR) { 21427c478bd9Sstevel@tonic-gate error = EISDIR; 21437c478bd9Sstevel@tonic-gate goto update_inode; 21447c478bd9Sstevel@tonic-gate } 21457c478bd9Sstevel@tonic-gate if (error = ufs_iaccess(ip, IWRITE, cr)) 21467c478bd9Sstevel@tonic-gate goto update_inode; 21477c478bd9Sstevel@tonic-gate 21487c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 21497c478bd9Sstevel@tonic-gate error = TRANS_ITRUNC(ip, vap->va_size, 0, cr); 21507c478bd9Sstevel@tonic-gate if (error) { 21517c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_WRITER); 21527c478bd9Sstevel@tonic-gate goto update_inode; 21537c478bd9Sstevel@tonic-gate } 21547c478bd9Sstevel@tonic-gate } 21557c478bd9Sstevel@tonic-gate 21567c478bd9Sstevel@tonic-gate if (ulp) { 21577c478bd9Sstevel@tonic-gate trans_size = (int)TOP_SETATTR_SIZE(ip); 21587c478bd9Sstevel@tonic-gate TRANS_BEGIN_CSYNC(ufsvfsp, issync, TOP_SETATTR, trans_size); 21597c478bd9Sstevel@tonic-gate ++dotrans; 21607c478bd9Sstevel@tonic-gate } 21617c478bd9Sstevel@tonic-gate 21627c478bd9Sstevel@tonic-gate /* 21637c478bd9Sstevel@tonic-gate * Acquire i_rwlock after TRANS_BEGIN_CSYNC() if this is a directory. 21647c478bd9Sstevel@tonic-gate * This follows the protocol established by 21657c478bd9Sstevel@tonic-gate * ufs_link/create/remove/rename/mkdir/rmdir/symlink. 21667c478bd9Sstevel@tonic-gate */ 21677c478bd9Sstevel@tonic-gate if (vp->v_type == VDIR) { 2168*02ff05a9Svsakar ufs_tryirwlock_trans(&ip->i_rwlock, RW_WRITER, TOP_SETATTR, 2169*02ff05a9Svsakar retry_dir); 2170*02ff05a9Svsakar if (indeadlock) 2171*02ff05a9Svsakar goto again; 21727c478bd9Sstevel@tonic-gate dorwlock = 1; 21737c478bd9Sstevel@tonic-gate } 21747c478bd9Sstevel@tonic-gate 21757c478bd9Sstevel@tonic-gate /* 21767c478bd9Sstevel@tonic-gate * Grab quota lock if we are changing the file's owner. 21777c478bd9Sstevel@tonic-gate */ 21787c478bd9Sstevel@tonic-gate if (mask & AT_UID) { 21797c478bd9Sstevel@tonic-gate rw_enter(&ufsvfsp->vfs_dqrwlock, RW_READER); 21807c478bd9Sstevel@tonic-gate dodqlock = 1; 21817c478bd9Sstevel@tonic-gate } 21827c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_WRITER); 21837c478bd9Sstevel@tonic-gate 21847c478bd9Sstevel@tonic-gate oldva.va_mode = ip->i_mode; 21857c478bd9Sstevel@tonic-gate oldva.va_uid = ip->i_uid; 21867c478bd9Sstevel@tonic-gate oldva.va_gid = ip->i_gid; 21877c478bd9Sstevel@tonic-gate 21887c478bd9Sstevel@tonic-gate vap->va_mask &= ~AT_SIZE; 21897c478bd9Sstevel@tonic-gate /* 21907c478bd9Sstevel@tonic-gate * ufs_iaccess is "close enough"; that's because it doesn't 21917c478bd9Sstevel@tonic-gate * map the defines. 21927c478bd9Sstevel@tonic-gate */ 21937c478bd9Sstevel@tonic-gate error = secpolicy_vnode_setattr(cr, vp, vap, &oldva, flags, 21947c478bd9Sstevel@tonic-gate ufs_iaccess, ip); 21957c478bd9Sstevel@tonic-gate if (error) 21967c478bd9Sstevel@tonic-gate goto update_inode; 21977c478bd9Sstevel@tonic-gate 21987c478bd9Sstevel@tonic-gate mask = vap->va_mask; 21997c478bd9Sstevel@tonic-gate 22007c478bd9Sstevel@tonic-gate /* 22017c478bd9Sstevel@tonic-gate * Change file access modes. 22027c478bd9Sstevel@tonic-gate */ 22037c478bd9Sstevel@tonic-gate if (mask & AT_MODE) { 22047c478bd9Sstevel@tonic-gate ip->i_mode = (ip->i_mode & IFMT) | (vap->va_mode & ~IFMT); 22057c478bd9Sstevel@tonic-gate TRANS_INODE(ufsvfsp, ip); 22067c478bd9Sstevel@tonic-gate ip->i_flag |= ICHG; 22077c478bd9Sstevel@tonic-gate if (stickyhack) { 22087c478bd9Sstevel@tonic-gate mutex_enter(&vp->v_lock); 22097c478bd9Sstevel@tonic-gate if ((ip->i_mode & (ISVTX | IEXEC | IFDIR)) == ISVTX) 22107c478bd9Sstevel@tonic-gate vp->v_flag |= VSWAPLIKE; 22117c478bd9Sstevel@tonic-gate else 22127c478bd9Sstevel@tonic-gate vp->v_flag &= ~VSWAPLIKE; 22137c478bd9Sstevel@tonic-gate mutex_exit(&vp->v_lock); 22147c478bd9Sstevel@tonic-gate } 22157c478bd9Sstevel@tonic-gate } 22167c478bd9Sstevel@tonic-gate if (mask & (AT_UID|AT_GID)) { 22177c478bd9Sstevel@tonic-gate if (mask & AT_UID) { 22187c478bd9Sstevel@tonic-gate /* 22197c478bd9Sstevel@tonic-gate * Don't change ownership of the quota inode. 22207c478bd9Sstevel@tonic-gate */ 22217c478bd9Sstevel@tonic-gate if (ufsvfsp->vfs_qinod == ip) { 22227c478bd9Sstevel@tonic-gate ASSERT(ufsvfsp->vfs_qflags & MQ_ENABLED); 22237c478bd9Sstevel@tonic-gate error = EINVAL; 22247c478bd9Sstevel@tonic-gate goto update_inode; 22257c478bd9Sstevel@tonic-gate } 22267c478bd9Sstevel@tonic-gate 22277c478bd9Sstevel@tonic-gate /* 22287c478bd9Sstevel@tonic-gate * No real ownership change. 22297c478bd9Sstevel@tonic-gate */ 22307c478bd9Sstevel@tonic-gate if (ip->i_uid == vap->va_uid) { 22317c478bd9Sstevel@tonic-gate blocks = 0; 22327c478bd9Sstevel@tonic-gate owner_change = 0; 22337c478bd9Sstevel@tonic-gate } 22347c478bd9Sstevel@tonic-gate /* 22357c478bd9Sstevel@tonic-gate * Remove the blocks and the file, from the old user's 22367c478bd9Sstevel@tonic-gate * quota. 22377c478bd9Sstevel@tonic-gate */ 22387c478bd9Sstevel@tonic-gate else { 22397c478bd9Sstevel@tonic-gate blocks = ip->i_blocks; 22407c478bd9Sstevel@tonic-gate owner_change = 1; 22417c478bd9Sstevel@tonic-gate 22427c478bd9Sstevel@tonic-gate (void) chkdq(ip, -blocks, /* force */ 1, cr, 22437c478bd9Sstevel@tonic-gate (char **)NULL, (size_t *)NULL); 22447c478bd9Sstevel@tonic-gate (void) chkiq(ufsvfsp, /* change */ -1, ip, 22457c478bd9Sstevel@tonic-gate (uid_t)ip->i_uid, 22467c478bd9Sstevel@tonic-gate /* force */ 1, cr, 22477c478bd9Sstevel@tonic-gate (char **)NULL, (size_t *)NULL); 22487c478bd9Sstevel@tonic-gate dqrele(ip->i_dquot); 22497c478bd9Sstevel@tonic-gate } 22507c478bd9Sstevel@tonic-gate 22517c478bd9Sstevel@tonic-gate ip->i_uid = vap->va_uid; 22527c478bd9Sstevel@tonic-gate 22537c478bd9Sstevel@tonic-gate /* 22547c478bd9Sstevel@tonic-gate * There is a real ownership change. 22557c478bd9Sstevel@tonic-gate */ 22567c478bd9Sstevel@tonic-gate if (owner_change) { 22577c478bd9Sstevel@tonic-gate /* 22587c478bd9Sstevel@tonic-gate * Add the blocks and the file to the new 22597c478bd9Sstevel@tonic-gate * user's quota. 22607c478bd9Sstevel@tonic-gate */ 22617c478bd9Sstevel@tonic-gate ip->i_dquot = getinoquota(ip); 22627c478bd9Sstevel@tonic-gate (void) chkdq(ip, blocks, /* force */ 1, cr, 22637c478bd9Sstevel@tonic-gate &errmsg1, &len1); 22647c478bd9Sstevel@tonic-gate (void) chkiq(ufsvfsp, /* change */ 1, 22657c478bd9Sstevel@tonic-gate (struct inode *)NULL, 22667c478bd9Sstevel@tonic-gate (uid_t)ip->i_uid, 22677c478bd9Sstevel@tonic-gate /* force */ 1, cr, 22687c478bd9Sstevel@tonic-gate &errmsg2, &len2); 22697c478bd9Sstevel@tonic-gate } 22707c478bd9Sstevel@tonic-gate } 22717c478bd9Sstevel@tonic-gate if (mask & AT_GID) { 22727c478bd9Sstevel@tonic-gate ip->i_gid = vap->va_gid; 22737c478bd9Sstevel@tonic-gate } 22747c478bd9Sstevel@tonic-gate TRANS_INODE(ufsvfsp, ip); 22757c478bd9Sstevel@tonic-gate ip->i_flag |= ICHG; 22767c478bd9Sstevel@tonic-gate } 22777c478bd9Sstevel@tonic-gate /* 22787c478bd9Sstevel@tonic-gate * Change file access or modified times. 22797c478bd9Sstevel@tonic-gate */ 22807c478bd9Sstevel@tonic-gate if (mask & (AT_ATIME|AT_MTIME)) { 22817c478bd9Sstevel@tonic-gate /* Check that the time value is within ufs range */ 22827c478bd9Sstevel@tonic-gate if (((mask & AT_ATIME) && TIMESPEC_OVERFLOW(&vap->va_atime)) || 22837c478bd9Sstevel@tonic-gate ((mask & AT_MTIME) && TIMESPEC_OVERFLOW(&vap->va_mtime))) { 22847c478bd9Sstevel@tonic-gate error = EOVERFLOW; 22857c478bd9Sstevel@tonic-gate goto update_inode; 22867c478bd9Sstevel@tonic-gate } 22877c478bd9Sstevel@tonic-gate 22887c478bd9Sstevel@tonic-gate /* 22897c478bd9Sstevel@tonic-gate * if the "noaccess" mount option is set and only atime 22907c478bd9Sstevel@tonic-gate * update is requested, do nothing. No error is returned. 22917c478bd9Sstevel@tonic-gate */ 22927c478bd9Sstevel@tonic-gate if ((ufsvfsp->vfs_noatime) && 22937c478bd9Sstevel@tonic-gate ((mask & (AT_ATIME|AT_MTIME)) == AT_ATIME)) 22947c478bd9Sstevel@tonic-gate goto skip_atime; 22957c478bd9Sstevel@tonic-gate 22967c478bd9Sstevel@tonic-gate if (mask & AT_ATIME) { 22977c478bd9Sstevel@tonic-gate ip->i_atime.tv_sec = vap->va_atime.tv_sec; 22987c478bd9Sstevel@tonic-gate ip->i_atime.tv_usec = vap->va_atime.tv_nsec / 1000; 22997c478bd9Sstevel@tonic-gate ip->i_flag &= ~IACC; 23007c478bd9Sstevel@tonic-gate } 23017c478bd9Sstevel@tonic-gate if (mask & AT_MTIME) { 23027c478bd9Sstevel@tonic-gate ip->i_mtime.tv_sec = vap->va_mtime.tv_sec; 23037c478bd9Sstevel@tonic-gate ip->i_mtime.tv_usec = vap->va_mtime.tv_nsec / 1000; 23047c478bd9Sstevel@tonic-gate gethrestime(&now); 23057c478bd9Sstevel@tonic-gate if (now.tv_sec > TIME32_MAX) { 23067c478bd9Sstevel@tonic-gate /* 23077c478bd9Sstevel@tonic-gate * In 2038, ctime sticks forever.. 23087c478bd9Sstevel@tonic-gate */ 23097c478bd9Sstevel@tonic-gate ip->i_ctime.tv_sec = TIME32_MAX; 23107c478bd9Sstevel@tonic-gate ip->i_ctime.tv_usec = 0; 23117c478bd9Sstevel@tonic-gate } else { 23127c478bd9Sstevel@tonic-gate ip->i_ctime.tv_sec = now.tv_sec; 23137c478bd9Sstevel@tonic-gate ip->i_ctime.tv_usec = now.tv_nsec / 1000; 23147c478bd9Sstevel@tonic-gate } 23157c478bd9Sstevel@tonic-gate ip->i_flag &= ~(IUPD|ICHG); 23167c478bd9Sstevel@tonic-gate ip->i_flag |= IMODTIME; 23177c478bd9Sstevel@tonic-gate } 23187c478bd9Sstevel@tonic-gate TRANS_INODE(ufsvfsp, ip); 23197c478bd9Sstevel@tonic-gate ip->i_flag |= IMOD; 23207c478bd9Sstevel@tonic-gate } 23217c478bd9Sstevel@tonic-gate 23227c478bd9Sstevel@tonic-gate skip_atime: 23237c478bd9Sstevel@tonic-gate /* 23247c478bd9Sstevel@tonic-gate * The presence of a shadow inode may indicate an ACL, but does 23257c478bd9Sstevel@tonic-gate * not imply an ACL. Future FSD types should be handled here too 23267c478bd9Sstevel@tonic-gate * and check for the presence of the attribute-specific data 23277c478bd9Sstevel@tonic-gate * before referencing it. 23287c478bd9Sstevel@tonic-gate */ 23297c478bd9Sstevel@tonic-gate if (ip->i_shadow) { 23307c478bd9Sstevel@tonic-gate /* 23317c478bd9Sstevel@tonic-gate * XXX if ufs_iupdat is changed to sandbagged write fix 23327c478bd9Sstevel@tonic-gate * ufs_acl_setattr to push ip to keep acls consistent 23337c478bd9Sstevel@tonic-gate * 23347c478bd9Sstevel@tonic-gate * Suppress out of inodes messages if we will retry. 23357c478bd9Sstevel@tonic-gate */ 23367c478bd9Sstevel@tonic-gate if (retry) 23377c478bd9Sstevel@tonic-gate ip->i_flag |= IQUIET; 23387c478bd9Sstevel@tonic-gate error = ufs_acl_setattr(ip, vap, cr); 23397c478bd9Sstevel@tonic-gate ip->i_flag &= ~IQUIET; 23407c478bd9Sstevel@tonic-gate } 23417c478bd9Sstevel@tonic-gate 23427c478bd9Sstevel@tonic-gate update_inode: 23437c478bd9Sstevel@tonic-gate /* 23447c478bd9Sstevel@tonic-gate * Setattr always increases the sequence number 23457c478bd9Sstevel@tonic-gate */ 23467c478bd9Sstevel@tonic-gate ip->i_seq++; 23477c478bd9Sstevel@tonic-gate 23487c478bd9Sstevel@tonic-gate /* 23497c478bd9Sstevel@tonic-gate * if nfsd and not logging; push synchronously 23507c478bd9Sstevel@tonic-gate */ 23517c478bd9Sstevel@tonic-gate if ((curthread->t_flag & T_DONTPEND) && !TRANS_ISTRANS(ufsvfsp)) { 23527c478bd9Sstevel@tonic-gate ufs_iupdat(ip, 1); 23537c478bd9Sstevel@tonic-gate } else { 23547c478bd9Sstevel@tonic-gate ITIMES_NOLOCK(ip); 23557c478bd9Sstevel@tonic-gate } 23567c478bd9Sstevel@tonic-gate 23577c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 23587c478bd9Sstevel@tonic-gate if (dodqlock) { 23597c478bd9Sstevel@tonic-gate rw_exit(&ufsvfsp->vfs_dqrwlock); 23607c478bd9Sstevel@tonic-gate } 23617c478bd9Sstevel@tonic-gate if (dorwlock) 23627c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 23637c478bd9Sstevel@tonic-gate 23647c478bd9Sstevel@tonic-gate if (ulp) { 23657c478bd9Sstevel@tonic-gate if (dotrans) { 23667c478bd9Sstevel@tonic-gate int terr = 0; 23677c478bd9Sstevel@tonic-gate TRANS_END_CSYNC(ufsvfsp, terr, issync, TOP_SETATTR, 23687c478bd9Sstevel@tonic-gate trans_size); 23697c478bd9Sstevel@tonic-gate if (error == 0) 23707c478bd9Sstevel@tonic-gate error = terr; 23717c478bd9Sstevel@tonic-gate } 23727c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 23737c478bd9Sstevel@tonic-gate } 23747c478bd9Sstevel@tonic-gate out: 23757c478bd9Sstevel@tonic-gate /* 23767c478bd9Sstevel@tonic-gate * If out of inodes or blocks, see if we can free something 23777c478bd9Sstevel@tonic-gate * up from the delete queue. 23787c478bd9Sstevel@tonic-gate */ 23797c478bd9Sstevel@tonic-gate if ((error == ENOSPC) && retry && TRANS_ISTRANS(ufsvfsp)) { 23807c478bd9Sstevel@tonic-gate ufs_delete_drain_wait(ufsvfsp, 1); 23817c478bd9Sstevel@tonic-gate retry = 0; 23827c478bd9Sstevel@tonic-gate if (errmsg1 != NULL) 23837c478bd9Sstevel@tonic-gate kmem_free(errmsg1, len1); 23847c478bd9Sstevel@tonic-gate if (errmsg2 != NULL) 23857c478bd9Sstevel@tonic-gate kmem_free(errmsg2, len2); 23867c478bd9Sstevel@tonic-gate goto again; 23877c478bd9Sstevel@tonic-gate } 23887c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_SETATTR_END, 23897c478bd9Sstevel@tonic-gate "ufs_setattr_end:vp %p error %d", vp, error); 23907c478bd9Sstevel@tonic-gate if (errmsg1 != NULL) { 23917c478bd9Sstevel@tonic-gate uprintf(errmsg1); 23927c478bd9Sstevel@tonic-gate kmem_free(errmsg1, len1); 23937c478bd9Sstevel@tonic-gate } 23947c478bd9Sstevel@tonic-gate if (errmsg2 != NULL) { 23957c478bd9Sstevel@tonic-gate uprintf(errmsg2); 23967c478bd9Sstevel@tonic-gate kmem_free(errmsg2, len2); 23977c478bd9Sstevel@tonic-gate } 23987c478bd9Sstevel@tonic-gate return (error); 23997c478bd9Sstevel@tonic-gate } 24007c478bd9Sstevel@tonic-gate 24017c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 24027c478bd9Sstevel@tonic-gate static int 24037c478bd9Sstevel@tonic-gate ufs_access(struct vnode *vp, int mode, int flags, struct cred *cr) 24047c478bd9Sstevel@tonic-gate { 24057c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 24067c478bd9Sstevel@tonic-gate int error; 24077c478bd9Sstevel@tonic-gate 24087c478bd9Sstevel@tonic-gate TRACE_3(TR_FAC_UFS, TR_UFS_ACCESS_START, 24097c478bd9Sstevel@tonic-gate "ufs_access_start:vp %p mode %x flags %x", vp, mode, flags); 24107c478bd9Sstevel@tonic-gate 24117c478bd9Sstevel@tonic-gate if (ip->i_ufsvfs == NULL) 24127c478bd9Sstevel@tonic-gate return (EIO); 24137c478bd9Sstevel@tonic-gate 24147c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_READER); 24157c478bd9Sstevel@tonic-gate 24167c478bd9Sstevel@tonic-gate /* 24177c478bd9Sstevel@tonic-gate * The ufs_iaccess function wants to be called with 24187c478bd9Sstevel@tonic-gate * mode bits expressed as "ufs specific" bits. 24197c478bd9Sstevel@tonic-gate * I.e., VWRITE|VREAD|VEXEC do not make sense to 24207c478bd9Sstevel@tonic-gate * ufs_iaccess() but IWRITE|IREAD|IEXEC do. 24217c478bd9Sstevel@tonic-gate * But since they're the same we just pass the vnode mode 24227c478bd9Sstevel@tonic-gate * bit but just verify that assumption at compile time. 24237c478bd9Sstevel@tonic-gate */ 24247c478bd9Sstevel@tonic-gate #if IWRITE != VWRITE || IREAD != VREAD || IEXEC != VEXEC 24257c478bd9Sstevel@tonic-gate #error "ufs_access needs to map Vmodes to Imodes" 24267c478bd9Sstevel@tonic-gate #endif 24277c478bd9Sstevel@tonic-gate error = ufs_iaccess(ip, mode, cr); 24287c478bd9Sstevel@tonic-gate 24297c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 24307c478bd9Sstevel@tonic-gate 24317c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_ACCESS_END, 24327c478bd9Sstevel@tonic-gate "ufs_access_end:vp %p error %d", vp, error); 24337c478bd9Sstevel@tonic-gate return (error); 24347c478bd9Sstevel@tonic-gate } 24357c478bd9Sstevel@tonic-gate 24367c478bd9Sstevel@tonic-gate /* ARGSUSED */ 24377c478bd9Sstevel@tonic-gate static int 24387c478bd9Sstevel@tonic-gate ufs_readlink(struct vnode *vp, struct uio *uiop, struct cred *cr) 24397c478bd9Sstevel@tonic-gate { 24407c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 24417c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp; 24427c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 24437c478bd9Sstevel@tonic-gate int error; 24447c478bd9Sstevel@tonic-gate int fastsymlink; 24457c478bd9Sstevel@tonic-gate 24467c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_READLINK_START, 24477c478bd9Sstevel@tonic-gate "ufs_readlink_start:vp %p uiop %p", uiop, vp); 24487c478bd9Sstevel@tonic-gate 24497c478bd9Sstevel@tonic-gate if (vp->v_type != VLNK) { 24507c478bd9Sstevel@tonic-gate error = EINVAL; 24517c478bd9Sstevel@tonic-gate goto nolockout; 24527c478bd9Sstevel@tonic-gate } 24537c478bd9Sstevel@tonic-gate 24547c478bd9Sstevel@tonic-gate /* 24557c478bd9Sstevel@tonic-gate * If the symbolic link is empty there is nothing to read. 24567c478bd9Sstevel@tonic-gate * Fast-track these empty symbolic links 24577c478bd9Sstevel@tonic-gate */ 24587c478bd9Sstevel@tonic-gate if (ip->i_size == 0) { 24597c478bd9Sstevel@tonic-gate error = 0; 24607c478bd9Sstevel@tonic-gate goto nolockout; 24617c478bd9Sstevel@tonic-gate } 24627c478bd9Sstevel@tonic-gate 24637c478bd9Sstevel@tonic-gate ufsvfsp = ip->i_ufsvfs; 24647c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_READLINK_MASK); 24657c478bd9Sstevel@tonic-gate if (error) 24667c478bd9Sstevel@tonic-gate goto nolockout; 24677c478bd9Sstevel@tonic-gate /* 24687c478bd9Sstevel@tonic-gate * The ip->i_rwlock protects the data blocks used for FASTSYMLINK 24697c478bd9Sstevel@tonic-gate */ 24707c478bd9Sstevel@tonic-gate again: 24717c478bd9Sstevel@tonic-gate fastsymlink = 0; 24727c478bd9Sstevel@tonic-gate if (ip->i_flag & IFASTSYMLNK) { 24737c478bd9Sstevel@tonic-gate rw_enter(&ip->i_rwlock, RW_READER); 24747c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_READER); 24757c478bd9Sstevel@tonic-gate if (ip->i_flag & IFASTSYMLNK) { 24767c478bd9Sstevel@tonic-gate if (!ULOCKFS_IS_NOIACC(ITOUL(ip)) && 24777c478bd9Sstevel@tonic-gate (ip->i_fs->fs_ronly == 0) && 24787c478bd9Sstevel@tonic-gate (!ufsvfsp->vfs_noatime)) { 24797c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 24807c478bd9Sstevel@tonic-gate ip->i_flag |= IACC; 24817c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 24827c478bd9Sstevel@tonic-gate } 24837c478bd9Sstevel@tonic-gate error = uiomove((caddr_t)&ip->i_db[1], 24847c478bd9Sstevel@tonic-gate MIN(ip->i_size, uiop->uio_resid), 24857c478bd9Sstevel@tonic-gate UIO_READ, uiop); 24867c478bd9Sstevel@tonic-gate ITIMES(ip); 24877c478bd9Sstevel@tonic-gate ++fastsymlink; 24887c478bd9Sstevel@tonic-gate } 24897c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 24907c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 24917c478bd9Sstevel@tonic-gate } 24927c478bd9Sstevel@tonic-gate if (!fastsymlink) { 24937c478bd9Sstevel@tonic-gate ssize_t size; /* number of bytes read */ 24947c478bd9Sstevel@tonic-gate caddr_t basep; /* pointer to input data */ 24957c478bd9Sstevel@tonic-gate ino_t ino; 24967c478bd9Sstevel@tonic-gate long igen; 24977c478bd9Sstevel@tonic-gate struct uio tuio; /* temp uio struct */ 24987c478bd9Sstevel@tonic-gate struct uio *tuiop; 24997c478bd9Sstevel@tonic-gate iovec_t tiov; /* temp iovec struct */ 25007c478bd9Sstevel@tonic-gate char kbuf[FSL_SIZE]; /* buffer to hold fast symlink */ 25017c478bd9Sstevel@tonic-gate int tflag = 0; /* flag to indicate temp vars used */ 25027c478bd9Sstevel@tonic-gate 25037c478bd9Sstevel@tonic-gate ino = ip->i_number; 25047c478bd9Sstevel@tonic-gate igen = ip->i_gen; 25057c478bd9Sstevel@tonic-gate size = uiop->uio_resid; 25067c478bd9Sstevel@tonic-gate basep = uiop->uio_iov->iov_base; 25077c478bd9Sstevel@tonic-gate tuiop = uiop; 25087c478bd9Sstevel@tonic-gate 25097c478bd9Sstevel@tonic-gate rw_enter(&ip->i_rwlock, RW_WRITER); 25107c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_WRITER); 25117c478bd9Sstevel@tonic-gate if (ip->i_flag & IFASTSYMLNK) { 25127c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 25137c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 25147c478bd9Sstevel@tonic-gate goto again; 25157c478bd9Sstevel@tonic-gate } 25167c478bd9Sstevel@tonic-gate 25177c478bd9Sstevel@tonic-gate /* can this be a fast symlink and is it a user buffer? */ 25187c478bd9Sstevel@tonic-gate if (ip->i_size <= FSL_SIZE && 25197c478bd9Sstevel@tonic-gate (uiop->uio_segflg == UIO_USERSPACE || 25207c478bd9Sstevel@tonic-gate uiop->uio_segflg == UIO_USERISPACE)) { 25217c478bd9Sstevel@tonic-gate 25227c478bd9Sstevel@tonic-gate bzero(&tuio, sizeof (struct uio)); 25237c478bd9Sstevel@tonic-gate /* 25247c478bd9Sstevel@tonic-gate * setup a kernel buffer to read link into. this 25257c478bd9Sstevel@tonic-gate * is to fix a race condition where the user buffer 25267c478bd9Sstevel@tonic-gate * got corrupted before copying it into the inode. 25277c478bd9Sstevel@tonic-gate */ 25287c478bd9Sstevel@tonic-gate size = ip->i_size; 25297c478bd9Sstevel@tonic-gate tiov.iov_len = size; 25307c478bd9Sstevel@tonic-gate tiov.iov_base = kbuf; 25317c478bd9Sstevel@tonic-gate tuio.uio_iov = &tiov; 25327c478bd9Sstevel@tonic-gate tuio.uio_iovcnt = 1; 25337c478bd9Sstevel@tonic-gate tuio.uio_offset = uiop->uio_offset; 25347c478bd9Sstevel@tonic-gate tuio.uio_segflg = UIO_SYSSPACE; 25357c478bd9Sstevel@tonic-gate tuio.uio_fmode = uiop->uio_fmode; 25367c478bd9Sstevel@tonic-gate tuio.uio_extflg = uiop->uio_extflg; 25377c478bd9Sstevel@tonic-gate tuio.uio_limit = uiop->uio_limit; 25387c478bd9Sstevel@tonic-gate tuio.uio_resid = size; 25397c478bd9Sstevel@tonic-gate 25407c478bd9Sstevel@tonic-gate basep = tuio.uio_iov->iov_base; 25417c478bd9Sstevel@tonic-gate tuiop = &tuio; 25427c478bd9Sstevel@tonic-gate tflag = 1; 25437c478bd9Sstevel@tonic-gate } 25447c478bd9Sstevel@tonic-gate 25457c478bd9Sstevel@tonic-gate error = rdip(ip, tuiop, 0, cr); 25467c478bd9Sstevel@tonic-gate if (!(error == 0 && ip->i_number == ino && ip->i_gen == igen)) { 25477c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 25487c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 25497c478bd9Sstevel@tonic-gate goto out; 25507c478bd9Sstevel@tonic-gate } 25517c478bd9Sstevel@tonic-gate 25527c478bd9Sstevel@tonic-gate if (tflag == 0) 25537c478bd9Sstevel@tonic-gate size -= uiop->uio_resid; 25547c478bd9Sstevel@tonic-gate 25557c478bd9Sstevel@tonic-gate if ((tflag == 0 && ip->i_size <= FSL_SIZE && 25567c478bd9Sstevel@tonic-gate ip->i_size == size) || (tflag == 1 && 25577c478bd9Sstevel@tonic-gate tuio.uio_resid == 0)) { 25587c478bd9Sstevel@tonic-gate error = kcopy(basep, &ip->i_db[1], ip->i_size); 25597c478bd9Sstevel@tonic-gate if (error == 0) { 25607c478bd9Sstevel@tonic-gate ip->i_flag |= IFASTSYMLNK; 25617c478bd9Sstevel@tonic-gate /* 25627c478bd9Sstevel@tonic-gate * free page 25637c478bd9Sstevel@tonic-gate */ 25647c478bd9Sstevel@tonic-gate (void) VOP_PUTPAGE(ITOV(ip), 25657c478bd9Sstevel@tonic-gate (offset_t)0, PAGESIZE, 25667c478bd9Sstevel@tonic-gate (B_DONTNEED | B_FREE | B_FORCE | B_ASYNC), 25677c478bd9Sstevel@tonic-gate cr); 25687c478bd9Sstevel@tonic-gate } else { 25697c478bd9Sstevel@tonic-gate int i; 25707c478bd9Sstevel@tonic-gate /* error, clear garbage left behind */ 25717c478bd9Sstevel@tonic-gate for (i = 1; i < NDADDR; i++) 25727c478bd9Sstevel@tonic-gate ip->i_db[i] = 0; 25737c478bd9Sstevel@tonic-gate for (i = 0; i < NIADDR; i++) 25747c478bd9Sstevel@tonic-gate ip->i_ib[i] = 0; 25757c478bd9Sstevel@tonic-gate } 25767c478bd9Sstevel@tonic-gate } 25777c478bd9Sstevel@tonic-gate if (tflag == 1) { 25787c478bd9Sstevel@tonic-gate /* now, copy it into the user buffer */ 25797c478bd9Sstevel@tonic-gate error = uiomove((caddr_t)kbuf, 25807c478bd9Sstevel@tonic-gate MIN(size, uiop->uio_resid), 25817c478bd9Sstevel@tonic-gate UIO_READ, uiop); 25827c478bd9Sstevel@tonic-gate } 25837c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 25847c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 25857c478bd9Sstevel@tonic-gate } 25867c478bd9Sstevel@tonic-gate out: 25877c478bd9Sstevel@tonic-gate if (ulp) { 25887c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 25897c478bd9Sstevel@tonic-gate } 25907c478bd9Sstevel@tonic-gate nolockout: 25917c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_READLINK_END, 25927c478bd9Sstevel@tonic-gate "ufs_readlink_end:vp %p error %d", vp, error); 25937c478bd9Sstevel@tonic-gate 25947c478bd9Sstevel@tonic-gate return (error); 25957c478bd9Sstevel@tonic-gate } 25967c478bd9Sstevel@tonic-gate 25977c478bd9Sstevel@tonic-gate /* ARGSUSED */ 25987c478bd9Sstevel@tonic-gate static int 25997c478bd9Sstevel@tonic-gate ufs_fsync(struct vnode *vp, int syncflag, struct cred *cr) 26007c478bd9Sstevel@tonic-gate { 26017c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 26027c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp = ip->i_ufsvfs; 26037c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 26047c478bd9Sstevel@tonic-gate int error; 26057c478bd9Sstevel@tonic-gate 26067c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_FSYNC_START, 26077c478bd9Sstevel@tonic-gate "ufs_fsync_start:vp %p", vp); 26087c478bd9Sstevel@tonic-gate 26097c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_FSYNC_MASK); 26107c478bd9Sstevel@tonic-gate if (error) 26117c478bd9Sstevel@tonic-gate return (error); 26127c478bd9Sstevel@tonic-gate 26137c478bd9Sstevel@tonic-gate if (TRANS_ISTRANS(ufsvfsp)) { 26147c478bd9Sstevel@tonic-gate /* 26157c478bd9Sstevel@tonic-gate * First push out any data pages 26167c478bd9Sstevel@tonic-gate */ 26177c478bd9Sstevel@tonic-gate if (vn_has_cached_data(vp) && !(syncflag & FNODSYNC) && 26187c478bd9Sstevel@tonic-gate (vp->v_type != VCHR) && !(IS_SWAPVP(vp))) { 26197c478bd9Sstevel@tonic-gate error = VOP_PUTPAGE(vp, (offset_t)0, (size_t)0, 26207c478bd9Sstevel@tonic-gate 0, CRED()); 26217c478bd9Sstevel@tonic-gate if (error) 26227c478bd9Sstevel@tonic-gate goto out; 26237c478bd9Sstevel@tonic-gate } 26247c478bd9Sstevel@tonic-gate 26257c478bd9Sstevel@tonic-gate /* 26267c478bd9Sstevel@tonic-gate * Delta any delayed inode times updates 26277c478bd9Sstevel@tonic-gate * and push inode to log. 26287c478bd9Sstevel@tonic-gate * All other inode deltas will have already been delta'd 26297c478bd9Sstevel@tonic-gate * and will be pushed during the commit. 26307c478bd9Sstevel@tonic-gate */ 26317c478bd9Sstevel@tonic-gate if (!(syncflag & FDSYNC) && 26327c478bd9Sstevel@tonic-gate ((ip->i_flag & (IMOD|IMODACC)) == IMODACC)) { 26337c478bd9Sstevel@tonic-gate if (ulp) { 26347c478bd9Sstevel@tonic-gate TRANS_BEGIN_ASYNC(ufsvfsp, TOP_FSYNC, 26357c478bd9Sstevel@tonic-gate TOP_SYNCIP_SIZE); 26367c478bd9Sstevel@tonic-gate } 26377c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_READER); 26387c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 26397c478bd9Sstevel@tonic-gate ip->i_flag &= ~IMODTIME; 26407c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 26417c478bd9Sstevel@tonic-gate ufs_iupdat(ip, I_SYNC); 26427c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 26437c478bd9Sstevel@tonic-gate if (ulp) { 26447c478bd9Sstevel@tonic-gate TRANS_END_ASYNC(ufsvfsp, TOP_FSYNC, 26457c478bd9Sstevel@tonic-gate TOP_SYNCIP_SIZE); 26467c478bd9Sstevel@tonic-gate } 26477c478bd9Sstevel@tonic-gate } 26487c478bd9Sstevel@tonic-gate 26497c478bd9Sstevel@tonic-gate /* 26507c478bd9Sstevel@tonic-gate * Commit the Moby transaction 26517c478bd9Sstevel@tonic-gate * 26527c478bd9Sstevel@tonic-gate * Deltas have already been made so we just need to 26537c478bd9Sstevel@tonic-gate * commit them with a synchronous transaction. 26547c478bd9Sstevel@tonic-gate * TRANS_BEGIN_SYNC() will return an error 26557c478bd9Sstevel@tonic-gate * if there are no deltas to commit, for an 26567c478bd9Sstevel@tonic-gate * empty transaction. 26577c478bd9Sstevel@tonic-gate */ 26587c478bd9Sstevel@tonic-gate if (ulp) { 26597c478bd9Sstevel@tonic-gate TRANS_BEGIN_SYNC(ufsvfsp, TOP_FSYNC, TOP_COMMIT_SIZE, 26607c478bd9Sstevel@tonic-gate error); 26617c478bd9Sstevel@tonic-gate if (error) { 26627c478bd9Sstevel@tonic-gate error = 0; /* commit wasn't needed */ 26637c478bd9Sstevel@tonic-gate goto out; 26647c478bd9Sstevel@tonic-gate } 26657c478bd9Sstevel@tonic-gate TRANS_END_SYNC(ufsvfsp, error, TOP_FSYNC, 26667c478bd9Sstevel@tonic-gate TOP_COMMIT_SIZE); 26677c478bd9Sstevel@tonic-gate } 26687c478bd9Sstevel@tonic-gate } else { /* not logging */ 26697c478bd9Sstevel@tonic-gate if (!(IS_SWAPVP(vp))) 26707c478bd9Sstevel@tonic-gate if (syncflag & FNODSYNC) { 26717c478bd9Sstevel@tonic-gate /* Just update the inode only */ 26727c478bd9Sstevel@tonic-gate TRANS_IUPDAT(ip, 1); 26737c478bd9Sstevel@tonic-gate error = 0; 26747c478bd9Sstevel@tonic-gate } else if (syncflag & FDSYNC) 26757c478bd9Sstevel@tonic-gate /* Do data-synchronous writes */ 26767c478bd9Sstevel@tonic-gate error = TRANS_SYNCIP(ip, 0, I_DSYNC, TOP_FSYNC); 26777c478bd9Sstevel@tonic-gate else 26787c478bd9Sstevel@tonic-gate /* Do synchronous writes */ 26797c478bd9Sstevel@tonic-gate error = TRANS_SYNCIP(ip, 0, I_SYNC, TOP_FSYNC); 26807c478bd9Sstevel@tonic-gate 26817c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_WRITER); 26827c478bd9Sstevel@tonic-gate if (!error) 26837c478bd9Sstevel@tonic-gate error = ufs_sync_indir(ip); 26847c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 26857c478bd9Sstevel@tonic-gate } 26867c478bd9Sstevel@tonic-gate out: 26877c478bd9Sstevel@tonic-gate if (ulp) { 26887c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 26897c478bd9Sstevel@tonic-gate } 26907c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_FSYNC_END, 26917c478bd9Sstevel@tonic-gate "ufs_fsync_end:vp %p error %d", vp, error); 26927c478bd9Sstevel@tonic-gate return (error); 26937c478bd9Sstevel@tonic-gate } 26947c478bd9Sstevel@tonic-gate 26957c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 26967c478bd9Sstevel@tonic-gate static void 26977c478bd9Sstevel@tonic-gate ufs_inactive(struct vnode *vp, struct cred *cr) 26987c478bd9Sstevel@tonic-gate { 26997c478bd9Sstevel@tonic-gate ufs_iinactive(VTOI(vp)); 27007c478bd9Sstevel@tonic-gate } 27017c478bd9Sstevel@tonic-gate 27027c478bd9Sstevel@tonic-gate /* 27037c478bd9Sstevel@tonic-gate * Unix file system operations having to do with directory manipulation. 27047c478bd9Sstevel@tonic-gate */ 27057c478bd9Sstevel@tonic-gate int ufs_lookup_idle_count = 2; /* Number of inodes to idle each time */ 27067c478bd9Sstevel@tonic-gate /* ARGSUSED */ 27077c478bd9Sstevel@tonic-gate static int 27087c478bd9Sstevel@tonic-gate ufs_lookup(struct vnode *dvp, char *nm, struct vnode **vpp, 27097c478bd9Sstevel@tonic-gate struct pathname *pnp, int flags, struct vnode *rdir, struct cred *cr) 27107c478bd9Sstevel@tonic-gate { 27117c478bd9Sstevel@tonic-gate struct inode *ip; 27127c478bd9Sstevel@tonic-gate struct inode *sip; 27137c478bd9Sstevel@tonic-gate struct inode *xip; 27147c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp; 27157c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 27167c478bd9Sstevel@tonic-gate struct vnode *vp; 27177c478bd9Sstevel@tonic-gate int error; 27187c478bd9Sstevel@tonic-gate 27197c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_LOOKUP_START, 27207c478bd9Sstevel@tonic-gate "ufs_lookup_start:dvp %p name %s", dvp, nm); 27217c478bd9Sstevel@tonic-gate 27227c478bd9Sstevel@tonic-gate 27237c478bd9Sstevel@tonic-gate /* 27247c478bd9Sstevel@tonic-gate * Check flags for type of lookup (regular file or attribute file) 27257c478bd9Sstevel@tonic-gate */ 27267c478bd9Sstevel@tonic-gate 27277c478bd9Sstevel@tonic-gate ip = VTOI(dvp); 27287c478bd9Sstevel@tonic-gate 27297c478bd9Sstevel@tonic-gate if (flags & LOOKUP_XATTR) { 27307c478bd9Sstevel@tonic-gate 27317c478bd9Sstevel@tonic-gate /* 27327c478bd9Sstevel@tonic-gate * We don't allow recursive attributes... 27337c478bd9Sstevel@tonic-gate * Maybe someday we will. 27347c478bd9Sstevel@tonic-gate */ 27357c478bd9Sstevel@tonic-gate if ((ip->i_cflags & IXATTR)) { 27367c478bd9Sstevel@tonic-gate return (EINVAL); 27377c478bd9Sstevel@tonic-gate } 27387c478bd9Sstevel@tonic-gate 27397c478bd9Sstevel@tonic-gate if ((vp = dnlc_lookup(dvp, XATTR_DIR_NAME)) == NULL) { 27407c478bd9Sstevel@tonic-gate error = ufs_xattr_getattrdir(dvp, &sip, flags, cr); 27417c478bd9Sstevel@tonic-gate if (error) { 27427c478bd9Sstevel@tonic-gate *vpp = NULL; 27437c478bd9Sstevel@tonic-gate goto out; 27447c478bd9Sstevel@tonic-gate } 27457c478bd9Sstevel@tonic-gate 27467c478bd9Sstevel@tonic-gate vp = ITOV(sip); 27477c478bd9Sstevel@tonic-gate dnlc_update(dvp, XATTR_DIR_NAME, vp); 27487c478bd9Sstevel@tonic-gate } 27497c478bd9Sstevel@tonic-gate 27507c478bd9Sstevel@tonic-gate /* 27517c478bd9Sstevel@tonic-gate * Check accessibility of directory. 27527c478bd9Sstevel@tonic-gate */ 27537c478bd9Sstevel@tonic-gate if (vp == DNLC_NO_VNODE) { 27547c478bd9Sstevel@tonic-gate VN_RELE(vp); 27557c478bd9Sstevel@tonic-gate error = ENOENT; 27567c478bd9Sstevel@tonic-gate goto out; 27577c478bd9Sstevel@tonic-gate } 27587c478bd9Sstevel@tonic-gate if ((error = ufs_iaccess(VTOI(vp), IEXEC, cr)) != 0) { 27597c478bd9Sstevel@tonic-gate VN_RELE(vp); 27607c478bd9Sstevel@tonic-gate goto out; 27617c478bd9Sstevel@tonic-gate } 27627c478bd9Sstevel@tonic-gate 27637c478bd9Sstevel@tonic-gate *vpp = vp; 27647c478bd9Sstevel@tonic-gate return (0); 27657c478bd9Sstevel@tonic-gate } 27667c478bd9Sstevel@tonic-gate 27677c478bd9Sstevel@tonic-gate /* 27687c478bd9Sstevel@tonic-gate * Check for a null component, which we should treat as 27697c478bd9Sstevel@tonic-gate * looking at dvp from within it's parent, so we don't 27707c478bd9Sstevel@tonic-gate * need a call to ufs_iaccess(), as it has already been 27717c478bd9Sstevel@tonic-gate * done. 27727c478bd9Sstevel@tonic-gate */ 27737c478bd9Sstevel@tonic-gate if (nm[0] == 0) { 27747c478bd9Sstevel@tonic-gate VN_HOLD(dvp); 27757c478bd9Sstevel@tonic-gate error = 0; 27767c478bd9Sstevel@tonic-gate *vpp = dvp; 27777c478bd9Sstevel@tonic-gate goto out; 27787c478bd9Sstevel@tonic-gate } 27797c478bd9Sstevel@tonic-gate 27807c478bd9Sstevel@tonic-gate /* 27817c478bd9Sstevel@tonic-gate * Check for "." ie itself. this is a quick check and 27827c478bd9Sstevel@tonic-gate * avoids adding "." into the dnlc (which have been seen 27837c478bd9Sstevel@tonic-gate * to occupy >10% of the cache). 27847c478bd9Sstevel@tonic-gate */ 27857c478bd9Sstevel@tonic-gate if ((nm[0] == '.') && (nm[1] == 0)) { 27867c478bd9Sstevel@tonic-gate /* 27877c478bd9Sstevel@tonic-gate * Don't return without checking accessibility 27887c478bd9Sstevel@tonic-gate * of the directory. We only need the lock if 27897c478bd9Sstevel@tonic-gate * we are going to return it. 27907c478bd9Sstevel@tonic-gate */ 27917c478bd9Sstevel@tonic-gate if ((error = ufs_iaccess(ip, IEXEC, cr)) == 0) { 27927c478bd9Sstevel@tonic-gate VN_HOLD(dvp); 27937c478bd9Sstevel@tonic-gate *vpp = dvp; 27947c478bd9Sstevel@tonic-gate } 27957c478bd9Sstevel@tonic-gate goto out; 27967c478bd9Sstevel@tonic-gate } 27977c478bd9Sstevel@tonic-gate 27987c478bd9Sstevel@tonic-gate /* 27997c478bd9Sstevel@tonic-gate * Fast path: Check the directory name lookup cache. 28007c478bd9Sstevel@tonic-gate */ 28017c478bd9Sstevel@tonic-gate if (vp = dnlc_lookup(dvp, nm)) { 28027c478bd9Sstevel@tonic-gate /* 28037c478bd9Sstevel@tonic-gate * Check accessibility of directory. 28047c478bd9Sstevel@tonic-gate */ 28057c478bd9Sstevel@tonic-gate if ((error = ufs_iaccess(ip, IEXEC, cr)) != 0) { 28067c478bd9Sstevel@tonic-gate VN_RELE(vp); 28077c478bd9Sstevel@tonic-gate goto out; 28087c478bd9Sstevel@tonic-gate } 28097c478bd9Sstevel@tonic-gate if (vp == DNLC_NO_VNODE) { 28107c478bd9Sstevel@tonic-gate VN_RELE(vp); 28117c478bd9Sstevel@tonic-gate error = ENOENT; 28127c478bd9Sstevel@tonic-gate goto out; 28137c478bd9Sstevel@tonic-gate } 28147c478bd9Sstevel@tonic-gate xip = VTOI(vp); 28157c478bd9Sstevel@tonic-gate ulp = NULL; 28167c478bd9Sstevel@tonic-gate goto fastpath; 28177c478bd9Sstevel@tonic-gate } 28187c478bd9Sstevel@tonic-gate 28197c478bd9Sstevel@tonic-gate /* 28207c478bd9Sstevel@tonic-gate * Keep the idle queue from getting too long by 28217c478bd9Sstevel@tonic-gate * idling two inodes before attempting to allocate another. 28227c478bd9Sstevel@tonic-gate * This operation must be performed before entering 28237c478bd9Sstevel@tonic-gate * lockfs or a transaction. 28247c478bd9Sstevel@tonic-gate */ 28257c478bd9Sstevel@tonic-gate if (ufs_idle_q.uq_ne > ufs_idle_q.uq_hiwat) 28267c478bd9Sstevel@tonic-gate if ((curthread->t_flag & T_DONTBLOCK) == 0) { 28277c478bd9Sstevel@tonic-gate ins.in_lidles.value.ul += ufs_lookup_idle_count; 28287c478bd9Sstevel@tonic-gate ufs_idle_some(ufs_lookup_idle_count); 28297c478bd9Sstevel@tonic-gate } 28307c478bd9Sstevel@tonic-gate 2831*02ff05a9Svsakar retry_lookup: 28327c478bd9Sstevel@tonic-gate ufsvfsp = ip->i_ufsvfs; 28337c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_LOOKUP_MASK); 28347c478bd9Sstevel@tonic-gate if (error) 28357c478bd9Sstevel@tonic-gate goto out; 28367c478bd9Sstevel@tonic-gate 28377c478bd9Sstevel@tonic-gate error = ufs_dirlook(ip, nm, &xip, cr, 1); 28387c478bd9Sstevel@tonic-gate 28397c478bd9Sstevel@tonic-gate fastpath: 28407c478bd9Sstevel@tonic-gate if (error == 0) { 28417c478bd9Sstevel@tonic-gate ip = xip; 28427c478bd9Sstevel@tonic-gate *vpp = ITOV(ip); 28437c478bd9Sstevel@tonic-gate 28447c478bd9Sstevel@tonic-gate /* 28457c478bd9Sstevel@tonic-gate * If vnode is a device return special vnode instead. 28467c478bd9Sstevel@tonic-gate */ 28477c478bd9Sstevel@tonic-gate if (IS_DEVVP(*vpp)) { 28487c478bd9Sstevel@tonic-gate struct vnode *newvp; 28497c478bd9Sstevel@tonic-gate 28507c478bd9Sstevel@tonic-gate newvp = specvp(*vpp, (*vpp)->v_rdev, (*vpp)->v_type, 28517c478bd9Sstevel@tonic-gate cr); 28527c478bd9Sstevel@tonic-gate VN_RELE(*vpp); 28537c478bd9Sstevel@tonic-gate if (newvp == NULL) 28547c478bd9Sstevel@tonic-gate error = ENOSYS; 28557c478bd9Sstevel@tonic-gate else 28567c478bd9Sstevel@tonic-gate *vpp = newvp; 28577c478bd9Sstevel@tonic-gate } 28587c478bd9Sstevel@tonic-gate } 28597c478bd9Sstevel@tonic-gate if (ulp) { 28607c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 28617c478bd9Sstevel@tonic-gate } 28627c478bd9Sstevel@tonic-gate 2863*02ff05a9Svsakar if (error == EAGAIN) 2864*02ff05a9Svsakar goto retry_lookup; 2865*02ff05a9Svsakar 28667c478bd9Sstevel@tonic-gate out: 28677c478bd9Sstevel@tonic-gate TRACE_3(TR_FAC_UFS, TR_UFS_LOOKUP_END, 28687c478bd9Sstevel@tonic-gate "ufs_lookup_end:dvp %p name %s error %d", vpp, nm, error); 28697c478bd9Sstevel@tonic-gate return (error); 28707c478bd9Sstevel@tonic-gate } 28717c478bd9Sstevel@tonic-gate 28727c478bd9Sstevel@tonic-gate static int 28737c478bd9Sstevel@tonic-gate ufs_create(struct vnode *dvp, char *name, struct vattr *vap, enum vcexcl excl, 28747c478bd9Sstevel@tonic-gate int mode, struct vnode **vpp, struct cred *cr, int flag) 28757c478bd9Sstevel@tonic-gate { 28767c478bd9Sstevel@tonic-gate struct inode *ip; 28777c478bd9Sstevel@tonic-gate struct inode *xip; 28787c478bd9Sstevel@tonic-gate struct inode *dip; 28797c478bd9Sstevel@tonic-gate struct vnode *xvp; 28807c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp; 28817c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 28827c478bd9Sstevel@tonic-gate int error; 28837c478bd9Sstevel@tonic-gate int issync; 28847c478bd9Sstevel@tonic-gate int truncflag; 28857c478bd9Sstevel@tonic-gate int trans_size; 28867c478bd9Sstevel@tonic-gate int noentry; 28877c478bd9Sstevel@tonic-gate int defer_dip_seq_update = 0; /* need to defer update of dip->i_seq */ 28887c478bd9Sstevel@tonic-gate int retry = 1; 2889*02ff05a9Svsakar int indeadlock; 28907c478bd9Sstevel@tonic-gate 28917c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_CREATE_START, 28927c478bd9Sstevel@tonic-gate "ufs_create_start:dvp %p", dvp); 28937c478bd9Sstevel@tonic-gate 28947c478bd9Sstevel@tonic-gate again: 28957c478bd9Sstevel@tonic-gate ip = VTOI(dvp); 28967c478bd9Sstevel@tonic-gate ufsvfsp = ip->i_ufsvfs; 28977c478bd9Sstevel@tonic-gate truncflag = 0; 28987c478bd9Sstevel@tonic-gate 28997c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_CREATE_MASK); 29007c478bd9Sstevel@tonic-gate if (error) 29017c478bd9Sstevel@tonic-gate goto out; 29027c478bd9Sstevel@tonic-gate 29037c478bd9Sstevel@tonic-gate if (ulp) { 29047c478bd9Sstevel@tonic-gate trans_size = (int)TOP_CREATE_SIZE(ip); 29057c478bd9Sstevel@tonic-gate TRANS_BEGIN_CSYNC(ufsvfsp, issync, TOP_CREATE, trans_size); 29067c478bd9Sstevel@tonic-gate } 29077c478bd9Sstevel@tonic-gate 29087c478bd9Sstevel@tonic-gate if ((vap->va_mode & VSVTX) && secpolicy_vnode_stky_modify(cr) != 0) 29097c478bd9Sstevel@tonic-gate vap->va_mode &= ~VSVTX; 29107c478bd9Sstevel@tonic-gate 29117c478bd9Sstevel@tonic-gate if (*name == '\0') { 29127c478bd9Sstevel@tonic-gate /* 29137c478bd9Sstevel@tonic-gate * Null component name refers to the directory itself. 29147c478bd9Sstevel@tonic-gate */ 29157c478bd9Sstevel@tonic-gate VN_HOLD(dvp); 29167c478bd9Sstevel@tonic-gate /* 29177c478bd9Sstevel@tonic-gate * Even though this is an error case, we need to grab the 29187c478bd9Sstevel@tonic-gate * quota lock since the error handling code below is common. 29197c478bd9Sstevel@tonic-gate */ 29207c478bd9Sstevel@tonic-gate rw_enter(&ufsvfsp->vfs_dqrwlock, RW_READER); 29217c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_WRITER); 29227c478bd9Sstevel@tonic-gate error = EEXIST; 29237c478bd9Sstevel@tonic-gate } else { 29247c478bd9Sstevel@tonic-gate xip = NULL; 29257c478bd9Sstevel@tonic-gate noentry = 0; 2926*02ff05a9Svsakar /* 2927*02ff05a9Svsakar * ufs_tryirwlock_trans uses rw_tryenter and checks for SLOCK 2928*02ff05a9Svsakar * to avoid i_rwlock, ufs_lockfs_begin deadlock. If deadlock 2929*02ff05a9Svsakar * possible, retries the operation. 2930*02ff05a9Svsakar */ 2931*02ff05a9Svsakar ufs_tryirwlock_trans(&ip->i_rwlock, RW_WRITER, TOP_CREATE, 2932*02ff05a9Svsakar retry_dir); 2933*02ff05a9Svsakar if (indeadlock) 2934*02ff05a9Svsakar goto again; 2935*02ff05a9Svsakar 29367c478bd9Sstevel@tonic-gate xvp = dnlc_lookup(dvp, name); 29377c478bd9Sstevel@tonic-gate if (xvp == DNLC_NO_VNODE) { 29387c478bd9Sstevel@tonic-gate noentry = 1; 29397c478bd9Sstevel@tonic-gate VN_RELE(xvp); 29407c478bd9Sstevel@tonic-gate xvp = NULL; 29417c478bd9Sstevel@tonic-gate } 29427c478bd9Sstevel@tonic-gate if (xvp) { 29437c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 29447c478bd9Sstevel@tonic-gate if (error = ufs_iaccess(ip, IEXEC, cr)) { 29457c478bd9Sstevel@tonic-gate VN_RELE(xvp); 29467c478bd9Sstevel@tonic-gate } else { 29477c478bd9Sstevel@tonic-gate error = EEXIST; 29487c478bd9Sstevel@tonic-gate xip = VTOI(xvp); 29497c478bd9Sstevel@tonic-gate } 29507c478bd9Sstevel@tonic-gate } else { 29517c478bd9Sstevel@tonic-gate /* 29527c478bd9Sstevel@tonic-gate * Suppress file system full message if we will retry 29537c478bd9Sstevel@tonic-gate */ 29547c478bd9Sstevel@tonic-gate error = ufs_direnter_cm(ip, name, DE_CREATE, 29557c478bd9Sstevel@tonic-gate vap, &xip, cr, 29567c478bd9Sstevel@tonic-gate (noentry | (retry ? IQUIET : 0))); 2957*02ff05a9Svsakar if (error == EAGAIN) { 2958*02ff05a9Svsakar if (ulp) { 2959*02ff05a9Svsakar TRANS_END_CSYNC(ufsvfsp, error, issync, 2960*02ff05a9Svsakar TOP_CREATE, trans_size); 2961*02ff05a9Svsakar ufs_lockfs_end(ulp); 2962*02ff05a9Svsakar } 2963*02ff05a9Svsakar goto again; 2964*02ff05a9Svsakar } 29657c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 29667c478bd9Sstevel@tonic-gate } 29677c478bd9Sstevel@tonic-gate ip = xip; 29687c478bd9Sstevel@tonic-gate if (ip != NULL) { 29697c478bd9Sstevel@tonic-gate rw_enter(&ufsvfsp->vfs_dqrwlock, RW_READER); 29707c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_WRITER); 29717c478bd9Sstevel@tonic-gate } 29727c478bd9Sstevel@tonic-gate } 29737c478bd9Sstevel@tonic-gate 29747c478bd9Sstevel@tonic-gate /* 29757c478bd9Sstevel@tonic-gate * If the file already exists and this is a non-exclusive create, 29767c478bd9Sstevel@tonic-gate * check permissions and allow access for non-directories. 29777c478bd9Sstevel@tonic-gate * Read-only create of an existing directory is also allowed. 29787c478bd9Sstevel@tonic-gate * We fail an exclusive create of anything which already exists. 29797c478bd9Sstevel@tonic-gate */ 29807c478bd9Sstevel@tonic-gate if (error == EEXIST) { 29817c478bd9Sstevel@tonic-gate dip = VTOI(dvp); 29827c478bd9Sstevel@tonic-gate if (excl == NONEXCL) { 29837c478bd9Sstevel@tonic-gate if ((((ip->i_mode & IFMT) == IFDIR) || 29847c478bd9Sstevel@tonic-gate ((ip->i_mode & IFMT) == IFATTRDIR)) && 29857c478bd9Sstevel@tonic-gate (mode & IWRITE)) 29867c478bd9Sstevel@tonic-gate error = EISDIR; 29877c478bd9Sstevel@tonic-gate else if (mode) 29887c478bd9Sstevel@tonic-gate error = ufs_iaccess(ip, mode, cr); 29897c478bd9Sstevel@tonic-gate else 29907c478bd9Sstevel@tonic-gate error = 0; 29917c478bd9Sstevel@tonic-gate } 29927c478bd9Sstevel@tonic-gate if (error) { 29937c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 29947c478bd9Sstevel@tonic-gate rw_exit(&ufsvfsp->vfs_dqrwlock); 29957c478bd9Sstevel@tonic-gate VN_RELE(ITOV(ip)); 29967c478bd9Sstevel@tonic-gate goto unlock; 29977c478bd9Sstevel@tonic-gate } 29987c478bd9Sstevel@tonic-gate /* 29997c478bd9Sstevel@tonic-gate * If the error EEXIST was set, then i_seq can not 30007c478bd9Sstevel@tonic-gate * have been updated. The sequence number interface 30017c478bd9Sstevel@tonic-gate * is defined such that a non-error VOP_CREATE must 30027c478bd9Sstevel@tonic-gate * increase the dir va_seq it by at least one. If we 30037c478bd9Sstevel@tonic-gate * have cleared the error, increase i_seq. Note that 30047c478bd9Sstevel@tonic-gate * we are increasing the dir i_seq and in rare cases 30057c478bd9Sstevel@tonic-gate * ip may actually be from the dvp, so we already have 30067c478bd9Sstevel@tonic-gate * the locks and it will not be subject to truncation. 30077c478bd9Sstevel@tonic-gate * In case we have to update i_seq of the parent 30087c478bd9Sstevel@tonic-gate * directory dip, we have to defer it till we have 30097c478bd9Sstevel@tonic-gate * released our locks on ip due to lock ordering requirements. 30107c478bd9Sstevel@tonic-gate */ 30117c478bd9Sstevel@tonic-gate if (ip != dip) 30127c478bd9Sstevel@tonic-gate defer_dip_seq_update = 1; 30137c478bd9Sstevel@tonic-gate else 30147c478bd9Sstevel@tonic-gate ip->i_seq++; 30157c478bd9Sstevel@tonic-gate 30167c478bd9Sstevel@tonic-gate if (((ip->i_mode & IFMT) == IFREG) && 30177c478bd9Sstevel@tonic-gate (vap->va_mask & AT_SIZE) && vap->va_size == 0) { 30187c478bd9Sstevel@tonic-gate /* 30197c478bd9Sstevel@tonic-gate * Truncate regular files, if requested by caller. 30207c478bd9Sstevel@tonic-gate * Grab i_rwlock to make sure no one else is 30217c478bd9Sstevel@tonic-gate * currently writing to the file (we promised 30227c478bd9Sstevel@tonic-gate * bmap we would do this). 30237c478bd9Sstevel@tonic-gate * Must get the locks in the correct order. 30247c478bd9Sstevel@tonic-gate */ 30257c478bd9Sstevel@tonic-gate if (ip->i_size == 0) { 30267c478bd9Sstevel@tonic-gate ip->i_flag |= ICHG | IUPD; 30277c478bd9Sstevel@tonic-gate ip->i_seq++; 30287c478bd9Sstevel@tonic-gate TRANS_INODE(ufsvfsp, ip); 30297c478bd9Sstevel@tonic-gate } else { 30307c478bd9Sstevel@tonic-gate /* 30317c478bd9Sstevel@tonic-gate * Large Files: Why this check here? 30327c478bd9Sstevel@tonic-gate * Though we do it in vn_create() we really 30337c478bd9Sstevel@tonic-gate * want to guarantee that we do not destroy 30347c478bd9Sstevel@tonic-gate * Large file data by atomically checking 30357c478bd9Sstevel@tonic-gate * the size while holding the contents 30367c478bd9Sstevel@tonic-gate * lock. 30377c478bd9Sstevel@tonic-gate */ 30387c478bd9Sstevel@tonic-gate if (flag && !(flag & FOFFMAX) && 30397c478bd9Sstevel@tonic-gate ((ip->i_mode & IFMT) == IFREG) && 30407c478bd9Sstevel@tonic-gate (ip->i_size > (offset_t)MAXOFF32_T)) { 30417c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 30427c478bd9Sstevel@tonic-gate rw_exit(&ufsvfsp->vfs_dqrwlock); 30437c478bd9Sstevel@tonic-gate error = EOVERFLOW; 30447c478bd9Sstevel@tonic-gate goto unlock; 30457c478bd9Sstevel@tonic-gate } 30467c478bd9Sstevel@tonic-gate if (TRANS_ISTRANS(ufsvfsp)) 30477c478bd9Sstevel@tonic-gate truncflag++; 30487c478bd9Sstevel@tonic-gate else { 30497c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 30507c478bd9Sstevel@tonic-gate rw_exit(&ufsvfsp->vfs_dqrwlock); 3051*02ff05a9Svsakar ufs_tryirwlock_trans(&ip->i_rwlock, 3052*02ff05a9Svsakar RW_WRITER, TOP_CREATE, 3053*02ff05a9Svsakar retry_file); 3054*02ff05a9Svsakar if (indeadlock) { 3055*02ff05a9Svsakar VN_RELE(ITOV(ip)); 3056*02ff05a9Svsakar goto again; 3057*02ff05a9Svsakar } 30587c478bd9Sstevel@tonic-gate rw_enter(&ufsvfsp->vfs_dqrwlock, 30597c478bd9Sstevel@tonic-gate RW_READER); 30607c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_WRITER); 30617c478bd9Sstevel@tonic-gate (void) ufs_itrunc(ip, (u_offset_t)0, 0, 30627c478bd9Sstevel@tonic-gate cr); 30637c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 30647c478bd9Sstevel@tonic-gate } 30657c478bd9Sstevel@tonic-gate } 30667c478bd9Sstevel@tonic-gate } 30677c478bd9Sstevel@tonic-gate } 30687c478bd9Sstevel@tonic-gate 30697c478bd9Sstevel@tonic-gate if (error) { 30707c478bd9Sstevel@tonic-gate if (ip != NULL) { 30717c478bd9Sstevel@tonic-gate rw_exit(&ufsvfsp->vfs_dqrwlock); 30727c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 30737c478bd9Sstevel@tonic-gate } 30747c478bd9Sstevel@tonic-gate goto unlock; 30757c478bd9Sstevel@tonic-gate } 30767c478bd9Sstevel@tonic-gate 30777c478bd9Sstevel@tonic-gate *vpp = ITOV(ip); 30787c478bd9Sstevel@tonic-gate ITIMES(ip); 30797c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 30807c478bd9Sstevel@tonic-gate rw_exit(&ufsvfsp->vfs_dqrwlock); 30817c478bd9Sstevel@tonic-gate 30827c478bd9Sstevel@tonic-gate /* 30837c478bd9Sstevel@tonic-gate * If vnode is a device return special vnode instead. 30847c478bd9Sstevel@tonic-gate */ 30857c478bd9Sstevel@tonic-gate if (!error && IS_DEVVP(*vpp)) { 30867c478bd9Sstevel@tonic-gate struct vnode *newvp; 30877c478bd9Sstevel@tonic-gate 30887c478bd9Sstevel@tonic-gate newvp = specvp(*vpp, (*vpp)->v_rdev, (*vpp)->v_type, cr); 30897c478bd9Sstevel@tonic-gate VN_RELE(*vpp); 30907c478bd9Sstevel@tonic-gate if (newvp == NULL) { 30917c478bd9Sstevel@tonic-gate error = ENOSYS; 30927c478bd9Sstevel@tonic-gate goto unlock; 30937c478bd9Sstevel@tonic-gate } 30947c478bd9Sstevel@tonic-gate truncflag = 0; 30957c478bd9Sstevel@tonic-gate *vpp = newvp; 30967c478bd9Sstevel@tonic-gate } 30977c478bd9Sstevel@tonic-gate unlock: 30987c478bd9Sstevel@tonic-gate 30997c478bd9Sstevel@tonic-gate /* 31007c478bd9Sstevel@tonic-gate * Do the deferred update of the parent directory's sequence 31017c478bd9Sstevel@tonic-gate * number now. 31027c478bd9Sstevel@tonic-gate */ 31037c478bd9Sstevel@tonic-gate if (defer_dip_seq_update == 1) { 31047c478bd9Sstevel@tonic-gate rw_enter(&dip->i_contents, RW_READER); 31057c478bd9Sstevel@tonic-gate mutex_enter(&dip->i_tlock); 31067c478bd9Sstevel@tonic-gate dip->i_seq++; 31077c478bd9Sstevel@tonic-gate mutex_exit(&dip->i_tlock); 31087c478bd9Sstevel@tonic-gate rw_exit(&dip->i_contents); 31097c478bd9Sstevel@tonic-gate } 31107c478bd9Sstevel@tonic-gate 31117c478bd9Sstevel@tonic-gate if (ulp) { 31127c478bd9Sstevel@tonic-gate int terr = 0; 31137c478bd9Sstevel@tonic-gate 31147c478bd9Sstevel@tonic-gate TRANS_END_CSYNC(ufsvfsp, terr, issync, TOP_CREATE, 31157c478bd9Sstevel@tonic-gate trans_size); 31167c478bd9Sstevel@tonic-gate 31177c478bd9Sstevel@tonic-gate /* 31187c478bd9Sstevel@tonic-gate * If we haven't had a more interesting failure 31197c478bd9Sstevel@tonic-gate * already, then anything that might've happened 31207c478bd9Sstevel@tonic-gate * here should be reported. 31217c478bd9Sstevel@tonic-gate */ 31227c478bd9Sstevel@tonic-gate if (error == 0) 31237c478bd9Sstevel@tonic-gate error = terr; 31247c478bd9Sstevel@tonic-gate } 31257c478bd9Sstevel@tonic-gate 31267c478bd9Sstevel@tonic-gate if (!error && truncflag) { 3127*02ff05a9Svsakar ufs_tryirwlock(&ip->i_rwlock, RW_WRITER, retry_trunc); 3128*02ff05a9Svsakar if (indeadlock) { 3129*02ff05a9Svsakar if (ulp) 3130*02ff05a9Svsakar ufs_lockfs_end(ulp); 3131*02ff05a9Svsakar VN_RELE(ITOV(ip)); 3132*02ff05a9Svsakar goto again; 3133*02ff05a9Svsakar } 31347c478bd9Sstevel@tonic-gate (void) TRANS_ITRUNC(ip, (u_offset_t)0, 0, cr); 31357c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 31367c478bd9Sstevel@tonic-gate } 31377c478bd9Sstevel@tonic-gate 31387c478bd9Sstevel@tonic-gate if (ulp) 31397c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 31407c478bd9Sstevel@tonic-gate 31417c478bd9Sstevel@tonic-gate /* 31427c478bd9Sstevel@tonic-gate * If no inodes available, try to free one up out of the 31437c478bd9Sstevel@tonic-gate * pending delete queue. 31447c478bd9Sstevel@tonic-gate */ 31457c478bd9Sstevel@tonic-gate if ((error == ENOSPC) && retry && TRANS_ISTRANS(ufsvfsp)) { 31467c478bd9Sstevel@tonic-gate ufs_delete_drain_wait(ufsvfsp, 1); 31477c478bd9Sstevel@tonic-gate retry = 0; 31487c478bd9Sstevel@tonic-gate goto again; 31497c478bd9Sstevel@tonic-gate } 31507c478bd9Sstevel@tonic-gate 31517c478bd9Sstevel@tonic-gate out: 31527c478bd9Sstevel@tonic-gate TRACE_3(TR_FAC_UFS, TR_UFS_CREATE_END, 31537c478bd9Sstevel@tonic-gate "ufs_create_end:dvp %p name %s error %d", vpp, name, error); 31547c478bd9Sstevel@tonic-gate return (error); 31557c478bd9Sstevel@tonic-gate } 31567c478bd9Sstevel@tonic-gate 31577c478bd9Sstevel@tonic-gate extern int ufs_idle_max; 31587c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 31597c478bd9Sstevel@tonic-gate static int 31607c478bd9Sstevel@tonic-gate ufs_remove(struct vnode *vp, char *nm, struct cred *cr) 31617c478bd9Sstevel@tonic-gate { 31627c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 31637c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp = ip->i_ufsvfs; 31647c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 31657c478bd9Sstevel@tonic-gate vnode_t *rmvp = NULL; /* Vnode corresponding to name being removed */ 3166*02ff05a9Svsakar int indeadlock; 31677c478bd9Sstevel@tonic-gate int error; 31687c478bd9Sstevel@tonic-gate int issync; 31697c478bd9Sstevel@tonic-gate int trans_size; 31707c478bd9Sstevel@tonic-gate 31717c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_REMOVE_START, 31727c478bd9Sstevel@tonic-gate "ufs_remove_start:vp %p", vp); 31737c478bd9Sstevel@tonic-gate 31747c478bd9Sstevel@tonic-gate /* 31757c478bd9Sstevel@tonic-gate * don't let the delete queue get too long 31767c478bd9Sstevel@tonic-gate */ 31777c478bd9Sstevel@tonic-gate if (ufsvfsp == NULL) { 31787c478bd9Sstevel@tonic-gate error = EIO; 31797c478bd9Sstevel@tonic-gate goto out; 31807c478bd9Sstevel@tonic-gate } 31817c478bd9Sstevel@tonic-gate if (ufsvfsp->vfs_delete.uq_ne > ufs_idle_max) 31827c478bd9Sstevel@tonic-gate ufs_delete_drain(vp->v_vfsp, 1, 1); 31837c478bd9Sstevel@tonic-gate 3184*02ff05a9Svsakar retry_remove: 31857c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_REMOVE_MASK); 31867c478bd9Sstevel@tonic-gate if (error) 31877c478bd9Sstevel@tonic-gate goto out; 31887c478bd9Sstevel@tonic-gate 31897c478bd9Sstevel@tonic-gate if (ulp) 31907c478bd9Sstevel@tonic-gate TRANS_BEGIN_CSYNC(ufsvfsp, issync, TOP_REMOVE, 31917c478bd9Sstevel@tonic-gate trans_size = (int)TOP_REMOVE_SIZE(VTOI(vp))); 31927c478bd9Sstevel@tonic-gate 3193*02ff05a9Svsakar /* 3194*02ff05a9Svsakar * ufs_tryirwlock_trans uses rw_tryenter and checks for SLOCK 3195*02ff05a9Svsakar * to avoid i_rwlock, ufs_lockfs_begin deadlock. If deadlock 3196*02ff05a9Svsakar * possible, retries the operation. 3197*02ff05a9Svsakar */ 3198*02ff05a9Svsakar ufs_tryirwlock_trans(&ip->i_rwlock, RW_WRITER, TOP_REMOVE, retry); 3199*02ff05a9Svsakar if (indeadlock) 3200*02ff05a9Svsakar goto retry_remove; 32017c478bd9Sstevel@tonic-gate error = ufs_dirremove(ip, nm, (struct inode *)0, (struct vnode *)0, 32027c478bd9Sstevel@tonic-gate DR_REMOVE, cr, &rmvp); 32037c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 32047c478bd9Sstevel@tonic-gate 32057c478bd9Sstevel@tonic-gate if (ulp) { 32067c478bd9Sstevel@tonic-gate TRANS_END_CSYNC(ufsvfsp, error, issync, TOP_REMOVE, trans_size); 32077c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 32087c478bd9Sstevel@tonic-gate } 32097c478bd9Sstevel@tonic-gate 32107c478bd9Sstevel@tonic-gate /* 32117c478bd9Sstevel@tonic-gate * This must be called after the remove transaction is closed. 32127c478bd9Sstevel@tonic-gate */ 32137c478bd9Sstevel@tonic-gate if (rmvp != NULL) { 32147c478bd9Sstevel@tonic-gate /* Only send the event if there were no errors */ 32157c478bd9Sstevel@tonic-gate if (error == 0) 32167c478bd9Sstevel@tonic-gate vnevent_remove(rmvp); 32177c478bd9Sstevel@tonic-gate VN_RELE(rmvp); 32187c478bd9Sstevel@tonic-gate } 32197c478bd9Sstevel@tonic-gate out: 32207c478bd9Sstevel@tonic-gate TRACE_3(TR_FAC_UFS, TR_UFS_REMOVE_END, 32217c478bd9Sstevel@tonic-gate "ufs_remove_end:vp %p name %s error %d", vp, nm, error); 32227c478bd9Sstevel@tonic-gate return (error); 32237c478bd9Sstevel@tonic-gate } 32247c478bd9Sstevel@tonic-gate 32257c478bd9Sstevel@tonic-gate /* 32267c478bd9Sstevel@tonic-gate * Link a file or a directory. Only privileged processes are allowed to 32277c478bd9Sstevel@tonic-gate * make links to directories. 32287c478bd9Sstevel@tonic-gate */ 32297c478bd9Sstevel@tonic-gate static int 32307c478bd9Sstevel@tonic-gate ufs_link(struct vnode *tdvp, struct vnode *svp, char *tnm, struct cred *cr) 32317c478bd9Sstevel@tonic-gate { 32327c478bd9Sstevel@tonic-gate struct inode *sip; 32337c478bd9Sstevel@tonic-gate struct inode *tdp = VTOI(tdvp); 32347c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp = tdp->i_ufsvfs; 32357c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 32367c478bd9Sstevel@tonic-gate struct vnode *realvp; 32377c478bd9Sstevel@tonic-gate int error; 32387c478bd9Sstevel@tonic-gate int issync; 32397c478bd9Sstevel@tonic-gate int trans_size; 32407c478bd9Sstevel@tonic-gate int isdev; 3241*02ff05a9Svsakar int indeadlock; 32427c478bd9Sstevel@tonic-gate 32437c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_LINK_START, 32447c478bd9Sstevel@tonic-gate "ufs_link_start:tdvp %p", tdvp); 32457c478bd9Sstevel@tonic-gate 3246*02ff05a9Svsakar retry_link: 32477c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_LINK_MASK); 32487c478bd9Sstevel@tonic-gate if (error) 32497c478bd9Sstevel@tonic-gate goto out; 32507c478bd9Sstevel@tonic-gate 32517c478bd9Sstevel@tonic-gate if (ulp) 32527c478bd9Sstevel@tonic-gate TRANS_BEGIN_CSYNC(ufsvfsp, issync, TOP_LINK, 32537c478bd9Sstevel@tonic-gate trans_size = (int)TOP_LINK_SIZE(VTOI(tdvp))); 32547c478bd9Sstevel@tonic-gate 32557c478bd9Sstevel@tonic-gate if (VOP_REALVP(svp, &realvp) == 0) 32567c478bd9Sstevel@tonic-gate svp = realvp; 32577c478bd9Sstevel@tonic-gate 32587c478bd9Sstevel@tonic-gate /* 32597c478bd9Sstevel@tonic-gate * Make sure link for extended attributes is valid 32607c478bd9Sstevel@tonic-gate * We only support hard linking of attr in ATTRDIR to ATTRDIR 32617c478bd9Sstevel@tonic-gate * 32627c478bd9Sstevel@tonic-gate * Make certain we don't attempt to look at a device node as 32637c478bd9Sstevel@tonic-gate * a ufs inode. 32647c478bd9Sstevel@tonic-gate */ 32657c478bd9Sstevel@tonic-gate 32667c478bd9Sstevel@tonic-gate isdev = IS_DEVVP(svp); 32677c478bd9Sstevel@tonic-gate if (((isdev == 0) && ((VTOI(svp)->i_cflags & IXATTR) == 0) && 32687c478bd9Sstevel@tonic-gate ((tdp->i_mode & IFMT) == IFATTRDIR)) || 32697c478bd9Sstevel@tonic-gate ((isdev == 0) && (VTOI(svp)->i_cflags & IXATTR) && 32707c478bd9Sstevel@tonic-gate ((tdp->i_mode & IFMT) == IFDIR))) { 32717c478bd9Sstevel@tonic-gate error = EINVAL; 32727c478bd9Sstevel@tonic-gate goto unlock; 32737c478bd9Sstevel@tonic-gate } 32747c478bd9Sstevel@tonic-gate 32757c478bd9Sstevel@tonic-gate sip = VTOI(svp); 32767c478bd9Sstevel@tonic-gate if ((svp->v_type == VDIR && 32777c478bd9Sstevel@tonic-gate secpolicy_fs_linkdir(cr, ufsvfsp->vfs_vfs) != 0) || 32787c478bd9Sstevel@tonic-gate (sip->i_uid != crgetuid(cr) && secpolicy_basic_link(cr) != 0)) { 32797c478bd9Sstevel@tonic-gate error = EPERM; 32807c478bd9Sstevel@tonic-gate goto unlock; 32817c478bd9Sstevel@tonic-gate } 3282*02ff05a9Svsakar 3283*02ff05a9Svsakar /* 3284*02ff05a9Svsakar * ufs_tryirwlock_trans uses rw_tryenter and checks for SLOCK 3285*02ff05a9Svsakar * to avoid i_rwlock, ufs_lockfs_begin deadlock. If deadlock 3286*02ff05a9Svsakar * possible, retries the operation. 3287*02ff05a9Svsakar */ 3288*02ff05a9Svsakar ufs_tryirwlock_trans(&tdp->i_rwlock, RW_WRITER, TOP_LINK, retry); 3289*02ff05a9Svsakar if (indeadlock) 3290*02ff05a9Svsakar goto retry_link; 32917c478bd9Sstevel@tonic-gate error = ufs_direnter_lr(tdp, tnm, DE_LINK, (struct inode *)0, 32927c478bd9Sstevel@tonic-gate sip, cr, NULL); 32937c478bd9Sstevel@tonic-gate rw_exit(&tdp->i_rwlock); 32947c478bd9Sstevel@tonic-gate 32957c478bd9Sstevel@tonic-gate unlock: 32967c478bd9Sstevel@tonic-gate if (ulp) { 32977c478bd9Sstevel@tonic-gate TRANS_END_CSYNC(ufsvfsp, error, issync, TOP_LINK, trans_size); 32987c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 32997c478bd9Sstevel@tonic-gate } 33007c478bd9Sstevel@tonic-gate out: 33017c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_LINK_END, 33027c478bd9Sstevel@tonic-gate "ufs_link_end:tdvp %p error %d", tdvp, error); 33037c478bd9Sstevel@tonic-gate return (error); 33047c478bd9Sstevel@tonic-gate } 33057c478bd9Sstevel@tonic-gate 33067c478bd9Sstevel@tonic-gate uint64_t ufs_rename_retry_cnt; 33077c478bd9Sstevel@tonic-gate uint64_t ufs_rename_upgrade_retry_cnt; 33087c478bd9Sstevel@tonic-gate uint64_t ufs_rename_dircheck_retry_cnt; 33097c478bd9Sstevel@tonic-gate clock_t ufs_rename_backoff_delay = 1; 33107c478bd9Sstevel@tonic-gate 33117c478bd9Sstevel@tonic-gate /* 33127c478bd9Sstevel@tonic-gate * Rename a file or directory. 33137c478bd9Sstevel@tonic-gate * We are given the vnode and entry string of the source and the 33147c478bd9Sstevel@tonic-gate * vnode and entry string of the place we want to move the source 33157c478bd9Sstevel@tonic-gate * to (the target). The essential operation is: 33167c478bd9Sstevel@tonic-gate * unlink(target); 33177c478bd9Sstevel@tonic-gate * link(source, target); 33187c478bd9Sstevel@tonic-gate * unlink(source); 33197c478bd9Sstevel@tonic-gate * but "atomically". Can't do full commit without saving state in 33207c478bd9Sstevel@tonic-gate * the inode on disk, which isn't feasible at this time. Best we 33217c478bd9Sstevel@tonic-gate * can do is always guarantee that the TARGET exists. 33227c478bd9Sstevel@tonic-gate */ 3323baa4d099Sswilcox 33247c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 33257c478bd9Sstevel@tonic-gate static int 33267c478bd9Sstevel@tonic-gate ufs_rename( 33277c478bd9Sstevel@tonic-gate struct vnode *sdvp, /* old (source) parent vnode */ 33287c478bd9Sstevel@tonic-gate char *snm, /* old (source) entry name */ 33297c478bd9Sstevel@tonic-gate struct vnode *tdvp, /* new (target) parent vnode */ 33307c478bd9Sstevel@tonic-gate char *tnm, /* new (target) entry name */ 33317c478bd9Sstevel@tonic-gate struct cred *cr) 33327c478bd9Sstevel@tonic-gate { 33337c478bd9Sstevel@tonic-gate struct inode *sip = NULL; /* source inode */ 3334baa4d099Sswilcox struct inode *ip = NULL; /* check inode */ 33357c478bd9Sstevel@tonic-gate struct inode *sdp; /* old (source) parent inode */ 33367c478bd9Sstevel@tonic-gate struct inode *tdp; /* new (target) parent inode */ 33377c478bd9Sstevel@tonic-gate struct vnode *tvp = NULL; /* target vnode, if it exists */ 33387c478bd9Sstevel@tonic-gate struct vnode *realvp; 33397c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp; 33407c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 33415b024a5bSbatschul struct ufs_slot slot; 3342baa4d099Sswilcox timestruc_t now; 33437c478bd9Sstevel@tonic-gate int error; 33447c478bd9Sstevel@tonic-gate int issync; 33457c478bd9Sstevel@tonic-gate int trans_size; 3346*02ff05a9Svsakar krwlock_t *first_lock; 3347*02ff05a9Svsakar krwlock_t *second_lock; 3348*02ff05a9Svsakar krwlock_t *reverse_lock; 33497c478bd9Sstevel@tonic-gate 33507c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_RENAME_START, 33517c478bd9Sstevel@tonic-gate "ufs_rename_start:sdvp %p", sdvp); 33527c478bd9Sstevel@tonic-gate 33537c478bd9Sstevel@tonic-gate 33547c478bd9Sstevel@tonic-gate sdp = VTOI(sdvp); 3355baa4d099Sswilcox slot.fbp = NULL; 33567c478bd9Sstevel@tonic-gate ufsvfsp = sdp->i_ufsvfs; 3357*02ff05a9Svsakar retry_rename: 33587c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_RENAME_MASK); 33597c478bd9Sstevel@tonic-gate if (error) 33607c478bd9Sstevel@tonic-gate goto out; 33617c478bd9Sstevel@tonic-gate 33627c478bd9Sstevel@tonic-gate if (ulp) 33637c478bd9Sstevel@tonic-gate TRANS_BEGIN_CSYNC(ufsvfsp, issync, TOP_RENAME, 33647c478bd9Sstevel@tonic-gate trans_size = (int)TOP_RENAME_SIZE(sdp)); 33657c478bd9Sstevel@tonic-gate 33667c478bd9Sstevel@tonic-gate if (VOP_REALVP(tdvp, &realvp) == 0) 33677c478bd9Sstevel@tonic-gate tdvp = realvp; 33687c478bd9Sstevel@tonic-gate 33697c478bd9Sstevel@tonic-gate tdp = VTOI(tdvp); 33707c478bd9Sstevel@tonic-gate 3371baa4d099Sswilcox 33727c478bd9Sstevel@tonic-gate /* 33737c478bd9Sstevel@tonic-gate * We only allow renaming of attributes from ATTRDIR to ATTRDIR. 33747c478bd9Sstevel@tonic-gate */ 33757c478bd9Sstevel@tonic-gate if ((tdp->i_mode & IFMT) != (sdp->i_mode & IFMT)) { 33767c478bd9Sstevel@tonic-gate error = EINVAL; 33777c478bd9Sstevel@tonic-gate goto unlock; 33787c478bd9Sstevel@tonic-gate } 33797c478bd9Sstevel@tonic-gate 33807c478bd9Sstevel@tonic-gate /* 33817c478bd9Sstevel@tonic-gate * Look up inode of file we're supposed to rename. 33827c478bd9Sstevel@tonic-gate */ 3383baa4d099Sswilcox gethrestime(&now); 33847c478bd9Sstevel@tonic-gate if (error = ufs_dirlook(sdp, snm, &sip, cr, 0)) { 3385*02ff05a9Svsakar if (error == EAGAIN) { 3386*02ff05a9Svsakar if (ulp) { 3387*02ff05a9Svsakar TRANS_END_CSYNC(ufsvfsp, error, issync, 3388*02ff05a9Svsakar TOP_RENAME, trans_size); 3389*02ff05a9Svsakar ufs_lockfs_end(ulp); 3390*02ff05a9Svsakar } 3391*02ff05a9Svsakar goto retry_rename; 3392*02ff05a9Svsakar } 3393*02ff05a9Svsakar 33947c478bd9Sstevel@tonic-gate goto unlock; 33957c478bd9Sstevel@tonic-gate } 33967c478bd9Sstevel@tonic-gate 33977c478bd9Sstevel@tonic-gate /* 33987c478bd9Sstevel@tonic-gate * Lock both the source and target directories (they may be 33997c478bd9Sstevel@tonic-gate * the same) to provide the atomicity semantics that was 34007c478bd9Sstevel@tonic-gate * previously provided by the per file system vfs_rename_lock 34017c478bd9Sstevel@tonic-gate * 34027c478bd9Sstevel@tonic-gate * with vfs_rename_lock removed to allow simultaneous renames 34037c478bd9Sstevel@tonic-gate * within a file system, ufs_dircheckpath can deadlock while 34047c478bd9Sstevel@tonic-gate * traversing back to ensure that source is not a parent directory 34057c478bd9Sstevel@tonic-gate * of target parent directory. This is because we get into 34067c478bd9Sstevel@tonic-gate * ufs_dircheckpath with the sdp and tdp locks held as RW_WRITER. 34077c478bd9Sstevel@tonic-gate * If the tdp and sdp of the simultaneous renames happen to be 34087c478bd9Sstevel@tonic-gate * in the path of each other, it can lead to a deadlock. This 34097c478bd9Sstevel@tonic-gate * can be avoided by getting the locks as RW_READER here and then 34107c478bd9Sstevel@tonic-gate * upgrading to RW_WRITER after completing the ufs_dircheckpath. 3411*02ff05a9Svsakar * 3412*02ff05a9Svsakar * We hold the target directory's i_rwlock after calling 3413*02ff05a9Svsakar * ufs_lockfs_begin but in many other operations (like ufs_readdir) 3414*02ff05a9Svsakar * VOP_RWLOCK is explicitly called by the filesystem independent code 3415*02ff05a9Svsakar * before calling the file system operation. In these cases the order 3416*02ff05a9Svsakar * is reversed (i.e i_rwlock is taken first and then ufs_lockfs_begin 3417*02ff05a9Svsakar * is called). This is fine as long as ufs_lockfs_begin acts as a VOP 3418*02ff05a9Svsakar * counter but with ufs_quiesce setting the SLOCK bit this becomes a 3419*02ff05a9Svsakar * synchronizing object which might lead to a deadlock. So we use 3420*02ff05a9Svsakar * rw_tryenter instead of rw_enter. If we fail to get this lock and 3421*02ff05a9Svsakar * find that SLOCK bit is set, we call ufs_lockfs_end and restart the 3422*02ff05a9Svsakar * operation. 34237c478bd9Sstevel@tonic-gate */ 34247c478bd9Sstevel@tonic-gate retry: 3425*02ff05a9Svsakar first_lock = &tdp->i_rwlock; 3426*02ff05a9Svsakar second_lock = &sdp->i_rwlock; 3427*02ff05a9Svsakar retry_firstlock: 3428*02ff05a9Svsakar if (!rw_tryenter(first_lock, RW_READER)) { 34297c478bd9Sstevel@tonic-gate /* 3430*02ff05a9Svsakar * We didn't get the lock. Check if the SLOCK is set in the 3431*02ff05a9Svsakar * ufsvfs. If yes, we might be in a deadlock. Safer to give up 3432*02ff05a9Svsakar * and wait for SLOCK to be cleared. 34337c478bd9Sstevel@tonic-gate */ 3434*02ff05a9Svsakar 3435*02ff05a9Svsakar if (ulp && ULOCKFS_IS_SLOCK(ulp)) { 3436*02ff05a9Svsakar TRANS_END_CSYNC(ufsvfsp, error, issync, TOP_RENAME, 3437*02ff05a9Svsakar trans_size); 3438*02ff05a9Svsakar ufs_lockfs_end(ulp); 3439*02ff05a9Svsakar goto retry_rename; 3440*02ff05a9Svsakar 3441*02ff05a9Svsakar } else { 34427c478bd9Sstevel@tonic-gate /* 3443*02ff05a9Svsakar * SLOCK isn't set so this is a genuine synchronization 3444*02ff05a9Svsakar * case. Let's try again after giving them a breather. 34457c478bd9Sstevel@tonic-gate */ 3446*02ff05a9Svsakar delay(RETRY_LOCK_DELAY); 3447*02ff05a9Svsakar goto retry_firstlock; 3448*02ff05a9Svsakar } 3449*02ff05a9Svsakar } 3450*02ff05a9Svsakar /* 3451*02ff05a9Svsakar * Need to check if the tdp and sdp are same !!! 3452*02ff05a9Svsakar */ 3453*02ff05a9Svsakar if ((tdp != sdp) && (!rw_tryenter(second_lock, RW_READER))) { 3454*02ff05a9Svsakar /* 3455*02ff05a9Svsakar * We didn't get the lock. Check if the SLOCK is set in the 3456*02ff05a9Svsakar * ufsvfs. If yes, we might be in a deadlock. Safer to give up 3457*02ff05a9Svsakar * and wait for SLOCK to be cleared. 3458*02ff05a9Svsakar */ 3459*02ff05a9Svsakar 3460*02ff05a9Svsakar rw_exit(first_lock); 3461*02ff05a9Svsakar if (ulp && ULOCKFS_IS_SLOCK(ulp)) { 3462*02ff05a9Svsakar TRANS_END_CSYNC(ufsvfsp, error, issync, TOP_RENAME, 3463*02ff05a9Svsakar trans_size); 3464*02ff05a9Svsakar ufs_lockfs_end(ulp); 3465*02ff05a9Svsakar goto retry_rename; 3466*02ff05a9Svsakar 3467*02ff05a9Svsakar } else { 3468*02ff05a9Svsakar /* 3469*02ff05a9Svsakar * So we couldn't get the second level peer lock *and* 3470*02ff05a9Svsakar * the SLOCK bit isn't set. Too bad we can be 3471*02ff05a9Svsakar * contentding with someone wanting these locks otherway 3472*02ff05a9Svsakar * round. Reverse the locks in case there is a heavy 3473*02ff05a9Svsakar * contention for the second level lock. 3474*02ff05a9Svsakar */ 3475*02ff05a9Svsakar reverse_lock = first_lock; 3476*02ff05a9Svsakar first_lock = second_lock; 3477*02ff05a9Svsakar second_lock = reverse_lock; 3478*02ff05a9Svsakar ufs_rename_retry_cnt++; 3479*02ff05a9Svsakar goto retry_firstlock; 34807c478bd9Sstevel@tonic-gate } 34817c478bd9Sstevel@tonic-gate } 34827c478bd9Sstevel@tonic-gate 34837c478bd9Sstevel@tonic-gate if (sip == tdp) { 34847c478bd9Sstevel@tonic-gate error = EINVAL; 34857c478bd9Sstevel@tonic-gate goto errout; 34867c478bd9Sstevel@tonic-gate } 34877c478bd9Sstevel@tonic-gate /* 34887c478bd9Sstevel@tonic-gate * Make sure we can delete the source entry. This requires 34897c478bd9Sstevel@tonic-gate * write permission on the containing directory. 34907c478bd9Sstevel@tonic-gate * Check for sticky directories. 34917c478bd9Sstevel@tonic-gate */ 34927c478bd9Sstevel@tonic-gate rw_enter(&sdp->i_contents, RW_READER); 34937c478bd9Sstevel@tonic-gate rw_enter(&sip->i_contents, RW_READER); 34947c478bd9Sstevel@tonic-gate if ((error = ufs_iaccess(sdp, IWRITE, cr)) != 0 || 34957c478bd9Sstevel@tonic-gate (error = ufs_sticky_remove_access(sdp, sip, cr)) != 0) { 34967c478bd9Sstevel@tonic-gate rw_exit(&sip->i_contents); 34977c478bd9Sstevel@tonic-gate rw_exit(&sdp->i_contents); 34987c478bd9Sstevel@tonic-gate goto errout; 34997c478bd9Sstevel@tonic-gate } 35007c478bd9Sstevel@tonic-gate 35017c478bd9Sstevel@tonic-gate /* 35027c478bd9Sstevel@tonic-gate * If this is a rename of a directory and the parent is 35037c478bd9Sstevel@tonic-gate * different (".." must be changed), then the source 35047c478bd9Sstevel@tonic-gate * directory must not be in the directory hierarchy 35057c478bd9Sstevel@tonic-gate * above the target, as this would orphan everything 35067c478bd9Sstevel@tonic-gate * below the source directory. Also the user must have 35077c478bd9Sstevel@tonic-gate * write permission in the source so as to be able to 35087c478bd9Sstevel@tonic-gate * change "..". 35097c478bd9Sstevel@tonic-gate */ 35107c478bd9Sstevel@tonic-gate if ((((sip->i_mode & IFMT) == IFDIR) || 35117c478bd9Sstevel@tonic-gate ((sip->i_mode & IFMT) == IFATTRDIR)) && sdp != tdp) { 35127c478bd9Sstevel@tonic-gate ino_t inum; 35137c478bd9Sstevel@tonic-gate 35147c478bd9Sstevel@tonic-gate if ((error = ufs_iaccess(sip, IWRITE, cr))) { 35157c478bd9Sstevel@tonic-gate rw_exit(&sip->i_contents); 35167c478bd9Sstevel@tonic-gate rw_exit(&sdp->i_contents); 35177c478bd9Sstevel@tonic-gate goto errout; 35187c478bd9Sstevel@tonic-gate } 35197c478bd9Sstevel@tonic-gate inum = sip->i_number; 35207c478bd9Sstevel@tonic-gate rw_exit(&sip->i_contents); 35217c478bd9Sstevel@tonic-gate rw_exit(&sdp->i_contents); 35227c478bd9Sstevel@tonic-gate if ((error = ufs_dircheckpath(inum, tdp, sdp, cr))) { 35237c478bd9Sstevel@tonic-gate /* 35247c478bd9Sstevel@tonic-gate * If we got EAGAIN ufs_dircheckpath detected a 35257c478bd9Sstevel@tonic-gate * potential deadlock and backed out. We need 35267c478bd9Sstevel@tonic-gate * to retry the operation since sdp and tdp have 35277c478bd9Sstevel@tonic-gate * to be released to avoid the deadlock. 35287c478bd9Sstevel@tonic-gate */ 35297c478bd9Sstevel@tonic-gate if (error == EAGAIN) { 35307c478bd9Sstevel@tonic-gate rw_exit(&tdp->i_rwlock); 35317c478bd9Sstevel@tonic-gate if (tdp != sdp) 35327c478bd9Sstevel@tonic-gate rw_exit(&sdp->i_rwlock); 35337c478bd9Sstevel@tonic-gate delay(ufs_rename_backoff_delay); 35347c478bd9Sstevel@tonic-gate ufs_rename_dircheck_retry_cnt++; 35357c478bd9Sstevel@tonic-gate goto retry; 35367c478bd9Sstevel@tonic-gate } 35377c478bd9Sstevel@tonic-gate goto errout; 35387c478bd9Sstevel@tonic-gate } 35397c478bd9Sstevel@tonic-gate } else { 35407c478bd9Sstevel@tonic-gate rw_exit(&sip->i_contents); 35417c478bd9Sstevel@tonic-gate rw_exit(&sdp->i_contents); 35427c478bd9Sstevel@tonic-gate } 35437c478bd9Sstevel@tonic-gate 35447c478bd9Sstevel@tonic-gate 35457c478bd9Sstevel@tonic-gate /* 35467c478bd9Sstevel@tonic-gate * Check for renaming '.' or '..' or alias of '.' 35477c478bd9Sstevel@tonic-gate */ 35487c478bd9Sstevel@tonic-gate if (strcmp(snm, ".") == 0 || strcmp(snm, "..") == 0 || sdp == sip) { 35497c478bd9Sstevel@tonic-gate error = EINVAL; 35507c478bd9Sstevel@tonic-gate goto errout; 35517c478bd9Sstevel@tonic-gate } 35527c478bd9Sstevel@tonic-gate 35537c478bd9Sstevel@tonic-gate /* 35547c478bd9Sstevel@tonic-gate * Simultaneous renames can deadlock in ufs_dircheckpath since it 35557c478bd9Sstevel@tonic-gate * tries to traverse back the file tree with both tdp and sdp held 35567c478bd9Sstevel@tonic-gate * as RW_WRITER. To avoid that we have to hold the tdp and sdp locks 35577c478bd9Sstevel@tonic-gate * as RW_READERS till ufs_dircheckpath is done. 35587c478bd9Sstevel@tonic-gate * Now that ufs_dircheckpath is done with, we can upgrade the locks 35597c478bd9Sstevel@tonic-gate * to RW_WRITER. 35607c478bd9Sstevel@tonic-gate */ 35617c478bd9Sstevel@tonic-gate if (!rw_tryupgrade(&tdp->i_rwlock)) { 35627c478bd9Sstevel@tonic-gate /* 35637c478bd9Sstevel@tonic-gate * The upgrade failed. We got to give away the lock 35647c478bd9Sstevel@tonic-gate * as to avoid deadlocking with someone else who is 35657c478bd9Sstevel@tonic-gate * waiting for writer lock. With the lock gone, we 35667c478bd9Sstevel@tonic-gate * cannot be sure the checks done above will hold 35677c478bd9Sstevel@tonic-gate * good when we eventually get them back as writer. 35687c478bd9Sstevel@tonic-gate * So if we can't upgrade we drop the locks and retry 35697c478bd9Sstevel@tonic-gate * everything again. 35707c478bd9Sstevel@tonic-gate */ 35717c478bd9Sstevel@tonic-gate rw_exit(&tdp->i_rwlock); 35727c478bd9Sstevel@tonic-gate if (tdp != sdp) 35737c478bd9Sstevel@tonic-gate rw_exit(&sdp->i_rwlock); 35747c478bd9Sstevel@tonic-gate delay(ufs_rename_backoff_delay); 35757c478bd9Sstevel@tonic-gate ufs_rename_upgrade_retry_cnt++; 35767c478bd9Sstevel@tonic-gate goto retry; 35777c478bd9Sstevel@tonic-gate } 35787c478bd9Sstevel@tonic-gate if (tdp != sdp) { 35797c478bd9Sstevel@tonic-gate if (!rw_tryupgrade(&sdp->i_rwlock)) { 35807c478bd9Sstevel@tonic-gate /* 35817c478bd9Sstevel@tonic-gate * The upgrade failed. We got to give away the lock 35827c478bd9Sstevel@tonic-gate * as to avoid deadlocking with someone else who is 35837c478bd9Sstevel@tonic-gate * waiting for writer lock. With the lock gone, we 35847c478bd9Sstevel@tonic-gate * cannot be sure the checks done above will hold 35857c478bd9Sstevel@tonic-gate * good when we eventually get them back as writer. 35867c478bd9Sstevel@tonic-gate * So if we can't upgrade we drop the locks and retry 35877c478bd9Sstevel@tonic-gate * everything again. 35887c478bd9Sstevel@tonic-gate */ 35897c478bd9Sstevel@tonic-gate rw_exit(&tdp->i_rwlock); 35907c478bd9Sstevel@tonic-gate rw_exit(&sdp->i_rwlock); 35917c478bd9Sstevel@tonic-gate delay(ufs_rename_backoff_delay); 35927c478bd9Sstevel@tonic-gate ufs_rename_upgrade_retry_cnt++; 35937c478bd9Sstevel@tonic-gate goto retry; 35947c478bd9Sstevel@tonic-gate } 35957c478bd9Sstevel@tonic-gate } 3596baa4d099Sswilcox 3597baa4d099Sswilcox /* 3598baa4d099Sswilcox * Now that all the locks are held check to make sure another thread 3599baa4d099Sswilcox * didn't slip in and take out the sip. 3600baa4d099Sswilcox */ 3601baa4d099Sswilcox slot.status = NONE; 3602baa4d099Sswilcox if ((sip->i_ctime.tv_usec * 1000) > now.tv_nsec || 3603baa4d099Sswilcox sip->i_ctime.tv_sec > now.tv_sec) { 3604baa4d099Sswilcox rw_enter(&sdp->i_ufsvfs->vfs_dqrwlock, RW_READER); 3605baa4d099Sswilcox rw_enter(&sdp->i_contents, RW_WRITER); 3606baa4d099Sswilcox error = ufs_dircheckforname(sdp, snm, strlen(snm), &slot, 3607baa4d099Sswilcox &ip, cr, 0); 3608baa4d099Sswilcox rw_exit(&sdp->i_contents); 3609baa4d099Sswilcox rw_exit(&sdp->i_ufsvfs->vfs_dqrwlock); 3610baa4d099Sswilcox if (error) { 3611baa4d099Sswilcox goto errout; 3612baa4d099Sswilcox } 3613baa4d099Sswilcox if (ip == NULL) { 3614baa4d099Sswilcox error = ENOENT; 3615baa4d099Sswilcox goto errout; 3616baa4d099Sswilcox } else { 3617baa4d099Sswilcox /* 3618baa4d099Sswilcox * If the inode was found need to drop the v_count 3619baa4d099Sswilcox * so as not to keep the filesystem from being 3620baa4d099Sswilcox * unmounted at a later time. 3621baa4d099Sswilcox */ 3622baa4d099Sswilcox VN_RELE(ITOV(ip)); 3623baa4d099Sswilcox } 3624baa4d099Sswilcox 3625baa4d099Sswilcox /* 3626baa4d099Sswilcox * Release the slot.fbp that has the page mapped and 3627baa4d099Sswilcox * locked SE_SHARED, and could be used in in 3628baa4d099Sswilcox * ufs_direnter_lr() which needs to get the SE_EXCL lock 3629baa4d099Sswilcox * on said page. 3630baa4d099Sswilcox */ 3631baa4d099Sswilcox if (slot.fbp) { 3632baa4d099Sswilcox fbrelse(slot.fbp, S_OTHER); 3633baa4d099Sswilcox slot.fbp = NULL; 3634baa4d099Sswilcox } 3635baa4d099Sswilcox } 3636baa4d099Sswilcox 36377c478bd9Sstevel@tonic-gate /* 36387c478bd9Sstevel@tonic-gate * Link source to the target. If a target exists, return its 36397c478bd9Sstevel@tonic-gate * vnode pointer in tvp. We'll release it after sending the 36407c478bd9Sstevel@tonic-gate * vnevent. 36417c478bd9Sstevel@tonic-gate */ 36427c478bd9Sstevel@tonic-gate if (error = ufs_direnter_lr(tdp, tnm, DE_RENAME, sdp, sip, cr, &tvp)) { 36437c478bd9Sstevel@tonic-gate /* 36447c478bd9Sstevel@tonic-gate * ESAME isn't really an error; it indicates that the 36457c478bd9Sstevel@tonic-gate * operation should not be done because the source and target 36467c478bd9Sstevel@tonic-gate * are the same file, but that no error should be reported. 36477c478bd9Sstevel@tonic-gate */ 36487c478bd9Sstevel@tonic-gate if (error == ESAME) 36497c478bd9Sstevel@tonic-gate error = 0; 36507c478bd9Sstevel@tonic-gate goto errout; 36517c478bd9Sstevel@tonic-gate } 36527c478bd9Sstevel@tonic-gate 36537c478bd9Sstevel@tonic-gate /* 36547c478bd9Sstevel@tonic-gate * Unlink the source. 36557c478bd9Sstevel@tonic-gate * Remove the source entry. ufs_dirremove() checks that the entry 36567c478bd9Sstevel@tonic-gate * still reflects sip, and returns an error if it doesn't. 36577c478bd9Sstevel@tonic-gate * If the entry has changed just forget about it. Release 36587c478bd9Sstevel@tonic-gate * the source inode. 36597c478bd9Sstevel@tonic-gate */ 36607c478bd9Sstevel@tonic-gate if ((error = ufs_dirremove(sdp, snm, sip, (struct vnode *)0, 36617c478bd9Sstevel@tonic-gate DR_RENAME, cr, NULL)) == ENOENT) 36627c478bd9Sstevel@tonic-gate error = 0; 36637c478bd9Sstevel@tonic-gate 36647c478bd9Sstevel@tonic-gate errout: 3665baa4d099Sswilcox if (slot.fbp) 3666baa4d099Sswilcox fbrelse(slot.fbp, S_OTHER); 3667baa4d099Sswilcox 36687c478bd9Sstevel@tonic-gate rw_exit(&tdp->i_rwlock); 36697c478bd9Sstevel@tonic-gate if (sdp != tdp) { 36707c478bd9Sstevel@tonic-gate rw_exit(&sdp->i_rwlock); 36717c478bd9Sstevel@tonic-gate } 36727c478bd9Sstevel@tonic-gate 36737c478bd9Sstevel@tonic-gate unlock: 36747c478bd9Sstevel@tonic-gate if (ulp) { 36757c478bd9Sstevel@tonic-gate TRANS_END_CSYNC(ufsvfsp, error, issync, TOP_RENAME, trans_size); 36767c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 36777c478bd9Sstevel@tonic-gate } 36787c478bd9Sstevel@tonic-gate 36797c478bd9Sstevel@tonic-gate /* 36807c478bd9Sstevel@tonic-gate * If no errors, send the appropriate events on the source 36817c478bd9Sstevel@tonic-gate * and destination (a.k.a, target) vnodes, if they exist. 36827c478bd9Sstevel@tonic-gate * This has to be done after the rename transaction has closed. 36837c478bd9Sstevel@tonic-gate */ 36847c478bd9Sstevel@tonic-gate if (error == 0) { 36857c478bd9Sstevel@tonic-gate if (tvp != NULL) 36867c478bd9Sstevel@tonic-gate vnevent_rename_dest(tvp); 36877c478bd9Sstevel@tonic-gate /* 36887c478bd9Sstevel@tonic-gate * Note that if ufs_direnter_lr() returned ESAME then 36897c478bd9Sstevel@tonic-gate * this event will still be sent. This isn't expected 36907c478bd9Sstevel@tonic-gate * to be a problem for anticipated usage by consumers. 36917c478bd9Sstevel@tonic-gate */ 36927c478bd9Sstevel@tonic-gate if (sip != NULL) 36937c478bd9Sstevel@tonic-gate vnevent_rename_src(ITOV(sip)); 36947c478bd9Sstevel@tonic-gate } 36957c478bd9Sstevel@tonic-gate 36967c478bd9Sstevel@tonic-gate if (tvp != NULL) 36977c478bd9Sstevel@tonic-gate VN_RELE(tvp); 36987c478bd9Sstevel@tonic-gate 36997c478bd9Sstevel@tonic-gate if (sip != NULL) 37007c478bd9Sstevel@tonic-gate VN_RELE(ITOV(sip)); 37017c478bd9Sstevel@tonic-gate 37027c478bd9Sstevel@tonic-gate out: 37037c478bd9Sstevel@tonic-gate TRACE_5(TR_FAC_UFS, TR_UFS_RENAME_END, 37047c478bd9Sstevel@tonic-gate "ufs_rename_end:sdvp %p snm %s tdvp %p tnm %s error %d", 37057c478bd9Sstevel@tonic-gate sdvp, snm, tdvp, tnm, error); 37067c478bd9Sstevel@tonic-gate return (error); 37077c478bd9Sstevel@tonic-gate } 37087c478bd9Sstevel@tonic-gate 37097c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 37107c478bd9Sstevel@tonic-gate static int 37117c478bd9Sstevel@tonic-gate ufs_mkdir(struct vnode *dvp, char *dirname, struct vattr *vap, 37127c478bd9Sstevel@tonic-gate struct vnode **vpp, struct cred *cr) 37137c478bd9Sstevel@tonic-gate { 37147c478bd9Sstevel@tonic-gate struct inode *ip; 37157c478bd9Sstevel@tonic-gate struct inode *xip; 37167c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp; 37177c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 37187c478bd9Sstevel@tonic-gate int error; 37197c478bd9Sstevel@tonic-gate int issync; 37207c478bd9Sstevel@tonic-gate int trans_size; 3721*02ff05a9Svsakar int indeadlock; 37227c478bd9Sstevel@tonic-gate int retry = 1; 37237c478bd9Sstevel@tonic-gate 37247c478bd9Sstevel@tonic-gate ASSERT((vap->va_mask & (AT_TYPE|AT_MODE)) == (AT_TYPE|AT_MODE)); 37257c478bd9Sstevel@tonic-gate 37267c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_MKDIR_START, 37277c478bd9Sstevel@tonic-gate "ufs_mkdir_start:dvp %p", dvp); 37287c478bd9Sstevel@tonic-gate 37297c478bd9Sstevel@tonic-gate /* 37307c478bd9Sstevel@tonic-gate * Can't make directory in attr hidden dir 37317c478bd9Sstevel@tonic-gate */ 37327c478bd9Sstevel@tonic-gate if ((VTOI(dvp)->i_mode & IFMT) == IFATTRDIR) 37337c478bd9Sstevel@tonic-gate return (EINVAL); 37347c478bd9Sstevel@tonic-gate 37357c478bd9Sstevel@tonic-gate again: 37367c478bd9Sstevel@tonic-gate ip = VTOI(dvp); 37377c478bd9Sstevel@tonic-gate ufsvfsp = ip->i_ufsvfs; 37387c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_MKDIR_MASK); 37397c478bd9Sstevel@tonic-gate if (error) 37407c478bd9Sstevel@tonic-gate goto out; 37417c478bd9Sstevel@tonic-gate if (ulp) 37427c478bd9Sstevel@tonic-gate TRANS_BEGIN_CSYNC(ufsvfsp, issync, TOP_MKDIR, 37437c478bd9Sstevel@tonic-gate trans_size = (int)TOP_MKDIR_SIZE(ip)); 37447c478bd9Sstevel@tonic-gate 3745*02ff05a9Svsakar /* 3746*02ff05a9Svsakar * ufs_tryirwlock_trans uses rw_tryenter and checks for SLOCK 3747*02ff05a9Svsakar * to avoid i_rwlock, ufs_lockfs_begin deadlock. If deadlock 3748*02ff05a9Svsakar * possible, retries the operation. 3749*02ff05a9Svsakar */ 3750*02ff05a9Svsakar ufs_tryirwlock_trans(&ip->i_rwlock, RW_WRITER, TOP_MKDIR, retry); 3751*02ff05a9Svsakar if (indeadlock) 3752*02ff05a9Svsakar goto again; 37537c478bd9Sstevel@tonic-gate 37547c478bd9Sstevel@tonic-gate error = ufs_direnter_cm(ip, dirname, DE_MKDIR, vap, &xip, cr, 37557c478bd9Sstevel@tonic-gate (retry ? IQUIET : 0)); 3756*02ff05a9Svsakar if (error == EAGAIN) { 3757*02ff05a9Svsakar if (ulp) { 3758*02ff05a9Svsakar TRANS_END_CSYNC(ufsvfsp, error, issync, TOP_MKDIR, 3759*02ff05a9Svsakar trans_size); 3760*02ff05a9Svsakar ufs_lockfs_end(ulp); 3761*02ff05a9Svsakar } 3762*02ff05a9Svsakar goto again; 3763*02ff05a9Svsakar } 37647c478bd9Sstevel@tonic-gate 37657c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 37667c478bd9Sstevel@tonic-gate if (error == 0) { 37677c478bd9Sstevel@tonic-gate ip = xip; 37687c478bd9Sstevel@tonic-gate *vpp = ITOV(ip); 37697c478bd9Sstevel@tonic-gate } else if (error == EEXIST) 37707c478bd9Sstevel@tonic-gate VN_RELE(ITOV(xip)); 37717c478bd9Sstevel@tonic-gate 37727c478bd9Sstevel@tonic-gate if (ulp) { 37737c478bd9Sstevel@tonic-gate int terr = 0; 37747c478bd9Sstevel@tonic-gate TRANS_END_CSYNC(ufsvfsp, terr, issync, TOP_MKDIR, trans_size); 37757c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 37767c478bd9Sstevel@tonic-gate if (error == 0) 37777c478bd9Sstevel@tonic-gate error = terr; 37787c478bd9Sstevel@tonic-gate } 37797c478bd9Sstevel@tonic-gate out: 37807c478bd9Sstevel@tonic-gate if ((error == ENOSPC) && retry && TRANS_ISTRANS(ufsvfsp)) { 37817c478bd9Sstevel@tonic-gate ufs_delete_drain_wait(ufsvfsp, 1); 37827c478bd9Sstevel@tonic-gate retry = 0; 37837c478bd9Sstevel@tonic-gate goto again; 37847c478bd9Sstevel@tonic-gate } 37857c478bd9Sstevel@tonic-gate 37867c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_MKDIR_END, 37877c478bd9Sstevel@tonic-gate "ufs_mkdir_end:dvp %p error %d", dvp, error); 37887c478bd9Sstevel@tonic-gate return (error); 37897c478bd9Sstevel@tonic-gate } 37907c478bd9Sstevel@tonic-gate 37917c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 37927c478bd9Sstevel@tonic-gate static int 37937c478bd9Sstevel@tonic-gate ufs_rmdir(struct vnode *vp, char *nm, struct vnode *cdir, struct cred *cr) 37947c478bd9Sstevel@tonic-gate { 37957c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 37967c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp = ip->i_ufsvfs; 37977c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 37987c478bd9Sstevel@tonic-gate vnode_t *rmvp = NULL; /* Vnode of removed directory */ 37997c478bd9Sstevel@tonic-gate int error; 38007c478bd9Sstevel@tonic-gate int issync; 3801*02ff05a9Svsakar int trans_size; 3802*02ff05a9Svsakar int indeadlock; 38037c478bd9Sstevel@tonic-gate 38047c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_RMDIR_START, 38057c478bd9Sstevel@tonic-gate "ufs_rmdir_start:vp %p", vp); 38067c478bd9Sstevel@tonic-gate 38077c478bd9Sstevel@tonic-gate /* 38087c478bd9Sstevel@tonic-gate * don't let the delete queue get too long 38097c478bd9Sstevel@tonic-gate */ 38107c478bd9Sstevel@tonic-gate if (ufsvfsp == NULL) { 38117c478bd9Sstevel@tonic-gate error = EIO; 38127c478bd9Sstevel@tonic-gate goto out; 38137c478bd9Sstevel@tonic-gate } 38147c478bd9Sstevel@tonic-gate if (ufsvfsp->vfs_delete.uq_ne > ufs_idle_max) 38157c478bd9Sstevel@tonic-gate ufs_delete_drain(vp->v_vfsp, 1, 1); 38167c478bd9Sstevel@tonic-gate 3817*02ff05a9Svsakar retry_rmdir: 38187c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_RMDIR_MASK); 38197c478bd9Sstevel@tonic-gate if (error) 38207c478bd9Sstevel@tonic-gate goto out; 38217c478bd9Sstevel@tonic-gate 38227c478bd9Sstevel@tonic-gate if (ulp) 3823*02ff05a9Svsakar TRANS_BEGIN_CSYNC(ufsvfsp, issync, TOP_RMDIR, 3824*02ff05a9Svsakar trans_size = TOP_RMDIR_SIZE); 38257c478bd9Sstevel@tonic-gate 3826*02ff05a9Svsakar /* 3827*02ff05a9Svsakar * ufs_tryirwlock_trans uses rw_tryenter and checks for SLOCK 3828*02ff05a9Svsakar * to avoid i_rwlock, ufs_lockfs_begin deadlock. If deadlock 3829*02ff05a9Svsakar * possible, retries the operation. 3830*02ff05a9Svsakar */ 3831*02ff05a9Svsakar ufs_tryirwlock_trans(&ip->i_rwlock, RW_WRITER, TOP_RMDIR, retry); 3832*02ff05a9Svsakar if (indeadlock) 3833*02ff05a9Svsakar goto retry_rmdir; 38347c478bd9Sstevel@tonic-gate error = ufs_dirremove(ip, nm, (struct inode *)0, cdir, DR_RMDIR, cr, 38357c478bd9Sstevel@tonic-gate &rmvp); 38367c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 38377c478bd9Sstevel@tonic-gate 38387c478bd9Sstevel@tonic-gate if (ulp) { 38397c478bd9Sstevel@tonic-gate TRANS_END_CSYNC(ufsvfsp, error, issync, TOP_RMDIR, 3840*02ff05a9Svsakar trans_size); 38417c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 38427c478bd9Sstevel@tonic-gate } 38437c478bd9Sstevel@tonic-gate 38447c478bd9Sstevel@tonic-gate /* 38457c478bd9Sstevel@tonic-gate * This must be done AFTER the rmdir transaction has closed. 38467c478bd9Sstevel@tonic-gate */ 38477c478bd9Sstevel@tonic-gate if (rmvp != NULL) { 38487c478bd9Sstevel@tonic-gate /* Only send the event if there were no errors */ 38497c478bd9Sstevel@tonic-gate if (error == 0) 38507c478bd9Sstevel@tonic-gate vnevent_rmdir(rmvp); 38517c478bd9Sstevel@tonic-gate VN_RELE(rmvp); 38527c478bd9Sstevel@tonic-gate } 38537c478bd9Sstevel@tonic-gate out: 38547c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_RMDIR_END, 38557c478bd9Sstevel@tonic-gate "ufs_rmdir_end:vp %p error %d", vp, error); 38567c478bd9Sstevel@tonic-gate 38577c478bd9Sstevel@tonic-gate return (error); 38587c478bd9Sstevel@tonic-gate } 38597c478bd9Sstevel@tonic-gate 38607c478bd9Sstevel@tonic-gate /* ARGSUSED */ 38617c478bd9Sstevel@tonic-gate static int 38627c478bd9Sstevel@tonic-gate ufs_readdir( 38637c478bd9Sstevel@tonic-gate struct vnode *vp, 38647c478bd9Sstevel@tonic-gate struct uio *uiop, 38657c478bd9Sstevel@tonic-gate struct cred *cr, 38667c478bd9Sstevel@tonic-gate int *eofp) 38677c478bd9Sstevel@tonic-gate { 38687c478bd9Sstevel@tonic-gate struct iovec *iovp; 38697c478bd9Sstevel@tonic-gate struct inode *ip; 38707c478bd9Sstevel@tonic-gate struct direct *idp; 38717c478bd9Sstevel@tonic-gate struct dirent64 *odp; 38727c478bd9Sstevel@tonic-gate struct fbuf *fbp; 38737c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp; 38747c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 38757c478bd9Sstevel@tonic-gate caddr_t outbuf; 38767c478bd9Sstevel@tonic-gate size_t bufsize; 38777c478bd9Sstevel@tonic-gate uint_t offset; 38787c478bd9Sstevel@tonic-gate uint_t bytes_wanted, total_bytes_wanted; 38797c478bd9Sstevel@tonic-gate int incount = 0; 38807c478bd9Sstevel@tonic-gate int outcount = 0; 38817c478bd9Sstevel@tonic-gate int error; 38827c478bd9Sstevel@tonic-gate 38837c478bd9Sstevel@tonic-gate ip = VTOI(vp); 38847c478bd9Sstevel@tonic-gate ASSERT(RW_READ_HELD(&ip->i_rwlock)); 38857c478bd9Sstevel@tonic-gate 38867c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_READDIR_START, 38877c478bd9Sstevel@tonic-gate "ufs_readdir_start:vp %p uiop %p", vp, uiop); 38887c478bd9Sstevel@tonic-gate 38897c478bd9Sstevel@tonic-gate if (uiop->uio_loffset >= MAXOFF32_T) { 38907c478bd9Sstevel@tonic-gate if (eofp) 38917c478bd9Sstevel@tonic-gate *eofp = 1; 38927c478bd9Sstevel@tonic-gate return (0); 38937c478bd9Sstevel@tonic-gate } 38947c478bd9Sstevel@tonic-gate 38957c478bd9Sstevel@tonic-gate /* 38967c478bd9Sstevel@tonic-gate * Check if we have been called with a valid iov_len 38977c478bd9Sstevel@tonic-gate * and bail out if not, otherwise we may potentially loop 38987c478bd9Sstevel@tonic-gate * forever further down. 38997c478bd9Sstevel@tonic-gate */ 39007c478bd9Sstevel@tonic-gate if (uiop->uio_iov->iov_len <= 0) { 39017c478bd9Sstevel@tonic-gate error = EINVAL; 39027c478bd9Sstevel@tonic-gate goto out; 39037c478bd9Sstevel@tonic-gate } 39047c478bd9Sstevel@tonic-gate 39057c478bd9Sstevel@tonic-gate /* 39067c478bd9Sstevel@tonic-gate * Large Files: When we come here we are guaranteed that 39077c478bd9Sstevel@tonic-gate * uio_offset can be used safely. The high word is zero. 39087c478bd9Sstevel@tonic-gate */ 39097c478bd9Sstevel@tonic-gate 39107c478bd9Sstevel@tonic-gate ufsvfsp = ip->i_ufsvfs; 39117c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_READDIR_MASK); 39127c478bd9Sstevel@tonic-gate if (error) 39137c478bd9Sstevel@tonic-gate goto out; 39147c478bd9Sstevel@tonic-gate 39157c478bd9Sstevel@tonic-gate iovp = uiop->uio_iov; 39167c478bd9Sstevel@tonic-gate total_bytes_wanted = iovp->iov_len; 39177c478bd9Sstevel@tonic-gate 39187c478bd9Sstevel@tonic-gate /* Large Files: directory files should not be "large" */ 39197c478bd9Sstevel@tonic-gate 39207c478bd9Sstevel@tonic-gate ASSERT(ip->i_size <= MAXOFF32_T); 39217c478bd9Sstevel@tonic-gate 39227c478bd9Sstevel@tonic-gate /* Force offset to be valid (to guard against bogus lseek() values) */ 39237c478bd9Sstevel@tonic-gate offset = (uint_t)uiop->uio_offset & ~(DIRBLKSIZ - 1); 39247c478bd9Sstevel@tonic-gate 39257c478bd9Sstevel@tonic-gate /* Quit if at end of file or link count of zero (posix) */ 39267c478bd9Sstevel@tonic-gate if (offset >= (uint_t)ip->i_size || ip->i_nlink <= 0) { 39277c478bd9Sstevel@tonic-gate if (eofp) 39287c478bd9Sstevel@tonic-gate *eofp = 1; 39297c478bd9Sstevel@tonic-gate error = 0; 39307c478bd9Sstevel@tonic-gate goto unlock; 39317c478bd9Sstevel@tonic-gate } 39327c478bd9Sstevel@tonic-gate 39337c478bd9Sstevel@tonic-gate /* 39347c478bd9Sstevel@tonic-gate * Get space to change directory entries into fs independent format. 39357c478bd9Sstevel@tonic-gate * Do fast alloc for the most commonly used-request size (filesystem 39367c478bd9Sstevel@tonic-gate * block size). 39377c478bd9Sstevel@tonic-gate */ 39387c478bd9Sstevel@tonic-gate if (uiop->uio_segflg != UIO_SYSSPACE || uiop->uio_iovcnt != 1) { 39397c478bd9Sstevel@tonic-gate bufsize = total_bytes_wanted; 39407c478bd9Sstevel@tonic-gate outbuf = kmem_alloc(bufsize, KM_SLEEP); 39417c478bd9Sstevel@tonic-gate odp = (struct dirent64 *)outbuf; 39427c478bd9Sstevel@tonic-gate } else { 39437c478bd9Sstevel@tonic-gate bufsize = total_bytes_wanted; 39447c478bd9Sstevel@tonic-gate odp = (struct dirent64 *)iovp->iov_base; 39457c478bd9Sstevel@tonic-gate } 39467c478bd9Sstevel@tonic-gate 39477c478bd9Sstevel@tonic-gate nextblk: 39487c478bd9Sstevel@tonic-gate bytes_wanted = total_bytes_wanted; 39497c478bd9Sstevel@tonic-gate 39507c478bd9Sstevel@tonic-gate /* Truncate request to file size */ 39517c478bd9Sstevel@tonic-gate if (offset + bytes_wanted > (int)ip->i_size) 39527c478bd9Sstevel@tonic-gate bytes_wanted = (int)(ip->i_size - offset); 39537c478bd9Sstevel@tonic-gate 39547c478bd9Sstevel@tonic-gate /* Comply with MAXBSIZE boundary restrictions of fbread() */ 39557c478bd9Sstevel@tonic-gate if ((offset & MAXBOFFSET) + bytes_wanted > MAXBSIZE) 39567c478bd9Sstevel@tonic-gate bytes_wanted = MAXBSIZE - (offset & MAXBOFFSET); 39577c478bd9Sstevel@tonic-gate 39587c478bd9Sstevel@tonic-gate /* 39597c478bd9Sstevel@tonic-gate * Read in the next chunk. 39607c478bd9Sstevel@tonic-gate * We are still holding the i_rwlock. 39617c478bd9Sstevel@tonic-gate */ 39627c478bd9Sstevel@tonic-gate error = fbread(vp, (offset_t)offset, bytes_wanted, S_OTHER, &fbp); 39637c478bd9Sstevel@tonic-gate 39647c478bd9Sstevel@tonic-gate if (error) 39657c478bd9Sstevel@tonic-gate goto update_inode; 39667c478bd9Sstevel@tonic-gate if (!ULOCKFS_IS_NOIACC(ITOUL(ip)) && (ip->i_fs->fs_ronly == 0) && 39677c478bd9Sstevel@tonic-gate (!ufsvfsp->vfs_noatime)) { 39687c478bd9Sstevel@tonic-gate ip->i_flag |= IACC; 39697c478bd9Sstevel@tonic-gate } 39707c478bd9Sstevel@tonic-gate incount = 0; 39717c478bd9Sstevel@tonic-gate idp = (struct direct *)fbp->fb_addr; 39727c478bd9Sstevel@tonic-gate if (idp->d_ino == 0 && idp->d_reclen == 0 && 39737c478bd9Sstevel@tonic-gate idp->d_namlen == 0) { 39747c478bd9Sstevel@tonic-gate cmn_err(CE_WARN, "ufs_readdir: bad dir, inumber = %llu, " 39757c478bd9Sstevel@tonic-gate "fs = %s\n", 39767c478bd9Sstevel@tonic-gate (u_longlong_t)ip->i_number, ufsvfsp->vfs_fs->fs_fsmnt); 39777c478bd9Sstevel@tonic-gate fbrelse(fbp, S_OTHER); 39787c478bd9Sstevel@tonic-gate error = ENXIO; 39797c478bd9Sstevel@tonic-gate goto update_inode; 39807c478bd9Sstevel@tonic-gate } 39817c478bd9Sstevel@tonic-gate /* Transform to file-system independent format */ 39827c478bd9Sstevel@tonic-gate while (incount < bytes_wanted) { 39837c478bd9Sstevel@tonic-gate /* 39847c478bd9Sstevel@tonic-gate * If the current directory entry is mangled, then skip 39857c478bd9Sstevel@tonic-gate * to the next block. It would be nice to set the FSBAD 39867c478bd9Sstevel@tonic-gate * flag in the super-block so that a fsck is forced on 39877c478bd9Sstevel@tonic-gate * next reboot, but locking is a problem. 39887c478bd9Sstevel@tonic-gate */ 39897c478bd9Sstevel@tonic-gate if (idp->d_reclen & 0x3) { 39907c478bd9Sstevel@tonic-gate offset = (offset + DIRBLKSIZ) & ~(DIRBLKSIZ-1); 39917c478bd9Sstevel@tonic-gate break; 39927c478bd9Sstevel@tonic-gate } 39937c478bd9Sstevel@tonic-gate 39947c478bd9Sstevel@tonic-gate /* Skip to requested offset and skip empty entries */ 39957c478bd9Sstevel@tonic-gate if (idp->d_ino != 0 && offset >= (uint_t)uiop->uio_offset) { 39967c478bd9Sstevel@tonic-gate ushort_t this_reclen = 39977c478bd9Sstevel@tonic-gate DIRENT64_RECLEN(idp->d_namlen); 39987c478bd9Sstevel@tonic-gate /* Buffer too small for any entries */ 39997c478bd9Sstevel@tonic-gate if (!outcount && this_reclen > bufsize) { 40007c478bd9Sstevel@tonic-gate fbrelse(fbp, S_OTHER); 40017c478bd9Sstevel@tonic-gate error = EINVAL; 40027c478bd9Sstevel@tonic-gate goto update_inode; 40037c478bd9Sstevel@tonic-gate } 40047c478bd9Sstevel@tonic-gate /* If would overrun the buffer, quit */ 40057c478bd9Sstevel@tonic-gate if (outcount + this_reclen > bufsize) { 40067c478bd9Sstevel@tonic-gate break; 40077c478bd9Sstevel@tonic-gate } 40087c478bd9Sstevel@tonic-gate /* Take this entry */ 40097c478bd9Sstevel@tonic-gate odp->d_ino = (ino64_t)idp->d_ino; 40107c478bd9Sstevel@tonic-gate odp->d_reclen = (ushort_t)this_reclen; 40117c478bd9Sstevel@tonic-gate odp->d_off = (offset_t)(offset + idp->d_reclen); 40127c478bd9Sstevel@tonic-gate 40137c478bd9Sstevel@tonic-gate /* use strncpy(9f) to zero out uninitialized bytes */ 40147c478bd9Sstevel@tonic-gate 40157c478bd9Sstevel@tonic-gate ASSERT(strlen(idp->d_name) + 1 <= 40167c478bd9Sstevel@tonic-gate DIRENT64_NAMELEN(this_reclen)); 40177c478bd9Sstevel@tonic-gate (void) strncpy(odp->d_name, idp->d_name, 40187c478bd9Sstevel@tonic-gate DIRENT64_NAMELEN(this_reclen)); 40197c478bd9Sstevel@tonic-gate outcount += odp->d_reclen; 40207c478bd9Sstevel@tonic-gate odp = (struct dirent64 *)((intptr_t)odp + 40217c478bd9Sstevel@tonic-gate odp->d_reclen); 40227c478bd9Sstevel@tonic-gate ASSERT(outcount <= bufsize); 40237c478bd9Sstevel@tonic-gate } 40247c478bd9Sstevel@tonic-gate if (idp->d_reclen) { 40257c478bd9Sstevel@tonic-gate incount += idp->d_reclen; 40267c478bd9Sstevel@tonic-gate offset += idp->d_reclen; 40277c478bd9Sstevel@tonic-gate idp = (struct direct *)((intptr_t)idp + idp->d_reclen); 40287c478bd9Sstevel@tonic-gate } else { 40297c478bd9Sstevel@tonic-gate offset = (offset + DIRBLKSIZ) & ~(DIRBLKSIZ-1); 40307c478bd9Sstevel@tonic-gate break; 40317c478bd9Sstevel@tonic-gate } 40327c478bd9Sstevel@tonic-gate } 40337c478bd9Sstevel@tonic-gate /* Release the chunk */ 40347c478bd9Sstevel@tonic-gate fbrelse(fbp, S_OTHER); 40357c478bd9Sstevel@tonic-gate 40367c478bd9Sstevel@tonic-gate /* Read whole block, but got no entries, read another if not eof */ 40377c478bd9Sstevel@tonic-gate 40387c478bd9Sstevel@tonic-gate /* 40397c478bd9Sstevel@tonic-gate * Large Files: casting i_size to int here is not a problem 40407c478bd9Sstevel@tonic-gate * because directory sizes are always less than MAXOFF32_T. 40417c478bd9Sstevel@tonic-gate * See assertion above. 40427c478bd9Sstevel@tonic-gate */ 40437c478bd9Sstevel@tonic-gate 40447c478bd9Sstevel@tonic-gate if (offset < (int)ip->i_size && !outcount) 40457c478bd9Sstevel@tonic-gate goto nextblk; 40467c478bd9Sstevel@tonic-gate 40477c478bd9Sstevel@tonic-gate /* Copy out the entry data */ 40487c478bd9Sstevel@tonic-gate if (uiop->uio_segflg == UIO_SYSSPACE && uiop->uio_iovcnt == 1) { 40497c478bd9Sstevel@tonic-gate iovp->iov_base += outcount; 40507c478bd9Sstevel@tonic-gate iovp->iov_len -= outcount; 40517c478bd9Sstevel@tonic-gate uiop->uio_resid -= outcount; 40527c478bd9Sstevel@tonic-gate uiop->uio_offset = offset; 40537c478bd9Sstevel@tonic-gate } else if ((error = uiomove(outbuf, (long)outcount, UIO_READ, 40547c478bd9Sstevel@tonic-gate uiop)) == 0) 40557c478bd9Sstevel@tonic-gate uiop->uio_offset = offset; 40567c478bd9Sstevel@tonic-gate update_inode: 40577c478bd9Sstevel@tonic-gate ITIMES(ip); 40587c478bd9Sstevel@tonic-gate if (uiop->uio_segflg != UIO_SYSSPACE || uiop->uio_iovcnt != 1) 40597c478bd9Sstevel@tonic-gate kmem_free(outbuf, bufsize); 40607c478bd9Sstevel@tonic-gate 40617c478bd9Sstevel@tonic-gate if (eofp && error == 0) 40627c478bd9Sstevel@tonic-gate *eofp = (uiop->uio_offset >= (int)ip->i_size); 40637c478bd9Sstevel@tonic-gate unlock: 40647c478bd9Sstevel@tonic-gate if (ulp) { 40657c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 40667c478bd9Sstevel@tonic-gate } 40677c478bd9Sstevel@tonic-gate out: 40687c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_READDIR_END, 40697c478bd9Sstevel@tonic-gate "ufs_readdir_end:vp %p error %d", vp, error); 40707c478bd9Sstevel@tonic-gate return (error); 40717c478bd9Sstevel@tonic-gate } 40727c478bd9Sstevel@tonic-gate 40737c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 40747c478bd9Sstevel@tonic-gate static int 40757c478bd9Sstevel@tonic-gate ufs_symlink( 40767c478bd9Sstevel@tonic-gate struct vnode *dvp, /* ptr to parent dir vnode */ 40777c478bd9Sstevel@tonic-gate char *linkname, /* name of symbolic link */ 40787c478bd9Sstevel@tonic-gate struct vattr *vap, /* attributes */ 40797c478bd9Sstevel@tonic-gate char *target, /* target path */ 40807c478bd9Sstevel@tonic-gate struct cred *cr) /* user credentials */ 40817c478bd9Sstevel@tonic-gate { 40827c478bd9Sstevel@tonic-gate struct inode *ip, *dip = VTOI(dvp); 40837c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp = dip->i_ufsvfs; 40847c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 40857c478bd9Sstevel@tonic-gate int error; 40867c478bd9Sstevel@tonic-gate int issync; 40877c478bd9Sstevel@tonic-gate int trans_size; 40887c478bd9Sstevel@tonic-gate int residual; 40897c478bd9Sstevel@tonic-gate int ioflag; 40907c478bd9Sstevel@tonic-gate int retry = 1; 40917c478bd9Sstevel@tonic-gate 40927c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_SYMLINK_START, 40937c478bd9Sstevel@tonic-gate "ufs_symlink_start:dvp %p", dvp); 40947c478bd9Sstevel@tonic-gate 40957c478bd9Sstevel@tonic-gate /* 40967c478bd9Sstevel@tonic-gate * No symlinks in attrdirs at this time 40977c478bd9Sstevel@tonic-gate */ 40987c478bd9Sstevel@tonic-gate if ((VTOI(dvp)->i_mode & IFMT) == IFATTRDIR) 40997c478bd9Sstevel@tonic-gate return (EINVAL); 41007c478bd9Sstevel@tonic-gate 41017c478bd9Sstevel@tonic-gate again: 41027c478bd9Sstevel@tonic-gate ip = (struct inode *)NULL; 41037c478bd9Sstevel@tonic-gate vap->va_type = VLNK; 41047c478bd9Sstevel@tonic-gate vap->va_rdev = 0; 41057c478bd9Sstevel@tonic-gate 41067c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_SYMLINK_MASK); 41077c478bd9Sstevel@tonic-gate if (error) 41087c478bd9Sstevel@tonic-gate goto out; 41097c478bd9Sstevel@tonic-gate 41107c478bd9Sstevel@tonic-gate if (ulp) 41117c478bd9Sstevel@tonic-gate TRANS_BEGIN_CSYNC(ufsvfsp, issync, TOP_SYMLINK, 41127c478bd9Sstevel@tonic-gate trans_size = (int)TOP_SYMLINK_SIZE(dip)); 41137c478bd9Sstevel@tonic-gate 41147c478bd9Sstevel@tonic-gate /* 41157c478bd9Sstevel@tonic-gate * We must create the inode before the directory entry, to avoid 41167c478bd9Sstevel@tonic-gate * racing with readlink(). ufs_dirmakeinode requires that we 41177c478bd9Sstevel@tonic-gate * hold the quota lock as reader, and directory locks as writer. 41187c478bd9Sstevel@tonic-gate */ 41197c478bd9Sstevel@tonic-gate 41207c478bd9Sstevel@tonic-gate rw_enter(&dip->i_rwlock, RW_WRITER); 41217c478bd9Sstevel@tonic-gate rw_enter(&ufsvfsp->vfs_dqrwlock, RW_READER); 41227c478bd9Sstevel@tonic-gate rw_enter(&dip->i_contents, RW_WRITER); 41237c478bd9Sstevel@tonic-gate 41247c478bd9Sstevel@tonic-gate /* 41257c478bd9Sstevel@tonic-gate * Suppress any out of inodes messages if we will retry on 41267c478bd9Sstevel@tonic-gate * ENOSP 41277c478bd9Sstevel@tonic-gate */ 41287c478bd9Sstevel@tonic-gate if (retry) 41297c478bd9Sstevel@tonic-gate dip->i_flag |= IQUIET; 41307c478bd9Sstevel@tonic-gate 41317c478bd9Sstevel@tonic-gate error = ufs_dirmakeinode(dip, &ip, vap, DE_SYMLINK, cr); 41327c478bd9Sstevel@tonic-gate 41337c478bd9Sstevel@tonic-gate dip->i_flag &= ~IQUIET; 41347c478bd9Sstevel@tonic-gate 41357c478bd9Sstevel@tonic-gate rw_exit(&dip->i_contents); 41367c478bd9Sstevel@tonic-gate rw_exit(&ufsvfsp->vfs_dqrwlock); 41377c478bd9Sstevel@tonic-gate rw_exit(&dip->i_rwlock); 41387c478bd9Sstevel@tonic-gate 41397c478bd9Sstevel@tonic-gate if (error) 41407c478bd9Sstevel@tonic-gate goto unlock; 41417c478bd9Sstevel@tonic-gate 41427c478bd9Sstevel@tonic-gate /* 41437c478bd9Sstevel@tonic-gate * OK. The inode has been created. Write out the data of the 41447c478bd9Sstevel@tonic-gate * symbolic link. Since symbolic links are metadata, and should 41457c478bd9Sstevel@tonic-gate * remain consistent across a system crash, we need to force the 41467c478bd9Sstevel@tonic-gate * data out synchronously. 41477c478bd9Sstevel@tonic-gate * 41487c478bd9Sstevel@tonic-gate * (This is a change from the semantics in earlier releases, which 41497c478bd9Sstevel@tonic-gate * only created symbolic links synchronously if the semi-documented 41507c478bd9Sstevel@tonic-gate * 'syncdir' option was set, or if we were being invoked by the NFS 41517c478bd9Sstevel@tonic-gate * server, which requires symbolic links to be created synchronously.) 41527c478bd9Sstevel@tonic-gate * 41537c478bd9Sstevel@tonic-gate * We need to pass in a pointer for the residual length; otherwise 41547c478bd9Sstevel@tonic-gate * ufs_rdwri() will always return EIO if it can't write the data, 41557c478bd9Sstevel@tonic-gate * even if the error was really ENOSPC or EDQUOT. 41567c478bd9Sstevel@tonic-gate */ 41577c478bd9Sstevel@tonic-gate 41587c478bd9Sstevel@tonic-gate ioflag = FWRITE | FDSYNC; 41597c478bd9Sstevel@tonic-gate residual = 0; 41607c478bd9Sstevel@tonic-gate 41617c478bd9Sstevel@tonic-gate rw_enter(&ufsvfsp->vfs_dqrwlock, RW_READER); 41627c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_WRITER); 41637c478bd9Sstevel@tonic-gate 41647c478bd9Sstevel@tonic-gate /* 41657c478bd9Sstevel@tonic-gate * Suppress file system full messages if we will retry 41667c478bd9Sstevel@tonic-gate */ 41677c478bd9Sstevel@tonic-gate if (retry) 41687c478bd9Sstevel@tonic-gate ip->i_flag |= IQUIET; 41697c478bd9Sstevel@tonic-gate 41707c478bd9Sstevel@tonic-gate error = ufs_rdwri(UIO_WRITE, ioflag, ip, target, strlen(target), 41717c478bd9Sstevel@tonic-gate (offset_t)0, UIO_SYSSPACE, &residual, cr); 41727c478bd9Sstevel@tonic-gate 41737c478bd9Sstevel@tonic-gate ip->i_flag &= ~IQUIET; 41747c478bd9Sstevel@tonic-gate 41757c478bd9Sstevel@tonic-gate if (error) { 41767c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 41777c478bd9Sstevel@tonic-gate rw_exit(&ufsvfsp->vfs_dqrwlock); 41787c478bd9Sstevel@tonic-gate goto remove; 41797c478bd9Sstevel@tonic-gate } 41807c478bd9Sstevel@tonic-gate 41817c478bd9Sstevel@tonic-gate /* 41827c478bd9Sstevel@tonic-gate * If the link's data is small enough, we can cache it in the inode. 41837c478bd9Sstevel@tonic-gate * This is a "fast symbolic link". We don't use the first direct 41847c478bd9Sstevel@tonic-gate * block because that's actually used to point at the symbolic link's 41857c478bd9Sstevel@tonic-gate * contents on disk; but we know that none of the other direct or 41867c478bd9Sstevel@tonic-gate * indirect blocks can be used because symbolic links are restricted 41877c478bd9Sstevel@tonic-gate * to be smaller than a file system block. 41887c478bd9Sstevel@tonic-gate */ 41897c478bd9Sstevel@tonic-gate 41907c478bd9Sstevel@tonic-gate ASSERT(MAXPATHLEN <= VBSIZE(ITOV(ip))); 41917c478bd9Sstevel@tonic-gate 41927c478bd9Sstevel@tonic-gate if (ip->i_size > 0 && ip->i_size <= FSL_SIZE) { 41937c478bd9Sstevel@tonic-gate if (kcopy(target, &ip->i_db[1], ip->i_size) == 0) { 41947c478bd9Sstevel@tonic-gate ip->i_flag |= IFASTSYMLNK; 41957c478bd9Sstevel@tonic-gate } else { 41967c478bd9Sstevel@tonic-gate int i; 41977c478bd9Sstevel@tonic-gate /* error, clear garbage left behind */ 41987c478bd9Sstevel@tonic-gate for (i = 1; i < NDADDR; i++) 41997c478bd9Sstevel@tonic-gate ip->i_db[i] = 0; 42007c478bd9Sstevel@tonic-gate for (i = 0; i < NIADDR; i++) 42017c478bd9Sstevel@tonic-gate ip->i_ib[i] = 0; 42027c478bd9Sstevel@tonic-gate } 42037c478bd9Sstevel@tonic-gate } 42047c478bd9Sstevel@tonic-gate 42057c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 42067c478bd9Sstevel@tonic-gate rw_exit(&ufsvfsp->vfs_dqrwlock); 42077c478bd9Sstevel@tonic-gate 42087c478bd9Sstevel@tonic-gate /* 42097c478bd9Sstevel@tonic-gate * OK. We've successfully created the symbolic link. All that 42107c478bd9Sstevel@tonic-gate * remains is to insert it into the appropriate directory. 42117c478bd9Sstevel@tonic-gate */ 42127c478bd9Sstevel@tonic-gate 42137c478bd9Sstevel@tonic-gate rw_enter(&dip->i_rwlock, RW_WRITER); 42147c478bd9Sstevel@tonic-gate error = ufs_direnter_lr(dip, linkname, DE_SYMLINK, NULL, ip, cr, NULL); 42157c478bd9Sstevel@tonic-gate rw_exit(&dip->i_rwlock); 42167c478bd9Sstevel@tonic-gate 42177c478bd9Sstevel@tonic-gate /* 42187c478bd9Sstevel@tonic-gate * Fall through into remove-on-error code. We're either done, or we 42197c478bd9Sstevel@tonic-gate * need to remove the inode (if we couldn't insert it). 42207c478bd9Sstevel@tonic-gate */ 42217c478bd9Sstevel@tonic-gate 42227c478bd9Sstevel@tonic-gate remove: 42237c478bd9Sstevel@tonic-gate if (error && (ip != NULL)) { 42247c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_WRITER); 42257c478bd9Sstevel@tonic-gate ip->i_nlink--; 42267c478bd9Sstevel@tonic-gate ip->i_flag |= ICHG; 42277c478bd9Sstevel@tonic-gate ip->i_seq++; 42287c478bd9Sstevel@tonic-gate ufs_setreclaim(ip); 42297c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 42307c478bd9Sstevel@tonic-gate } 42317c478bd9Sstevel@tonic-gate 42327c478bd9Sstevel@tonic-gate unlock: 42337c478bd9Sstevel@tonic-gate if (ip != NULL) 42347c478bd9Sstevel@tonic-gate VN_RELE(ITOV(ip)); 42357c478bd9Sstevel@tonic-gate 42367c478bd9Sstevel@tonic-gate if (ulp) { 42377c478bd9Sstevel@tonic-gate int terr = 0; 42387c478bd9Sstevel@tonic-gate 42397c478bd9Sstevel@tonic-gate TRANS_END_CSYNC(ufsvfsp, terr, issync, TOP_SYMLINK, 42407c478bd9Sstevel@tonic-gate trans_size); 42417c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 42427c478bd9Sstevel@tonic-gate if (error == 0) 42437c478bd9Sstevel@tonic-gate error = terr; 42447c478bd9Sstevel@tonic-gate } 42457c478bd9Sstevel@tonic-gate 42467c478bd9Sstevel@tonic-gate /* 42477c478bd9Sstevel@tonic-gate * We may have failed due to lack of an inode or of a block to 42487c478bd9Sstevel@tonic-gate * store the target in. Try flushing the delete queue to free 42497c478bd9Sstevel@tonic-gate * logically-available things up and try again. 42507c478bd9Sstevel@tonic-gate */ 42517c478bd9Sstevel@tonic-gate if ((error == ENOSPC) && retry && TRANS_ISTRANS(ufsvfsp)) { 42527c478bd9Sstevel@tonic-gate ufs_delete_drain_wait(ufsvfsp, 1); 42537c478bd9Sstevel@tonic-gate retry = 0; 42547c478bd9Sstevel@tonic-gate goto again; 42557c478bd9Sstevel@tonic-gate } 42567c478bd9Sstevel@tonic-gate 42577c478bd9Sstevel@tonic-gate out: 42587c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_SYMLINK_END, 42597c478bd9Sstevel@tonic-gate "ufs_symlink_end:dvp %p error %d", dvp, error); 42607c478bd9Sstevel@tonic-gate return (error); 42617c478bd9Sstevel@tonic-gate } 42627c478bd9Sstevel@tonic-gate 42637c478bd9Sstevel@tonic-gate /* 42647c478bd9Sstevel@tonic-gate * Ufs specific routine used to do ufs io. 42657c478bd9Sstevel@tonic-gate */ 42667c478bd9Sstevel@tonic-gate int 42677c478bd9Sstevel@tonic-gate ufs_rdwri(enum uio_rw rw, int ioflag, struct inode *ip, caddr_t base, 42687c478bd9Sstevel@tonic-gate ssize_t len, offset_t offset, enum uio_seg seg, int *aresid, 42697c478bd9Sstevel@tonic-gate struct cred *cr) 42707c478bd9Sstevel@tonic-gate { 42717c478bd9Sstevel@tonic-gate struct uio auio; 42727c478bd9Sstevel@tonic-gate struct iovec aiov; 42737c478bd9Sstevel@tonic-gate int error; 42747c478bd9Sstevel@tonic-gate 42757c478bd9Sstevel@tonic-gate ASSERT(RW_LOCK_HELD(&ip->i_contents)); 42767c478bd9Sstevel@tonic-gate 42777c478bd9Sstevel@tonic-gate bzero((caddr_t)&auio, sizeof (uio_t)); 42787c478bd9Sstevel@tonic-gate bzero((caddr_t)&aiov, sizeof (iovec_t)); 42797c478bd9Sstevel@tonic-gate 42807c478bd9Sstevel@tonic-gate aiov.iov_base = base; 42817c478bd9Sstevel@tonic-gate aiov.iov_len = len; 42827c478bd9Sstevel@tonic-gate auio.uio_iov = &aiov; 42837c478bd9Sstevel@tonic-gate auio.uio_iovcnt = 1; 42847c478bd9Sstevel@tonic-gate auio.uio_loffset = offset; 42857c478bd9Sstevel@tonic-gate auio.uio_segflg = (short)seg; 42867c478bd9Sstevel@tonic-gate auio.uio_resid = len; 42877c478bd9Sstevel@tonic-gate 42887c478bd9Sstevel@tonic-gate if (rw == UIO_WRITE) { 42897c478bd9Sstevel@tonic-gate auio.uio_fmode = FWRITE; 42907c478bd9Sstevel@tonic-gate auio.uio_extflg = UIO_COPY_DEFAULT; 42917c478bd9Sstevel@tonic-gate auio.uio_llimit = curproc->p_fsz_ctl; 42927c478bd9Sstevel@tonic-gate error = wrip(ip, &auio, ioflag, cr); 42937c478bd9Sstevel@tonic-gate } else { 42947c478bd9Sstevel@tonic-gate auio.uio_fmode = FREAD; 42957c478bd9Sstevel@tonic-gate auio.uio_extflg = UIO_COPY_CACHED; 42967c478bd9Sstevel@tonic-gate auio.uio_llimit = MAXOFFSET_T; 42977c478bd9Sstevel@tonic-gate error = rdip(ip, &auio, ioflag, cr); 42987c478bd9Sstevel@tonic-gate } 42997c478bd9Sstevel@tonic-gate 43007c478bd9Sstevel@tonic-gate if (aresid) { 43017c478bd9Sstevel@tonic-gate *aresid = auio.uio_resid; 43027c478bd9Sstevel@tonic-gate } else if (auio.uio_resid) { 43037c478bd9Sstevel@tonic-gate error = EIO; 43047c478bd9Sstevel@tonic-gate } 43057c478bd9Sstevel@tonic-gate return (error); 43067c478bd9Sstevel@tonic-gate } 43077c478bd9Sstevel@tonic-gate 43087c478bd9Sstevel@tonic-gate static int 43097c478bd9Sstevel@tonic-gate ufs_fid(vp, fidp) 43107c478bd9Sstevel@tonic-gate struct vnode *vp; 43117c478bd9Sstevel@tonic-gate struct fid *fidp; 43127c478bd9Sstevel@tonic-gate { 43137c478bd9Sstevel@tonic-gate struct ufid *ufid; 43147c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 43157c478bd9Sstevel@tonic-gate 43167c478bd9Sstevel@tonic-gate if (ip->i_ufsvfs == NULL) 43177c478bd9Sstevel@tonic-gate return (EIO); 43187c478bd9Sstevel@tonic-gate 43197c478bd9Sstevel@tonic-gate if (fidp->fid_len < (sizeof (struct ufid) - sizeof (ushort_t))) { 43207c478bd9Sstevel@tonic-gate fidp->fid_len = sizeof (struct ufid) - sizeof (ushort_t); 43217c478bd9Sstevel@tonic-gate return (ENOSPC); 43227c478bd9Sstevel@tonic-gate } 43237c478bd9Sstevel@tonic-gate 43247c478bd9Sstevel@tonic-gate ufid = (struct ufid *)fidp; 43257c478bd9Sstevel@tonic-gate bzero((char *)ufid, sizeof (struct ufid)); 43267c478bd9Sstevel@tonic-gate ufid->ufid_len = sizeof (struct ufid) - sizeof (ushort_t); 43277c478bd9Sstevel@tonic-gate ufid->ufid_ino = ip->i_number; 43287c478bd9Sstevel@tonic-gate ufid->ufid_gen = ip->i_gen; 43297c478bd9Sstevel@tonic-gate 43307c478bd9Sstevel@tonic-gate return (0); 43317c478bd9Sstevel@tonic-gate } 43327c478bd9Sstevel@tonic-gate 43337c478bd9Sstevel@tonic-gate /* ARGSUSED2 */ 43347c478bd9Sstevel@tonic-gate static int 43357c478bd9Sstevel@tonic-gate ufs_rwlock(struct vnode *vp, int write_lock, caller_context_t *ctp) 43367c478bd9Sstevel@tonic-gate { 43377c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 43387c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp; 43397c478bd9Sstevel@tonic-gate int forcedirectio; 43407c478bd9Sstevel@tonic-gate 43417c478bd9Sstevel@tonic-gate /* 43427c478bd9Sstevel@tonic-gate * Read case is easy. 43437c478bd9Sstevel@tonic-gate */ 43447c478bd9Sstevel@tonic-gate if (!write_lock) { 43457c478bd9Sstevel@tonic-gate rw_enter(&ip->i_rwlock, RW_READER); 43467c478bd9Sstevel@tonic-gate return (V_WRITELOCK_FALSE); 43477c478bd9Sstevel@tonic-gate } 43487c478bd9Sstevel@tonic-gate 43497c478bd9Sstevel@tonic-gate /* 43507c478bd9Sstevel@tonic-gate * Caller has requested a writer lock, but that inhibits any 43517c478bd9Sstevel@tonic-gate * concurrency in the VOPs that follow. Acquire the lock shared 43527c478bd9Sstevel@tonic-gate * and defer exclusive access until it is known to be needed in 43537c478bd9Sstevel@tonic-gate * other VOP handlers. Some cases can be determined here. 43547c478bd9Sstevel@tonic-gate */ 43557c478bd9Sstevel@tonic-gate 43567c478bd9Sstevel@tonic-gate /* 43577c478bd9Sstevel@tonic-gate * If directio is not set, there is no chance of concurrency, 43587c478bd9Sstevel@tonic-gate * so just acquire the lock exclusive. Beware of a forced 43597c478bd9Sstevel@tonic-gate * unmount before looking at the mount option. 43607c478bd9Sstevel@tonic-gate */ 43617c478bd9Sstevel@tonic-gate ufsvfsp = ip->i_ufsvfs; 43627c478bd9Sstevel@tonic-gate forcedirectio = ufsvfsp ? ufsvfsp->vfs_forcedirectio : 0; 43637c478bd9Sstevel@tonic-gate if (!(ip->i_flag & IDIRECTIO || forcedirectio) || 43647c478bd9Sstevel@tonic-gate !ufs_allow_shared_writes) { 43657c478bd9Sstevel@tonic-gate rw_enter(&ip->i_rwlock, RW_WRITER); 43667c478bd9Sstevel@tonic-gate return (V_WRITELOCK_TRUE); 43677c478bd9Sstevel@tonic-gate } 43687c478bd9Sstevel@tonic-gate 43697c478bd9Sstevel@tonic-gate /* 43707c478bd9Sstevel@tonic-gate * Mandatory locking forces acquiring i_rwlock exclusive. 43717c478bd9Sstevel@tonic-gate */ 43727c478bd9Sstevel@tonic-gate if (MANDLOCK(vp, ip->i_mode)) { 43737c478bd9Sstevel@tonic-gate rw_enter(&ip->i_rwlock, RW_WRITER); 43747c478bd9Sstevel@tonic-gate return (V_WRITELOCK_TRUE); 43757c478bd9Sstevel@tonic-gate } 43767c478bd9Sstevel@tonic-gate 43777c478bd9Sstevel@tonic-gate /* 43787c478bd9Sstevel@tonic-gate * Acquire the lock shared in case a concurrent write follows. 43797c478bd9Sstevel@tonic-gate * Mandatory locking could have become enabled before the lock 43807c478bd9Sstevel@tonic-gate * was acquired. Re-check and upgrade if needed. 43817c478bd9Sstevel@tonic-gate */ 43827c478bd9Sstevel@tonic-gate rw_enter(&ip->i_rwlock, RW_READER); 43837c478bd9Sstevel@tonic-gate if (MANDLOCK(vp, ip->i_mode)) { 43847c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 43857c478bd9Sstevel@tonic-gate rw_enter(&ip->i_rwlock, RW_WRITER); 43867c478bd9Sstevel@tonic-gate return (V_WRITELOCK_TRUE); 43877c478bd9Sstevel@tonic-gate } 43887c478bd9Sstevel@tonic-gate return (V_WRITELOCK_FALSE); 43897c478bd9Sstevel@tonic-gate } 43907c478bd9Sstevel@tonic-gate 43917c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 43927c478bd9Sstevel@tonic-gate static void 43937c478bd9Sstevel@tonic-gate ufs_rwunlock(struct vnode *vp, int write_lock, caller_context_t *ctp) 43947c478bd9Sstevel@tonic-gate { 43957c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 43967c478bd9Sstevel@tonic-gate 43977c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 43987c478bd9Sstevel@tonic-gate } 43997c478bd9Sstevel@tonic-gate 44007c478bd9Sstevel@tonic-gate /* ARGSUSED */ 44017c478bd9Sstevel@tonic-gate static int 44027c478bd9Sstevel@tonic-gate ufs_seek(struct vnode *vp, offset_t ooff, offset_t *noffp) 44037c478bd9Sstevel@tonic-gate { 44047c478bd9Sstevel@tonic-gate return ((*noffp < 0 || *noffp > MAXOFFSET_T) ? EINVAL : 0); 44057c478bd9Sstevel@tonic-gate } 44067c478bd9Sstevel@tonic-gate 44077c478bd9Sstevel@tonic-gate /* ARGSUSED */ 44087c478bd9Sstevel@tonic-gate static int 44097c478bd9Sstevel@tonic-gate ufs_frlock(struct vnode *vp, int cmd, struct flock64 *bfp, int flag, 44107c478bd9Sstevel@tonic-gate offset_t offset, struct flk_callback *flk_cbp, struct cred *cr) 44117c478bd9Sstevel@tonic-gate { 44127c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 44137c478bd9Sstevel@tonic-gate 44147c478bd9Sstevel@tonic-gate if (ip->i_ufsvfs == NULL) 44157c478bd9Sstevel@tonic-gate return (EIO); 44167c478bd9Sstevel@tonic-gate 44177c478bd9Sstevel@tonic-gate /* 44187c478bd9Sstevel@tonic-gate * If file is being mapped, disallow frlock. 44197c478bd9Sstevel@tonic-gate * XXX I am not holding tlock while checking i_mapcnt because the 44207c478bd9Sstevel@tonic-gate * current locking strategy drops all locks before calling fs_frlock. 44217c478bd9Sstevel@tonic-gate * So, mapcnt could change before we enter fs_frlock making is 44227c478bd9Sstevel@tonic-gate * meaningless to have held tlock in the first place. 44237c478bd9Sstevel@tonic-gate */ 44247c478bd9Sstevel@tonic-gate if (ip->i_mapcnt > 0 && MANDLOCK(vp, ip->i_mode)) 44257c478bd9Sstevel@tonic-gate return (EAGAIN); 44267c478bd9Sstevel@tonic-gate return (fs_frlock(vp, cmd, bfp, flag, offset, flk_cbp, cr)); 44277c478bd9Sstevel@tonic-gate } 44287c478bd9Sstevel@tonic-gate 44297c478bd9Sstevel@tonic-gate /* ARGSUSED */ 44307c478bd9Sstevel@tonic-gate static int 4431303bf60bSsdebnath ufs_space(struct vnode *vp, int cmd, struct flock64 *bfp, int flag, 4432303bf60bSsdebnath offset_t offset, cred_t *cr, caller_context_t *ct) 44337c478bd9Sstevel@tonic-gate { 4434303bf60bSsdebnath struct ufsvfs *ufsvfsp = VTOI(vp)->i_ufsvfs; 44357c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 44367c478bd9Sstevel@tonic-gate int error; 44377c478bd9Sstevel@tonic-gate 4438303bf60bSsdebnath if ((error = convoff(vp, bfp, 0, offset)) == 0) { 4439303bf60bSsdebnath if (cmd == F_FREESP) { 4440303bf60bSsdebnath error = ufs_lockfs_begin(ufsvfsp, &ulp, 4441303bf60bSsdebnath ULOCKFS_SPACE_MASK); 4442303bf60bSsdebnath if (error) 4443303bf60bSsdebnath return (error); 4444303bf60bSsdebnath error = ufs_freesp(vp, bfp, flag, cr); 4445303bf60bSsdebnath } else if (cmd == F_ALLOCSP) { 4446303bf60bSsdebnath error = ufs_lockfs_begin(ufsvfsp, &ulp, 4447303bf60bSsdebnath ULOCKFS_FALLOCATE_MASK); 4448303bf60bSsdebnath if (error) 4449303bf60bSsdebnath return (error); 4450303bf60bSsdebnath error = ufs_allocsp(vp, bfp, cr); 4451303bf60bSsdebnath } else 4452303bf60bSsdebnath return (EINVAL); /* Command not handled here */ 44537c478bd9Sstevel@tonic-gate 4454303bf60bSsdebnath if (ulp) 4455303bf60bSsdebnath ufs_lockfs_end(ulp); 44567c478bd9Sstevel@tonic-gate 4457303bf60bSsdebnath } 44587c478bd9Sstevel@tonic-gate return (error); 44597c478bd9Sstevel@tonic-gate } 44607c478bd9Sstevel@tonic-gate 44617c478bd9Sstevel@tonic-gate /* 44627c478bd9Sstevel@tonic-gate * Used to determine if read ahead should be done. Also used to 44637c478bd9Sstevel@tonic-gate * to determine when write back occurs. 44647c478bd9Sstevel@tonic-gate */ 44657c478bd9Sstevel@tonic-gate #define CLUSTSZ(ip) ((ip)->i_ufsvfs->vfs_ioclustsz) 44667c478bd9Sstevel@tonic-gate 44677c478bd9Sstevel@tonic-gate /* 44687c478bd9Sstevel@tonic-gate * A faster version of ufs_getpage. 44697c478bd9Sstevel@tonic-gate * 44707c478bd9Sstevel@tonic-gate * We optimize by inlining the pvn_getpages iterator, eliminating 44717c478bd9Sstevel@tonic-gate * calls to bmap_read if file doesn't have UFS holes, and avoiding 44727c478bd9Sstevel@tonic-gate * the overhead of page_exists(). 44737c478bd9Sstevel@tonic-gate * 44747c478bd9Sstevel@tonic-gate * When files has UFS_HOLES and ufs_getpage is called with S_READ, 44757c478bd9Sstevel@tonic-gate * we set *protp to PROT_READ to avoid calling bmap_read. This approach 44767c478bd9Sstevel@tonic-gate * victimizes performance when a file with UFS holes is faulted 44777c478bd9Sstevel@tonic-gate * first in the S_READ mode, and then in the S_WRITE mode. We will get 44787c478bd9Sstevel@tonic-gate * two MMU faults in this case. 44797c478bd9Sstevel@tonic-gate * 44807c478bd9Sstevel@tonic-gate * XXX - the inode fields which control the sequential mode are not 44817c478bd9Sstevel@tonic-gate * protected by any mutex. The read ahead will act wild if 44827c478bd9Sstevel@tonic-gate * multiple processes will access the file concurrently and 44837c478bd9Sstevel@tonic-gate * some of them in sequential mode. One particulary bad case 44847c478bd9Sstevel@tonic-gate * is if another thread will change the value of i_nextrio between 44857c478bd9Sstevel@tonic-gate * the time this thread tests the i_nextrio value and then reads it 44867c478bd9Sstevel@tonic-gate * again to use it as the offset for the read ahead. 44877c478bd9Sstevel@tonic-gate */ 44887c478bd9Sstevel@tonic-gate static int 44897c478bd9Sstevel@tonic-gate ufs_getpage(struct vnode *vp, offset_t off, size_t len, uint_t *protp, 44907c478bd9Sstevel@tonic-gate page_t *plarr[], size_t plsz, struct seg *seg, caddr_t addr, 44917c478bd9Sstevel@tonic-gate enum seg_rw rw, struct cred *cr) 44927c478bd9Sstevel@tonic-gate { 44937c478bd9Sstevel@tonic-gate u_offset_t uoff = (u_offset_t)off; /* type conversion */ 44947c478bd9Sstevel@tonic-gate u_offset_t pgoff; 44957c478bd9Sstevel@tonic-gate u_offset_t eoff; 44967c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 44977c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp = ip->i_ufsvfs; 44987c478bd9Sstevel@tonic-gate struct fs *fs; 44997c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 45007c478bd9Sstevel@tonic-gate page_t **pl; 45017c478bd9Sstevel@tonic-gate caddr_t pgaddr; 45027c478bd9Sstevel@tonic-gate krw_t rwtype; 45037c478bd9Sstevel@tonic-gate int err; 45047c478bd9Sstevel@tonic-gate int has_holes; 45057c478bd9Sstevel@tonic-gate int beyond_eof; 45067c478bd9Sstevel@tonic-gate int seqmode; 45077c478bd9Sstevel@tonic-gate int pgsize = PAGESIZE; 45087c478bd9Sstevel@tonic-gate int dolock; 45097c478bd9Sstevel@tonic-gate int do_qlock; 45107c478bd9Sstevel@tonic-gate int trans_size; 45117c478bd9Sstevel@tonic-gate 45127c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_GETPAGE_START, 45137c478bd9Sstevel@tonic-gate "ufs_getpage_start:vp %p", vp); 45147c478bd9Sstevel@tonic-gate 45157c478bd9Sstevel@tonic-gate ASSERT((uoff & PAGEOFFSET) == 0); 45167c478bd9Sstevel@tonic-gate 45177c478bd9Sstevel@tonic-gate if (protp) 45187c478bd9Sstevel@tonic-gate *protp = PROT_ALL; 45197c478bd9Sstevel@tonic-gate 45207c478bd9Sstevel@tonic-gate /* 45217c478bd9Sstevel@tonic-gate * Obey the lockfs protocol 45227c478bd9Sstevel@tonic-gate */ 45237c478bd9Sstevel@tonic-gate err = ufs_lockfs_begin_getpage(ufsvfsp, &ulp, seg, 45247c478bd9Sstevel@tonic-gate rw == S_READ || rw == S_EXEC, protp); 45257c478bd9Sstevel@tonic-gate if (err) 45267c478bd9Sstevel@tonic-gate goto out; 45277c478bd9Sstevel@tonic-gate 45287c478bd9Sstevel@tonic-gate fs = ufsvfsp->vfs_fs; 45297c478bd9Sstevel@tonic-gate 45307c478bd9Sstevel@tonic-gate if (ulp && (rw == S_CREATE || rw == S_WRITE) && 45317c478bd9Sstevel@tonic-gate !(vp->v_flag & VISSWAP)) { 45327c478bd9Sstevel@tonic-gate /* 45337c478bd9Sstevel@tonic-gate * Try to start a transaction, will return if blocking is 45347c478bd9Sstevel@tonic-gate * expected to occur and the address space is not the 45357c478bd9Sstevel@tonic-gate * kernel address space. 45367c478bd9Sstevel@tonic-gate */ 45377c478bd9Sstevel@tonic-gate trans_size = TOP_GETPAGE_SIZE(ip); 45387c478bd9Sstevel@tonic-gate if (seg->s_as != &kas) { 45397c478bd9Sstevel@tonic-gate TRANS_TRY_BEGIN_ASYNC(ufsvfsp, TOP_GETPAGE, 45407c478bd9Sstevel@tonic-gate trans_size, err) 45417c478bd9Sstevel@tonic-gate if (err == EWOULDBLOCK) { 45427c478bd9Sstevel@tonic-gate /* 45437c478bd9Sstevel@tonic-gate * Use EDEADLK here because the VM code 45447c478bd9Sstevel@tonic-gate * can normally never see this error. 45457c478bd9Sstevel@tonic-gate */ 45467c478bd9Sstevel@tonic-gate err = EDEADLK; 45477c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 45487c478bd9Sstevel@tonic-gate goto out; 45497c478bd9Sstevel@tonic-gate } 45507c478bd9Sstevel@tonic-gate } else { 45517c478bd9Sstevel@tonic-gate TRANS_BEGIN_ASYNC(ufsvfsp, TOP_GETPAGE, trans_size); 45527c478bd9Sstevel@tonic-gate } 45537c478bd9Sstevel@tonic-gate } 45547c478bd9Sstevel@tonic-gate 45557c478bd9Sstevel@tonic-gate if (vp->v_flag & VNOMAP) { 45567c478bd9Sstevel@tonic-gate err = ENOSYS; 45577c478bd9Sstevel@tonic-gate goto unlock; 45587c478bd9Sstevel@tonic-gate } 45597c478bd9Sstevel@tonic-gate 45607c478bd9Sstevel@tonic-gate seqmode = ip->i_nextr == uoff && rw != S_CREATE; 45617c478bd9Sstevel@tonic-gate 45627c478bd9Sstevel@tonic-gate rwtype = RW_READER; /* start as a reader */ 45637c478bd9Sstevel@tonic-gate dolock = (rw_owner(&ip->i_contents) != curthread); 45647c478bd9Sstevel@tonic-gate /* 45657c478bd9Sstevel@tonic-gate * If this thread owns the lock, i.e., this thread grabbed it 45667c478bd9Sstevel@tonic-gate * as writer somewhere above, then we don't need to grab the 45677c478bd9Sstevel@tonic-gate * lock as reader in this routine. 45687c478bd9Sstevel@tonic-gate */ 45697c478bd9Sstevel@tonic-gate do_qlock = (rw_owner(&ufsvfsp->vfs_dqrwlock) != curthread); 45707c478bd9Sstevel@tonic-gate 45717c478bd9Sstevel@tonic-gate retrylock: 45727c478bd9Sstevel@tonic-gate if (dolock) { 45737c478bd9Sstevel@tonic-gate /* 45747c478bd9Sstevel@tonic-gate * Grab the quota lock if we need to call 45757c478bd9Sstevel@tonic-gate * bmap_write() below (with i_contents as writer). 45767c478bd9Sstevel@tonic-gate */ 45777c478bd9Sstevel@tonic-gate if (do_qlock && rwtype == RW_WRITER) 45787c478bd9Sstevel@tonic-gate rw_enter(&ufsvfsp->vfs_dqrwlock, RW_READER); 45797c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, rwtype); 45807c478bd9Sstevel@tonic-gate } 45817c478bd9Sstevel@tonic-gate 45827c478bd9Sstevel@tonic-gate /* 45837c478bd9Sstevel@tonic-gate * We may be getting called as a side effect of a bmap using 45847c478bd9Sstevel@tonic-gate * fbread() when the blocks might be being allocated and the 45857c478bd9Sstevel@tonic-gate * size has not yet been up'ed. In this case we want to be 45867c478bd9Sstevel@tonic-gate * able to return zero pages if we get back UFS_HOLE from 45877c478bd9Sstevel@tonic-gate * calling bmap for a non write case here. We also might have 45887c478bd9Sstevel@tonic-gate * to read some frags from the disk into a page if we are 45897c478bd9Sstevel@tonic-gate * extending the number of frags for a given lbn in bmap(). 45907c478bd9Sstevel@tonic-gate * Large Files: The read of i_size here is atomic because 45917c478bd9Sstevel@tonic-gate * i_contents is held here. If dolock is zero, the lock 45927c478bd9Sstevel@tonic-gate * is held in bmap routines. 45937c478bd9Sstevel@tonic-gate */ 45947c478bd9Sstevel@tonic-gate beyond_eof = uoff + len > ip->i_size + PAGEOFFSET; 45957c478bd9Sstevel@tonic-gate if (beyond_eof && seg != segkmap) { 45967c478bd9Sstevel@tonic-gate if (dolock) { 45977c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 45987c478bd9Sstevel@tonic-gate if (do_qlock && rwtype == RW_WRITER) 45997c478bd9Sstevel@tonic-gate rw_exit(&ufsvfsp->vfs_dqrwlock); 46007c478bd9Sstevel@tonic-gate } 46017c478bd9Sstevel@tonic-gate err = EFAULT; 46027c478bd9Sstevel@tonic-gate goto unlock; 46037c478bd9Sstevel@tonic-gate } 46047c478bd9Sstevel@tonic-gate 46057c478bd9Sstevel@tonic-gate /* 46067c478bd9Sstevel@tonic-gate * Must hold i_contents lock throughout the call to pvn_getpages 46077c478bd9Sstevel@tonic-gate * since locked pages are returned from each call to ufs_getapage. 46087c478bd9Sstevel@tonic-gate * Must *not* return locked pages and then try for contents lock 46097c478bd9Sstevel@tonic-gate * due to lock ordering requirements (inode > page) 46107c478bd9Sstevel@tonic-gate */ 46117c478bd9Sstevel@tonic-gate 46127c478bd9Sstevel@tonic-gate has_holes = bmap_has_holes(ip); 46137c478bd9Sstevel@tonic-gate 46147c478bd9Sstevel@tonic-gate if ((rw == S_WRITE || rw == S_CREATE) && has_holes && !beyond_eof) { 46157c478bd9Sstevel@tonic-gate int blk_size; 46167c478bd9Sstevel@tonic-gate u_offset_t offset; 46177c478bd9Sstevel@tonic-gate 46187c478bd9Sstevel@tonic-gate /* 46197c478bd9Sstevel@tonic-gate * We must acquire the RW_WRITER lock in order to 46207c478bd9Sstevel@tonic-gate * call bmap_write(). 46217c478bd9Sstevel@tonic-gate */ 46227c478bd9Sstevel@tonic-gate if (dolock && rwtype == RW_READER) { 46237c478bd9Sstevel@tonic-gate rwtype = RW_WRITER; 46247c478bd9Sstevel@tonic-gate 46257c478bd9Sstevel@tonic-gate /* 46267c478bd9Sstevel@tonic-gate * Grab the quota lock before 46277c478bd9Sstevel@tonic-gate * upgrading i_contents, but if we can't grab it 46287c478bd9Sstevel@tonic-gate * don't wait here due to lock order: 46297c478bd9Sstevel@tonic-gate * vfs_dqrwlock > i_contents. 46307c478bd9Sstevel@tonic-gate */ 46317c478bd9Sstevel@tonic-gate if (do_qlock && rw_tryenter(&ufsvfsp->vfs_dqrwlock, 46327c478bd9Sstevel@tonic-gate RW_READER) == 0) { 46337c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 46347c478bd9Sstevel@tonic-gate goto retrylock; 46357c478bd9Sstevel@tonic-gate } 46367c478bd9Sstevel@tonic-gate if (!rw_tryupgrade(&ip->i_contents)) { 46377c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 46387c478bd9Sstevel@tonic-gate if (do_qlock) 46397c478bd9Sstevel@tonic-gate rw_exit(&ufsvfsp->vfs_dqrwlock); 46407c478bd9Sstevel@tonic-gate goto retrylock; 46417c478bd9Sstevel@tonic-gate } 46427c478bd9Sstevel@tonic-gate } 46437c478bd9Sstevel@tonic-gate 46447c478bd9Sstevel@tonic-gate /* 46457c478bd9Sstevel@tonic-gate * May be allocating disk blocks for holes here as 46467c478bd9Sstevel@tonic-gate * a result of mmap faults. write(2) does the bmap_write 46477c478bd9Sstevel@tonic-gate * in rdip/wrip, not here. We are not dealing with frags 46487c478bd9Sstevel@tonic-gate * in this case. 46497c478bd9Sstevel@tonic-gate */ 46507c478bd9Sstevel@tonic-gate /* 46517c478bd9Sstevel@tonic-gate * Large Files: We cast fs_bmask field to offset_t 46527c478bd9Sstevel@tonic-gate * just as we do for MAXBMASK because uoff is a 64-bit 46537c478bd9Sstevel@tonic-gate * data type. fs_bmask will still be a 32-bit type 46547c478bd9Sstevel@tonic-gate * as we cannot change any ondisk data structures. 46557c478bd9Sstevel@tonic-gate */ 46567c478bd9Sstevel@tonic-gate 46577c478bd9Sstevel@tonic-gate offset = uoff & (offset_t)fs->fs_bmask; 46587c478bd9Sstevel@tonic-gate while (offset < uoff + len) { 46597c478bd9Sstevel@tonic-gate blk_size = (int)blksize(fs, ip, lblkno(fs, offset)); 4660303bf60bSsdebnath err = bmap_write(ip, offset, blk_size, 4661303bf60bSsdebnath BI_NORMAL, NULL, cr); 46627c478bd9Sstevel@tonic-gate if (ip->i_flag & (ICHG|IUPD)) 46637c478bd9Sstevel@tonic-gate ip->i_seq++; 46647c478bd9Sstevel@tonic-gate if (err) 46657c478bd9Sstevel@tonic-gate goto update_inode; 46667c478bd9Sstevel@tonic-gate offset += blk_size; /* XXX - make this contig */ 46677c478bd9Sstevel@tonic-gate } 46687c478bd9Sstevel@tonic-gate } 46697c478bd9Sstevel@tonic-gate 46707c478bd9Sstevel@tonic-gate /* 46717c478bd9Sstevel@tonic-gate * Can be a reader from now on. 46727c478bd9Sstevel@tonic-gate */ 46737c478bd9Sstevel@tonic-gate if (dolock && rwtype == RW_WRITER) { 46747c478bd9Sstevel@tonic-gate rw_downgrade(&ip->i_contents); 46757c478bd9Sstevel@tonic-gate /* 46767c478bd9Sstevel@tonic-gate * We can release vfs_dqrwlock early so do it, but make 46777c478bd9Sstevel@tonic-gate * sure we don't try to release it again at the bottom. 46787c478bd9Sstevel@tonic-gate */ 46797c478bd9Sstevel@tonic-gate if (do_qlock) { 46807c478bd9Sstevel@tonic-gate rw_exit(&ufsvfsp->vfs_dqrwlock); 46817c478bd9Sstevel@tonic-gate do_qlock = 0; 46827c478bd9Sstevel@tonic-gate } 46837c478bd9Sstevel@tonic-gate } 46847c478bd9Sstevel@tonic-gate 46857c478bd9Sstevel@tonic-gate /* 46867c478bd9Sstevel@tonic-gate * We remove PROT_WRITE in cases when the file has UFS holes 46877c478bd9Sstevel@tonic-gate * because we don't want to call bmap_read() to check each 46887c478bd9Sstevel@tonic-gate * page if it is backed with a disk block. 46897c478bd9Sstevel@tonic-gate */ 46907c478bd9Sstevel@tonic-gate if (protp && has_holes && rw != S_WRITE && rw != S_CREATE) 46917c478bd9Sstevel@tonic-gate *protp &= ~PROT_WRITE; 46927c478bd9Sstevel@tonic-gate 46937c478bd9Sstevel@tonic-gate err = 0; 46947c478bd9Sstevel@tonic-gate 46957c478bd9Sstevel@tonic-gate /* 46967c478bd9Sstevel@tonic-gate * The loop looks up pages in the range [off, off + len). 46977c478bd9Sstevel@tonic-gate * For each page, we first check if we should initiate an asynchronous 46987c478bd9Sstevel@tonic-gate * read ahead before we call page_lookup (we may sleep in page_lookup 46997c478bd9Sstevel@tonic-gate * for a previously initiated disk read). 47007c478bd9Sstevel@tonic-gate */ 47017c478bd9Sstevel@tonic-gate eoff = (uoff + len); 47027c478bd9Sstevel@tonic-gate for (pgoff = uoff, pgaddr = addr, pl = plarr; 47037c478bd9Sstevel@tonic-gate pgoff < eoff; /* empty */) { 47047c478bd9Sstevel@tonic-gate page_t *pp; 47057c478bd9Sstevel@tonic-gate u_offset_t nextrio; 47067c478bd9Sstevel@tonic-gate se_t se; 47077c478bd9Sstevel@tonic-gate int retval; 47087c478bd9Sstevel@tonic-gate 47097c478bd9Sstevel@tonic-gate se = ((rw == S_CREATE || rw == S_OTHER) ? SE_EXCL : SE_SHARED); 47107c478bd9Sstevel@tonic-gate 47117c478bd9Sstevel@tonic-gate /* Handle async getpage (faultahead) */ 47127c478bd9Sstevel@tonic-gate if (plarr == NULL) { 47137c478bd9Sstevel@tonic-gate ip->i_nextrio = pgoff; 47147c478bd9Sstevel@tonic-gate (void) ufs_getpage_ra(vp, pgoff, seg, pgaddr); 47157c478bd9Sstevel@tonic-gate pgoff += pgsize; 47167c478bd9Sstevel@tonic-gate pgaddr += pgsize; 47177c478bd9Sstevel@tonic-gate continue; 47187c478bd9Sstevel@tonic-gate } 47197c478bd9Sstevel@tonic-gate /* 47207c478bd9Sstevel@tonic-gate * Check if we should initiate read ahead of next cluster. 47217c478bd9Sstevel@tonic-gate * We call page_exists only when we need to confirm that 47227c478bd9Sstevel@tonic-gate * we have the current page before we initiate the read ahead. 47237c478bd9Sstevel@tonic-gate */ 47247c478bd9Sstevel@tonic-gate nextrio = ip->i_nextrio; 47257c478bd9Sstevel@tonic-gate if (seqmode && 47267c478bd9Sstevel@tonic-gate pgoff + CLUSTSZ(ip) >= nextrio && pgoff <= nextrio && 47277c478bd9Sstevel@tonic-gate nextrio < ip->i_size && page_exists(vp, pgoff)) { 47287c478bd9Sstevel@tonic-gate retval = ufs_getpage_ra(vp, pgoff, seg, pgaddr); 47297c478bd9Sstevel@tonic-gate /* 47307c478bd9Sstevel@tonic-gate * We always read ahead the next cluster of data 47317c478bd9Sstevel@tonic-gate * starting from i_nextrio. If the page (vp,nextrio) 47327c478bd9Sstevel@tonic-gate * is actually in core at this point, the routine 47337c478bd9Sstevel@tonic-gate * ufs_getpage_ra() will stop pre-fetching data 47347c478bd9Sstevel@tonic-gate * until we read that page in a synchronized manner 47357c478bd9Sstevel@tonic-gate * through ufs_getpage_miss(). So, we should increase 47367c478bd9Sstevel@tonic-gate * i_nextrio if the page (vp, nextrio) exists. 47377c478bd9Sstevel@tonic-gate */ 47387c478bd9Sstevel@tonic-gate if ((retval == 0) && page_exists(vp, nextrio)) { 47397c478bd9Sstevel@tonic-gate ip->i_nextrio = nextrio + pgsize; 47407c478bd9Sstevel@tonic-gate } 47417c478bd9Sstevel@tonic-gate } 47427c478bd9Sstevel@tonic-gate 47437c478bd9Sstevel@tonic-gate if ((pp = page_lookup(vp, pgoff, se)) != NULL) { 47447c478bd9Sstevel@tonic-gate /* 47457c478bd9Sstevel@tonic-gate * We found the page in the page cache. 47467c478bd9Sstevel@tonic-gate */ 47477c478bd9Sstevel@tonic-gate *pl++ = pp; 47487c478bd9Sstevel@tonic-gate pgoff += pgsize; 47497c478bd9Sstevel@tonic-gate pgaddr += pgsize; 47507c478bd9Sstevel@tonic-gate len -= pgsize; 47517c478bd9Sstevel@tonic-gate plsz -= pgsize; 47527c478bd9Sstevel@tonic-gate } else { 47537c478bd9Sstevel@tonic-gate /* 47547c478bd9Sstevel@tonic-gate * We have to create the page, or read it from disk. 47557c478bd9Sstevel@tonic-gate */ 47567c478bd9Sstevel@tonic-gate if (err = ufs_getpage_miss(vp, pgoff, len, seg, pgaddr, 47577c478bd9Sstevel@tonic-gate pl, plsz, rw, seqmode)) 47587c478bd9Sstevel@tonic-gate goto error; 47597c478bd9Sstevel@tonic-gate 47607c478bd9Sstevel@tonic-gate while (*pl != NULL) { 47617c478bd9Sstevel@tonic-gate pl++; 47627c478bd9Sstevel@tonic-gate pgoff += pgsize; 47637c478bd9Sstevel@tonic-gate pgaddr += pgsize; 47647c478bd9Sstevel@tonic-gate len -= pgsize; 47657c478bd9Sstevel@tonic-gate plsz -= pgsize; 47667c478bd9Sstevel@tonic-gate } 47677c478bd9Sstevel@tonic-gate } 47687c478bd9Sstevel@tonic-gate } 47697c478bd9Sstevel@tonic-gate 47707c478bd9Sstevel@tonic-gate /* 47717c478bd9Sstevel@tonic-gate * Return pages up to plsz if they are in the page cache. 47727c478bd9Sstevel@tonic-gate * We cannot return pages if there is a chance that they are 47737c478bd9Sstevel@tonic-gate * backed with a UFS hole and rw is S_WRITE or S_CREATE. 47747c478bd9Sstevel@tonic-gate */ 47757c478bd9Sstevel@tonic-gate if (plarr && !(has_holes && (rw == S_WRITE || rw == S_CREATE))) { 47767c478bd9Sstevel@tonic-gate 47777c478bd9Sstevel@tonic-gate ASSERT((protp == NULL) || 47787c478bd9Sstevel@tonic-gate !(has_holes && (*protp & PROT_WRITE))); 47797c478bd9Sstevel@tonic-gate 47807c478bd9Sstevel@tonic-gate eoff = pgoff + plsz; 47817c478bd9Sstevel@tonic-gate while (pgoff < eoff) { 47827c478bd9Sstevel@tonic-gate page_t *pp; 47837c478bd9Sstevel@tonic-gate 47847c478bd9Sstevel@tonic-gate if ((pp = page_lookup_nowait(vp, pgoff, 47857c478bd9Sstevel@tonic-gate SE_SHARED)) == NULL) 47867c478bd9Sstevel@tonic-gate break; 47877c478bd9Sstevel@tonic-gate 47887c478bd9Sstevel@tonic-gate *pl++ = pp; 47897c478bd9Sstevel@tonic-gate pgoff += pgsize; 47907c478bd9Sstevel@tonic-gate plsz -= pgsize; 47917c478bd9Sstevel@tonic-gate } 47927c478bd9Sstevel@tonic-gate } 47937c478bd9Sstevel@tonic-gate 47947c478bd9Sstevel@tonic-gate if (plarr) 47957c478bd9Sstevel@tonic-gate *pl = NULL; /* Terminate page list */ 47967c478bd9Sstevel@tonic-gate ip->i_nextr = pgoff; 47977c478bd9Sstevel@tonic-gate 47987c478bd9Sstevel@tonic-gate error: 47997c478bd9Sstevel@tonic-gate if (err && plarr) { 48007c478bd9Sstevel@tonic-gate /* 48017c478bd9Sstevel@tonic-gate * Release any pages we have locked. 48027c478bd9Sstevel@tonic-gate */ 48037c478bd9Sstevel@tonic-gate while (pl > &plarr[0]) 48047c478bd9Sstevel@tonic-gate page_unlock(*--pl); 48057c478bd9Sstevel@tonic-gate 48067c478bd9Sstevel@tonic-gate plarr[0] = NULL; 48077c478bd9Sstevel@tonic-gate } 48087c478bd9Sstevel@tonic-gate 48097c478bd9Sstevel@tonic-gate update_inode: 48107c478bd9Sstevel@tonic-gate /* 48117c478bd9Sstevel@tonic-gate * If the inode is not already marked for IACC (in rdip() for read) 48127c478bd9Sstevel@tonic-gate * and the inode is not marked for no access time update (in wrip() 48137c478bd9Sstevel@tonic-gate * for write) then update the inode access time and mod time now. 48147c478bd9Sstevel@tonic-gate */ 48157c478bd9Sstevel@tonic-gate if ((ip->i_flag & (IACC | INOACC)) == 0) { 48167c478bd9Sstevel@tonic-gate if ((rw != S_OTHER) && (ip->i_mode & IFMT) != IFDIR) { 48177c478bd9Sstevel@tonic-gate if (!ULOCKFS_IS_NOIACC(ITOUL(ip)) && 48187c478bd9Sstevel@tonic-gate (fs->fs_ronly == 0) && 48197c478bd9Sstevel@tonic-gate (!ufsvfsp->vfs_noatime)) { 48207c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 48217c478bd9Sstevel@tonic-gate ip->i_flag |= IACC; 48227c478bd9Sstevel@tonic-gate ITIMES_NOLOCK(ip); 48237c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 48247c478bd9Sstevel@tonic-gate } 48257c478bd9Sstevel@tonic-gate } 48267c478bd9Sstevel@tonic-gate } 48277c478bd9Sstevel@tonic-gate 48287c478bd9Sstevel@tonic-gate if (dolock) { 48297c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 48307c478bd9Sstevel@tonic-gate if (do_qlock && rwtype == RW_WRITER) 48317c478bd9Sstevel@tonic-gate rw_exit(&ufsvfsp->vfs_dqrwlock); 48327c478bd9Sstevel@tonic-gate } 48337c478bd9Sstevel@tonic-gate 48347c478bd9Sstevel@tonic-gate unlock: 48357c478bd9Sstevel@tonic-gate if (ulp) { 48367c478bd9Sstevel@tonic-gate if ((rw == S_CREATE || rw == S_WRITE) && 48377c478bd9Sstevel@tonic-gate !(vp->v_flag & VISSWAP)) { 48387c478bd9Sstevel@tonic-gate TRANS_END_ASYNC(ufsvfsp, TOP_GETPAGE, trans_size); 48397c478bd9Sstevel@tonic-gate } 48407c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 48417c478bd9Sstevel@tonic-gate } 48427c478bd9Sstevel@tonic-gate out: 48437c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_GETPAGE_END, 48447c478bd9Sstevel@tonic-gate "ufs_getpage_end:vp %p error %d", vp, err); 48457c478bd9Sstevel@tonic-gate return (err); 48467c478bd9Sstevel@tonic-gate } 48477c478bd9Sstevel@tonic-gate 48487c478bd9Sstevel@tonic-gate /* 48497c478bd9Sstevel@tonic-gate * ufs_getpage_miss is called when ufs_getpage missed the page in the page 48507c478bd9Sstevel@tonic-gate * cache. The page is either read from the disk, or it's created. 48517c478bd9Sstevel@tonic-gate * A page is created (without disk read) if rw == S_CREATE, or if 48527c478bd9Sstevel@tonic-gate * the page is not backed with a real disk block (UFS hole). 48537c478bd9Sstevel@tonic-gate */ 48547c478bd9Sstevel@tonic-gate /* ARGSUSED */ 48557c478bd9Sstevel@tonic-gate static int 48567c478bd9Sstevel@tonic-gate ufs_getpage_miss(struct vnode *vp, u_offset_t off, size_t len, struct seg *seg, 48577c478bd9Sstevel@tonic-gate caddr_t addr, page_t *pl[], size_t plsz, enum seg_rw rw, int seq) 48587c478bd9Sstevel@tonic-gate { 48597c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 48607c478bd9Sstevel@tonic-gate page_t *pp; 48617c478bd9Sstevel@tonic-gate daddr_t bn; 48627c478bd9Sstevel@tonic-gate size_t io_len; 4863303bf60bSsdebnath int crpage = 0; 48647c478bd9Sstevel@tonic-gate int err; 48657c478bd9Sstevel@tonic-gate int contig; 48667c478bd9Sstevel@tonic-gate int bsize = ip->i_fs->fs_bsize; 48677c478bd9Sstevel@tonic-gate 48687c478bd9Sstevel@tonic-gate /* 48697c478bd9Sstevel@tonic-gate * Figure out whether the page can be created, or must be 48707c478bd9Sstevel@tonic-gate * must be read from the disk. 48717c478bd9Sstevel@tonic-gate */ 48727c478bd9Sstevel@tonic-gate if (rw == S_CREATE) 48737c478bd9Sstevel@tonic-gate crpage = 1; 48747c478bd9Sstevel@tonic-gate else { 48757c478bd9Sstevel@tonic-gate contig = 0; 48767c478bd9Sstevel@tonic-gate if (err = bmap_read(ip, off, &bn, &contig)) 48777c478bd9Sstevel@tonic-gate return (err); 4878303bf60bSsdebnath 48797c478bd9Sstevel@tonic-gate crpage = (bn == UFS_HOLE); 4880303bf60bSsdebnath 4881303bf60bSsdebnath /* 4882303bf60bSsdebnath * If its also a fallocated block that hasn't been written to 4883303bf60bSsdebnath * yet, we will treat it just like a UFS_HOLE and create 4884303bf60bSsdebnath * a zero page for it 4885303bf60bSsdebnath */ 4886303bf60bSsdebnath if (ISFALLOCBLK(ip, bn)) 4887303bf60bSsdebnath crpage = 1; 48887c478bd9Sstevel@tonic-gate } 48897c478bd9Sstevel@tonic-gate 48907c478bd9Sstevel@tonic-gate if (crpage) { 48917c478bd9Sstevel@tonic-gate if ((pp = page_create_va(vp, off, PAGESIZE, PG_WAIT, seg, 48927c478bd9Sstevel@tonic-gate addr)) == NULL) { 48937c478bd9Sstevel@tonic-gate return (ufs_fault(vp, 48947c478bd9Sstevel@tonic-gate "ufs_getpage_miss: page_create == NULL")); 48957c478bd9Sstevel@tonic-gate } 48967c478bd9Sstevel@tonic-gate 48977c478bd9Sstevel@tonic-gate if (rw != S_CREATE) 48987c478bd9Sstevel@tonic-gate pagezero(pp, 0, PAGESIZE); 4899303bf60bSsdebnath 49007c478bd9Sstevel@tonic-gate io_len = PAGESIZE; 49017c478bd9Sstevel@tonic-gate } else { 49027c478bd9Sstevel@tonic-gate u_offset_t io_off; 49037c478bd9Sstevel@tonic-gate uint_t xlen; 49047c478bd9Sstevel@tonic-gate struct buf *bp; 49057c478bd9Sstevel@tonic-gate ufsvfs_t *ufsvfsp = ip->i_ufsvfs; 49067c478bd9Sstevel@tonic-gate 49077c478bd9Sstevel@tonic-gate /* 49087c478bd9Sstevel@tonic-gate * If access is not in sequential order, we read from disk 49097c478bd9Sstevel@tonic-gate * in bsize units. 49107c478bd9Sstevel@tonic-gate * 49117c478bd9Sstevel@tonic-gate * We limit the size of the transfer to bsize if we are reading 49127c478bd9Sstevel@tonic-gate * from the beginning of the file. Note in this situation we 49137c478bd9Sstevel@tonic-gate * will hedge our bets and initiate an async read ahead of 49147c478bd9Sstevel@tonic-gate * the second block. 49157c478bd9Sstevel@tonic-gate */ 49167c478bd9Sstevel@tonic-gate if (!seq || off == 0) 49177c478bd9Sstevel@tonic-gate contig = MIN(contig, bsize); 49187c478bd9Sstevel@tonic-gate 49197c478bd9Sstevel@tonic-gate pp = pvn_read_kluster(vp, off, seg, addr, &io_off, 49207c478bd9Sstevel@tonic-gate &io_len, off, contig, 0); 49217c478bd9Sstevel@tonic-gate 49227c478bd9Sstevel@tonic-gate /* 49237c478bd9Sstevel@tonic-gate * Some other thread has entered the page. 49247c478bd9Sstevel@tonic-gate * ufs_getpage will retry page_lookup. 49257c478bd9Sstevel@tonic-gate */ 49267c478bd9Sstevel@tonic-gate if (pp == NULL) { 49277c478bd9Sstevel@tonic-gate pl[0] = NULL; 49287c478bd9Sstevel@tonic-gate return (0); 49297c478bd9Sstevel@tonic-gate } 49307c478bd9Sstevel@tonic-gate 49317c478bd9Sstevel@tonic-gate /* 49327c478bd9Sstevel@tonic-gate * Zero part of the page which we are not 49337c478bd9Sstevel@tonic-gate * going to read from the disk. 49347c478bd9Sstevel@tonic-gate */ 49357c478bd9Sstevel@tonic-gate xlen = io_len & PAGEOFFSET; 49367c478bd9Sstevel@tonic-gate if (xlen != 0) 49377c478bd9Sstevel@tonic-gate pagezero(pp->p_prev, xlen, PAGESIZE - xlen); 49387c478bd9Sstevel@tonic-gate 49397c478bd9Sstevel@tonic-gate bp = pageio_setup(pp, io_len, ip->i_devvp, B_READ); 49407c478bd9Sstevel@tonic-gate bp->b_edev = ip->i_dev; 49417c478bd9Sstevel@tonic-gate bp->b_dev = cmpdev(ip->i_dev); 49427c478bd9Sstevel@tonic-gate bp->b_blkno = bn; 49437c478bd9Sstevel@tonic-gate bp->b_un.b_addr = (caddr_t)0; 49447c478bd9Sstevel@tonic-gate bp->b_file = ip->i_vnode; 49457c478bd9Sstevel@tonic-gate bp->b_offset = off; 49467c478bd9Sstevel@tonic-gate 49477c478bd9Sstevel@tonic-gate if (ufsvfsp->vfs_log) { 49487c478bd9Sstevel@tonic-gate lufs_read_strategy(ufsvfsp->vfs_log, bp); 49497c478bd9Sstevel@tonic-gate } else if (ufsvfsp->vfs_snapshot) { 49507c478bd9Sstevel@tonic-gate fssnap_strategy(&ufsvfsp->vfs_snapshot, bp); 49517c478bd9Sstevel@tonic-gate } else { 49527c478bd9Sstevel@tonic-gate ufsvfsp->vfs_iotstamp = lbolt; 49537c478bd9Sstevel@tonic-gate ub.ub_getpages.value.ul++; 49547c478bd9Sstevel@tonic-gate (void) bdev_strategy(bp); 49557c478bd9Sstevel@tonic-gate lwp_stat_update(LWP_STAT_INBLK, 1); 49567c478bd9Sstevel@tonic-gate } 49577c478bd9Sstevel@tonic-gate 49587c478bd9Sstevel@tonic-gate ip->i_nextrio = off + ((io_len + PAGESIZE - 1) & PAGEMASK); 49597c478bd9Sstevel@tonic-gate 49607c478bd9Sstevel@tonic-gate /* 49617c478bd9Sstevel@tonic-gate * If the file access is sequential, initiate read ahead 49627c478bd9Sstevel@tonic-gate * of the next cluster. 49637c478bd9Sstevel@tonic-gate */ 49647c478bd9Sstevel@tonic-gate if (seq && ip->i_nextrio < ip->i_size) 49657c478bd9Sstevel@tonic-gate (void) ufs_getpage_ra(vp, off, seg, addr); 49667c478bd9Sstevel@tonic-gate err = biowait(bp); 49677c478bd9Sstevel@tonic-gate pageio_done(bp); 49687c478bd9Sstevel@tonic-gate 49697c478bd9Sstevel@tonic-gate if (err) { 49707c478bd9Sstevel@tonic-gate pvn_read_done(pp, B_ERROR); 49717c478bd9Sstevel@tonic-gate return (err); 49727c478bd9Sstevel@tonic-gate } 49737c478bd9Sstevel@tonic-gate } 49747c478bd9Sstevel@tonic-gate 49757c478bd9Sstevel@tonic-gate pvn_plist_init(pp, pl, plsz, off, io_len, rw); 49767c478bd9Sstevel@tonic-gate return (0); 49777c478bd9Sstevel@tonic-gate } 49787c478bd9Sstevel@tonic-gate 49797c478bd9Sstevel@tonic-gate /* 49807c478bd9Sstevel@tonic-gate * Read ahead a cluster from the disk. Returns the length in bytes. 49817c478bd9Sstevel@tonic-gate */ 49827c478bd9Sstevel@tonic-gate static int 49837c478bd9Sstevel@tonic-gate ufs_getpage_ra(struct vnode *vp, u_offset_t off, struct seg *seg, caddr_t addr) 49847c478bd9Sstevel@tonic-gate { 49857c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 49867c478bd9Sstevel@tonic-gate page_t *pp; 49877c478bd9Sstevel@tonic-gate u_offset_t io_off = ip->i_nextrio; 49887c478bd9Sstevel@tonic-gate ufsvfs_t *ufsvfsp; 49897c478bd9Sstevel@tonic-gate caddr_t addr2 = addr + (io_off - off); 49907c478bd9Sstevel@tonic-gate struct buf *bp; 49917c478bd9Sstevel@tonic-gate daddr_t bn; 49927c478bd9Sstevel@tonic-gate size_t io_len; 4993303bf60bSsdebnath int err; 49947c478bd9Sstevel@tonic-gate int contig; 49957c478bd9Sstevel@tonic-gate int xlen; 49967c478bd9Sstevel@tonic-gate int bsize = ip->i_fs->fs_bsize; 49977c478bd9Sstevel@tonic-gate 49987c478bd9Sstevel@tonic-gate /* 49997c478bd9Sstevel@tonic-gate * If the directio advisory is in effect on this file, 50007c478bd9Sstevel@tonic-gate * then do not do buffered read ahead. Read ahead makes 50017c478bd9Sstevel@tonic-gate * it more difficult on threads using directio as they 50027c478bd9Sstevel@tonic-gate * will be forced to flush the pages from this vnode. 50037c478bd9Sstevel@tonic-gate */ 50047c478bd9Sstevel@tonic-gate if ((ufsvfsp = ip->i_ufsvfs) == NULL) 50057c478bd9Sstevel@tonic-gate return (0); 50067c478bd9Sstevel@tonic-gate if (ip->i_flag & IDIRECTIO || ufsvfsp->vfs_forcedirectio) 50077c478bd9Sstevel@tonic-gate return (0); 50087c478bd9Sstevel@tonic-gate 50097c478bd9Sstevel@tonic-gate /* 50107c478bd9Sstevel@tonic-gate * Is this test needed? 50117c478bd9Sstevel@tonic-gate */ 50127c478bd9Sstevel@tonic-gate if (addr2 >= seg->s_base + seg->s_size) 50137c478bd9Sstevel@tonic-gate return (0); 50147c478bd9Sstevel@tonic-gate 50157c478bd9Sstevel@tonic-gate contig = 0; 5016303bf60bSsdebnath err = bmap_read(ip, io_off, &bn, &contig); 5017303bf60bSsdebnath /* 5018303bf60bSsdebnath * If its a UFS_HOLE or a fallocated block, do not perform 5019303bf60bSsdebnath * any read ahead's since there probably is nothing to read ahead 5020303bf60bSsdebnath */ 5021303bf60bSsdebnath if (err || bn == UFS_HOLE || ISFALLOCBLK(ip, bn)) 50227c478bd9Sstevel@tonic-gate return (0); 50237c478bd9Sstevel@tonic-gate 50247c478bd9Sstevel@tonic-gate /* 50257c478bd9Sstevel@tonic-gate * Limit the transfer size to bsize if this is the 2nd block. 50267c478bd9Sstevel@tonic-gate */ 50277c478bd9Sstevel@tonic-gate if (io_off == (u_offset_t)bsize) 50287c478bd9Sstevel@tonic-gate contig = MIN(contig, bsize); 50297c478bd9Sstevel@tonic-gate 50307c478bd9Sstevel@tonic-gate if ((pp = pvn_read_kluster(vp, io_off, seg, addr2, &io_off, 50317c478bd9Sstevel@tonic-gate &io_len, io_off, contig, 1)) == NULL) 50327c478bd9Sstevel@tonic-gate return (0); 50337c478bd9Sstevel@tonic-gate 50347c478bd9Sstevel@tonic-gate /* 50357c478bd9Sstevel@tonic-gate * Zero part of page which we are not going to read from disk 50367c478bd9Sstevel@tonic-gate */ 50377c478bd9Sstevel@tonic-gate if ((xlen = (io_len & PAGEOFFSET)) > 0) 50387c478bd9Sstevel@tonic-gate pagezero(pp->p_prev, xlen, PAGESIZE - xlen); 50397c478bd9Sstevel@tonic-gate 50407c478bd9Sstevel@tonic-gate ip->i_nextrio = (io_off + io_len + PAGESIZE - 1) & PAGEMASK; 50417c478bd9Sstevel@tonic-gate 50427c478bd9Sstevel@tonic-gate bp = pageio_setup(pp, io_len, ip->i_devvp, B_READ | B_ASYNC); 50437c478bd9Sstevel@tonic-gate bp->b_edev = ip->i_dev; 50447c478bd9Sstevel@tonic-gate bp->b_dev = cmpdev(ip->i_dev); 50457c478bd9Sstevel@tonic-gate bp->b_blkno = bn; 50467c478bd9Sstevel@tonic-gate bp->b_un.b_addr = (caddr_t)0; 50477c478bd9Sstevel@tonic-gate bp->b_file = ip->i_vnode; 50487c478bd9Sstevel@tonic-gate bp->b_offset = off; 50497c478bd9Sstevel@tonic-gate 50507c478bd9Sstevel@tonic-gate if (ufsvfsp->vfs_log) { 50517c478bd9Sstevel@tonic-gate lufs_read_strategy(ufsvfsp->vfs_log, bp); 50527c478bd9Sstevel@tonic-gate } else if (ufsvfsp->vfs_snapshot) { 50537c478bd9Sstevel@tonic-gate fssnap_strategy(&ufsvfsp->vfs_snapshot, bp); 50547c478bd9Sstevel@tonic-gate } else { 50557c478bd9Sstevel@tonic-gate ufsvfsp->vfs_iotstamp = lbolt; 50567c478bd9Sstevel@tonic-gate ub.ub_getras.value.ul++; 50577c478bd9Sstevel@tonic-gate (void) bdev_strategy(bp); 50587c478bd9Sstevel@tonic-gate lwp_stat_update(LWP_STAT_INBLK, 1); 50597c478bd9Sstevel@tonic-gate } 50607c478bd9Sstevel@tonic-gate 50617c478bd9Sstevel@tonic-gate return (io_len); 50627c478bd9Sstevel@tonic-gate } 50637c478bd9Sstevel@tonic-gate 50647c478bd9Sstevel@tonic-gate int ufs_delay = 1; 50657c478bd9Sstevel@tonic-gate /* 50667c478bd9Sstevel@tonic-gate * Flags are composed of {B_INVAL, B_FREE, B_DONTNEED, B_FORCE, B_ASYNC} 50677c478bd9Sstevel@tonic-gate * 50687c478bd9Sstevel@tonic-gate * LMXXX - the inode really ought to contain a pointer to one of these 50697c478bd9Sstevel@tonic-gate * async args. Stuff gunk in there and just hand the whole mess off. 50707c478bd9Sstevel@tonic-gate * This would replace i_delaylen, i_delayoff. 50717c478bd9Sstevel@tonic-gate */ 50727c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 50737c478bd9Sstevel@tonic-gate static int 50747c478bd9Sstevel@tonic-gate ufs_putpage(struct vnode *vp, offset_t off, size_t len, int flags, 50757c478bd9Sstevel@tonic-gate struct cred *cr) 50767c478bd9Sstevel@tonic-gate { 50777c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 50787c478bd9Sstevel@tonic-gate int err = 0; 50797c478bd9Sstevel@tonic-gate 50807c478bd9Sstevel@tonic-gate if (vp->v_count == 0) { 50817c478bd9Sstevel@tonic-gate return (ufs_fault(vp, "ufs_putpage: bad v_count == 0")); 50827c478bd9Sstevel@tonic-gate } 50837c478bd9Sstevel@tonic-gate 50847c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_PUTPAGE_START, 50857c478bd9Sstevel@tonic-gate "ufs_putpage_start:vp %p", vp); 50867c478bd9Sstevel@tonic-gate 50877c478bd9Sstevel@tonic-gate /* 50887c478bd9Sstevel@tonic-gate * XXX - Why should this check be made here? 50897c478bd9Sstevel@tonic-gate */ 50907c478bd9Sstevel@tonic-gate if (vp->v_flag & VNOMAP) { 50917c478bd9Sstevel@tonic-gate err = ENOSYS; 50927c478bd9Sstevel@tonic-gate goto errout; 50937c478bd9Sstevel@tonic-gate } 50947c478bd9Sstevel@tonic-gate 50957c478bd9Sstevel@tonic-gate if (ip->i_ufsvfs == NULL) { 50967c478bd9Sstevel@tonic-gate err = EIO; 50977c478bd9Sstevel@tonic-gate goto errout; 50987c478bd9Sstevel@tonic-gate } 50997c478bd9Sstevel@tonic-gate 51007c478bd9Sstevel@tonic-gate if (flags & B_ASYNC) { 51017c478bd9Sstevel@tonic-gate if (ufs_delay && len && 51027c478bd9Sstevel@tonic-gate (flags & ~(B_ASYNC|B_DONTNEED|B_FREE)) == 0) { 51037c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 51047c478bd9Sstevel@tonic-gate /* 51057c478bd9Sstevel@tonic-gate * If nobody stalled, start a new cluster. 51067c478bd9Sstevel@tonic-gate */ 51077c478bd9Sstevel@tonic-gate if (ip->i_delaylen == 0) { 51087c478bd9Sstevel@tonic-gate ip->i_delayoff = off; 51097c478bd9Sstevel@tonic-gate ip->i_delaylen = len; 51107c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 51117c478bd9Sstevel@tonic-gate goto errout; 51127c478bd9Sstevel@tonic-gate } 51137c478bd9Sstevel@tonic-gate /* 51147c478bd9Sstevel@tonic-gate * If we have a full cluster or they are not contig, 51157c478bd9Sstevel@tonic-gate * then push last cluster and start over. 51167c478bd9Sstevel@tonic-gate */ 51177c478bd9Sstevel@tonic-gate if (ip->i_delaylen >= CLUSTSZ(ip) || 51187c478bd9Sstevel@tonic-gate ip->i_delayoff + ip->i_delaylen != off) { 51197c478bd9Sstevel@tonic-gate u_offset_t doff; 51207c478bd9Sstevel@tonic-gate size_t dlen; 51217c478bd9Sstevel@tonic-gate 51227c478bd9Sstevel@tonic-gate doff = ip->i_delayoff; 51237c478bd9Sstevel@tonic-gate dlen = ip->i_delaylen; 51247c478bd9Sstevel@tonic-gate ip->i_delayoff = off; 51257c478bd9Sstevel@tonic-gate ip->i_delaylen = len; 51267c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 51277c478bd9Sstevel@tonic-gate err = ufs_putpages(vp, doff, dlen, 51287c478bd9Sstevel@tonic-gate flags, cr); 51297c478bd9Sstevel@tonic-gate /* LMXXX - flags are new val, not old */ 51307c478bd9Sstevel@tonic-gate goto errout; 51317c478bd9Sstevel@tonic-gate } 51327c478bd9Sstevel@tonic-gate /* 51337c478bd9Sstevel@tonic-gate * There is something there, it's not full, and 51347c478bd9Sstevel@tonic-gate * it is contig. 51357c478bd9Sstevel@tonic-gate */ 51367c478bd9Sstevel@tonic-gate ip->i_delaylen += len; 51377c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 51387c478bd9Sstevel@tonic-gate goto errout; 51397c478bd9Sstevel@tonic-gate } 51407c478bd9Sstevel@tonic-gate /* 51417c478bd9Sstevel@tonic-gate * Must have weird flags or we are not clustering. 51427c478bd9Sstevel@tonic-gate */ 51437c478bd9Sstevel@tonic-gate } 51447c478bd9Sstevel@tonic-gate 51457c478bd9Sstevel@tonic-gate err = ufs_putpages(vp, off, len, flags, cr); 51467c478bd9Sstevel@tonic-gate 51477c478bd9Sstevel@tonic-gate errout: 51487c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_PUTPAGE_END, 51497c478bd9Sstevel@tonic-gate "ufs_putpage_end:vp %p error %d", vp, err); 51507c478bd9Sstevel@tonic-gate return (err); 51517c478bd9Sstevel@tonic-gate } 51527c478bd9Sstevel@tonic-gate 51537c478bd9Sstevel@tonic-gate /* 51547c478bd9Sstevel@tonic-gate * If len == 0, do from off to EOF. 51557c478bd9Sstevel@tonic-gate * 51567c478bd9Sstevel@tonic-gate * The normal cases should be len == 0 & off == 0 (entire vp list), 51577c478bd9Sstevel@tonic-gate * len == MAXBSIZE (from segmap_release actions), and len == PAGESIZE 51587c478bd9Sstevel@tonic-gate * (from pageout). 51597c478bd9Sstevel@tonic-gate */ 51607c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 51617c478bd9Sstevel@tonic-gate static int 51627c478bd9Sstevel@tonic-gate ufs_putpages( 51637c478bd9Sstevel@tonic-gate struct vnode *vp, 51647c478bd9Sstevel@tonic-gate offset_t off, 51657c478bd9Sstevel@tonic-gate size_t len, 51667c478bd9Sstevel@tonic-gate int flags, 51677c478bd9Sstevel@tonic-gate struct cred *cr) 51687c478bd9Sstevel@tonic-gate { 51697c478bd9Sstevel@tonic-gate u_offset_t io_off; 51707c478bd9Sstevel@tonic-gate u_offset_t eoff; 51717c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 51727c478bd9Sstevel@tonic-gate page_t *pp; 51737c478bd9Sstevel@tonic-gate size_t io_len; 51747c478bd9Sstevel@tonic-gate int err = 0; 51757c478bd9Sstevel@tonic-gate int dolock; 51767c478bd9Sstevel@tonic-gate 51777c478bd9Sstevel@tonic-gate if (vp->v_count == 0) 51787c478bd9Sstevel@tonic-gate return (ufs_fault(vp, "ufs_putpages: v_count == 0")); 51797c478bd9Sstevel@tonic-gate /* 51807c478bd9Sstevel@tonic-gate * Acquire the readers/write inode lock before locking 51817c478bd9Sstevel@tonic-gate * any pages in this inode. 51827c478bd9Sstevel@tonic-gate * The inode lock is held during i/o. 51837c478bd9Sstevel@tonic-gate */ 51847c478bd9Sstevel@tonic-gate if (len == 0) { 51857c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 51867c478bd9Sstevel@tonic-gate ip->i_delayoff = ip->i_delaylen = 0; 51877c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 51887c478bd9Sstevel@tonic-gate } 51897c478bd9Sstevel@tonic-gate dolock = (rw_owner(&ip->i_contents) != curthread); 51907c478bd9Sstevel@tonic-gate if (dolock) { 51917c478bd9Sstevel@tonic-gate /* 51927c478bd9Sstevel@tonic-gate * Must synchronize this thread and any possible thread 51937c478bd9Sstevel@tonic-gate * operating in the window of vulnerability in wrip(). 51947c478bd9Sstevel@tonic-gate * It is dangerous to allow both a thread doing a putpage 51957c478bd9Sstevel@tonic-gate * and a thread writing, so serialize them. The exception 51967c478bd9Sstevel@tonic-gate * is when the thread in wrip() does something which causes 51977c478bd9Sstevel@tonic-gate * a putpage operation. Then, the thread must be allowed 51987c478bd9Sstevel@tonic-gate * to continue. It may encounter a bmap_read problem in 51997c478bd9Sstevel@tonic-gate * ufs_putapage, but that is handled in ufs_putapage. 52007c478bd9Sstevel@tonic-gate * Allow async writers to proceed, we don't want to block 52017c478bd9Sstevel@tonic-gate * the pageout daemon. 52027c478bd9Sstevel@tonic-gate */ 52037c478bd9Sstevel@tonic-gate if (ip->i_writer == curthread) 52047c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_READER); 52057c478bd9Sstevel@tonic-gate else { 52067c478bd9Sstevel@tonic-gate for (;;) { 52077c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_READER); 52087c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 52097c478bd9Sstevel@tonic-gate /* 52107c478bd9Sstevel@tonic-gate * If there is no thread in the critical 52117c478bd9Sstevel@tonic-gate * section of wrip(), then proceed. 52127c478bd9Sstevel@tonic-gate * Otherwise, wait until there isn't one. 52137c478bd9Sstevel@tonic-gate */ 52147c478bd9Sstevel@tonic-gate if (ip->i_writer == NULL) { 52157c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 52167c478bd9Sstevel@tonic-gate break; 52177c478bd9Sstevel@tonic-gate } 52187c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 52197c478bd9Sstevel@tonic-gate /* 52207c478bd9Sstevel@tonic-gate * Bounce async writers when we have a writer 52217c478bd9Sstevel@tonic-gate * working on this file so we don't deadlock 52227c478bd9Sstevel@tonic-gate * the pageout daemon. 52237c478bd9Sstevel@tonic-gate */ 52247c478bd9Sstevel@tonic-gate if (flags & B_ASYNC) { 52257c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 52267c478bd9Sstevel@tonic-gate return (0); 52277c478bd9Sstevel@tonic-gate } 52287c478bd9Sstevel@tonic-gate cv_wait(&ip->i_wrcv, &ip->i_tlock); 52297c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 52307c478bd9Sstevel@tonic-gate } 52317c478bd9Sstevel@tonic-gate } 52327c478bd9Sstevel@tonic-gate } 52337c478bd9Sstevel@tonic-gate 52347c478bd9Sstevel@tonic-gate if (!vn_has_cached_data(vp)) { 52357c478bd9Sstevel@tonic-gate if (dolock) 52367c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 52377c478bd9Sstevel@tonic-gate return (0); 52387c478bd9Sstevel@tonic-gate } 52397c478bd9Sstevel@tonic-gate 52407c478bd9Sstevel@tonic-gate if (len == 0) { 52417c478bd9Sstevel@tonic-gate /* 52427c478bd9Sstevel@tonic-gate * Search the entire vp list for pages >= off. 52437c478bd9Sstevel@tonic-gate */ 52447c478bd9Sstevel@tonic-gate err = pvn_vplist_dirty(vp, (u_offset_t)off, ufs_putapage, 52457c478bd9Sstevel@tonic-gate flags, cr); 52467c478bd9Sstevel@tonic-gate } else { 52477c478bd9Sstevel@tonic-gate /* 52487c478bd9Sstevel@tonic-gate * Loop over all offsets in the range looking for 52497c478bd9Sstevel@tonic-gate * pages to deal with. 52507c478bd9Sstevel@tonic-gate */ 52517c478bd9Sstevel@tonic-gate if ((eoff = blkroundup(ip->i_fs, ip->i_size)) != 0) 52527c478bd9Sstevel@tonic-gate eoff = MIN(off + len, eoff); 52537c478bd9Sstevel@tonic-gate else 52547c478bd9Sstevel@tonic-gate eoff = off + len; 52557c478bd9Sstevel@tonic-gate 52567c478bd9Sstevel@tonic-gate for (io_off = off; io_off < eoff; io_off += io_len) { 52577c478bd9Sstevel@tonic-gate /* 52587c478bd9Sstevel@tonic-gate * If we are not invalidating, synchronously 52597c478bd9Sstevel@tonic-gate * freeing or writing pages, use the routine 52607c478bd9Sstevel@tonic-gate * page_lookup_nowait() to prevent reclaiming 52617c478bd9Sstevel@tonic-gate * them from the free list. 52627c478bd9Sstevel@tonic-gate */ 52637c478bd9Sstevel@tonic-gate if ((flags & B_INVAL) || ((flags & B_ASYNC) == 0)) { 52647c478bd9Sstevel@tonic-gate pp = page_lookup(vp, io_off, 52657c478bd9Sstevel@tonic-gate (flags & (B_INVAL | B_FREE)) ? 52667c478bd9Sstevel@tonic-gate SE_EXCL : SE_SHARED); 52677c478bd9Sstevel@tonic-gate } else { 52687c478bd9Sstevel@tonic-gate pp = page_lookup_nowait(vp, io_off, 52697c478bd9Sstevel@tonic-gate (flags & B_FREE) ? SE_EXCL : SE_SHARED); 52707c478bd9Sstevel@tonic-gate } 52717c478bd9Sstevel@tonic-gate 52727c478bd9Sstevel@tonic-gate if (pp == NULL || pvn_getdirty(pp, flags) == 0) 52737c478bd9Sstevel@tonic-gate io_len = PAGESIZE; 52747c478bd9Sstevel@tonic-gate else { 52757c478bd9Sstevel@tonic-gate u_offset_t *io_offp = &io_off; 52767c478bd9Sstevel@tonic-gate 52777c478bd9Sstevel@tonic-gate err = ufs_putapage(vp, pp, io_offp, &io_len, 52787c478bd9Sstevel@tonic-gate flags, cr); 52797c478bd9Sstevel@tonic-gate if (err != 0) 52807c478bd9Sstevel@tonic-gate break; 52817c478bd9Sstevel@tonic-gate /* 52827c478bd9Sstevel@tonic-gate * "io_off" and "io_len" are returned as 52837c478bd9Sstevel@tonic-gate * the range of pages we actually wrote. 52847c478bd9Sstevel@tonic-gate * This allows us to skip ahead more quickly 52857c478bd9Sstevel@tonic-gate * since several pages may've been dealt 52867c478bd9Sstevel@tonic-gate * with by this iteration of the loop. 52877c478bd9Sstevel@tonic-gate */ 52887c478bd9Sstevel@tonic-gate } 52897c478bd9Sstevel@tonic-gate } 52907c478bd9Sstevel@tonic-gate } 52917c478bd9Sstevel@tonic-gate if (err == 0 && off == 0 && (len == 0 || len >= ip->i_size)) { 52927c478bd9Sstevel@tonic-gate /* 52937c478bd9Sstevel@tonic-gate * We have just sync'ed back all the pages on 52947c478bd9Sstevel@tonic-gate * the inode, turn off the IMODTIME flag. 52957c478bd9Sstevel@tonic-gate */ 52967c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 52977c478bd9Sstevel@tonic-gate ip->i_flag &= ~IMODTIME; 52987c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 52997c478bd9Sstevel@tonic-gate } 53007c478bd9Sstevel@tonic-gate if (dolock) 53017c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 53027c478bd9Sstevel@tonic-gate return (err); 53037c478bd9Sstevel@tonic-gate } 53047c478bd9Sstevel@tonic-gate 53057c478bd9Sstevel@tonic-gate static void 53067c478bd9Sstevel@tonic-gate ufs_iodone(buf_t *bp) 53077c478bd9Sstevel@tonic-gate { 53087c478bd9Sstevel@tonic-gate struct inode *ip; 53097c478bd9Sstevel@tonic-gate 53107c478bd9Sstevel@tonic-gate ASSERT((bp->b_pages->p_vnode != NULL) && !(bp->b_flags & B_READ)); 53117c478bd9Sstevel@tonic-gate 53127c478bd9Sstevel@tonic-gate bp->b_iodone = NULL; 53137c478bd9Sstevel@tonic-gate 53147c478bd9Sstevel@tonic-gate ip = VTOI(bp->b_pages->p_vnode); 53157c478bd9Sstevel@tonic-gate 53167c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 53177c478bd9Sstevel@tonic-gate if (ip->i_writes >= ufs_LW) { 53187c478bd9Sstevel@tonic-gate if ((ip->i_writes -= bp->b_bcount) <= ufs_LW) 53197c478bd9Sstevel@tonic-gate if (ufs_WRITES) 53207c478bd9Sstevel@tonic-gate cv_broadcast(&ip->i_wrcv); /* wake all up */ 53217c478bd9Sstevel@tonic-gate } else { 53227c478bd9Sstevel@tonic-gate ip->i_writes -= bp->b_bcount; 53237c478bd9Sstevel@tonic-gate } 53247c478bd9Sstevel@tonic-gate 53257c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 53267c478bd9Sstevel@tonic-gate iodone(bp); 53277c478bd9Sstevel@tonic-gate } 53287c478bd9Sstevel@tonic-gate 53297c478bd9Sstevel@tonic-gate /* 53307c478bd9Sstevel@tonic-gate * Write out a single page, possibly klustering adjacent 53317c478bd9Sstevel@tonic-gate * dirty pages. The inode lock must be held. 53327c478bd9Sstevel@tonic-gate * 53337c478bd9Sstevel@tonic-gate * LMXXX - bsize < pagesize not done. 53347c478bd9Sstevel@tonic-gate */ 53357c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 53367c478bd9Sstevel@tonic-gate int 53377c478bd9Sstevel@tonic-gate ufs_putapage( 53387c478bd9Sstevel@tonic-gate struct vnode *vp, 53397c478bd9Sstevel@tonic-gate page_t *pp, 53407c478bd9Sstevel@tonic-gate u_offset_t *offp, 53417c478bd9Sstevel@tonic-gate size_t *lenp, /* return values */ 53427c478bd9Sstevel@tonic-gate int flags, 53437c478bd9Sstevel@tonic-gate struct cred *cr) 53447c478bd9Sstevel@tonic-gate { 53457c478bd9Sstevel@tonic-gate u_offset_t io_off; 53467c478bd9Sstevel@tonic-gate u_offset_t off; 53477c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 53487c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp = ip->i_ufsvfs; 53497c478bd9Sstevel@tonic-gate struct fs *fs; 53507c478bd9Sstevel@tonic-gate struct buf *bp; 53517c478bd9Sstevel@tonic-gate size_t io_len; 53527c478bd9Sstevel@tonic-gate daddr_t bn; 53537c478bd9Sstevel@tonic-gate int err; 53547c478bd9Sstevel@tonic-gate int contig; 53557c478bd9Sstevel@tonic-gate 53567c478bd9Sstevel@tonic-gate ASSERT(RW_LOCK_HELD(&ip->i_contents)); 53577c478bd9Sstevel@tonic-gate 53587c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_PUTAPAGE_START, 53597c478bd9Sstevel@tonic-gate "ufs_putapage_start:vp %p", vp); 53607c478bd9Sstevel@tonic-gate 53617c478bd9Sstevel@tonic-gate if (ufsvfsp == NULL) { 53627c478bd9Sstevel@tonic-gate err = EIO; 53637c478bd9Sstevel@tonic-gate goto out_trace; 53647c478bd9Sstevel@tonic-gate } 53657c478bd9Sstevel@tonic-gate 53667c478bd9Sstevel@tonic-gate fs = ip->i_fs; 53677c478bd9Sstevel@tonic-gate ASSERT(fs->fs_ronly == 0); 53687c478bd9Sstevel@tonic-gate 53697c478bd9Sstevel@tonic-gate /* 53707c478bd9Sstevel@tonic-gate * If the modified time on the inode has not already been 53717c478bd9Sstevel@tonic-gate * set elsewhere (e.g. for write/setattr) we set the time now. 53727c478bd9Sstevel@tonic-gate * This gives us approximate modified times for mmap'ed files 53737c478bd9Sstevel@tonic-gate * which are modified via stores in the user address space. 53747c478bd9Sstevel@tonic-gate */ 53757c478bd9Sstevel@tonic-gate if ((ip->i_flag & IMODTIME) == 0) { 53767c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 53777c478bd9Sstevel@tonic-gate ip->i_flag |= IUPD; 53787c478bd9Sstevel@tonic-gate ip->i_seq++; 53797c478bd9Sstevel@tonic-gate ITIMES_NOLOCK(ip); 53807c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 53817c478bd9Sstevel@tonic-gate } 53827c478bd9Sstevel@tonic-gate 53837c478bd9Sstevel@tonic-gate /* 53847c478bd9Sstevel@tonic-gate * Align the request to a block boundry (for old file systems), 53857c478bd9Sstevel@tonic-gate * and go ask bmap() how contiguous things are for this file. 53867c478bd9Sstevel@tonic-gate */ 53877c478bd9Sstevel@tonic-gate off = pp->p_offset & (offset_t)fs->fs_bmask; /* block align it */ 53887c478bd9Sstevel@tonic-gate contig = 0; 53897c478bd9Sstevel@tonic-gate err = bmap_read(ip, off, &bn, &contig); 53907c478bd9Sstevel@tonic-gate if (err) 53917c478bd9Sstevel@tonic-gate goto out; 53927c478bd9Sstevel@tonic-gate if (bn == UFS_HOLE) { /* putpage never allocates */ 53937c478bd9Sstevel@tonic-gate /* 53947c478bd9Sstevel@tonic-gate * logging device is in error mode; simply return EIO 53957c478bd9Sstevel@tonic-gate */ 53967c478bd9Sstevel@tonic-gate if (TRANS_ISERROR(ufsvfsp)) { 53977c478bd9Sstevel@tonic-gate err = EIO; 53987c478bd9Sstevel@tonic-gate goto out; 53997c478bd9Sstevel@tonic-gate } 54007c478bd9Sstevel@tonic-gate /* 54017c478bd9Sstevel@tonic-gate * Oops, the thread in the window in wrip() did some 54027c478bd9Sstevel@tonic-gate * sort of operation which caused a putpage in the bad 54037c478bd9Sstevel@tonic-gate * range. In this case, just return an error which will 54047c478bd9Sstevel@tonic-gate * cause the software modified bit on the page to set 54057c478bd9Sstevel@tonic-gate * and the page will get written out again later. 54067c478bd9Sstevel@tonic-gate */ 54077c478bd9Sstevel@tonic-gate if (ip->i_writer == curthread) { 54087c478bd9Sstevel@tonic-gate err = EIO; 54097c478bd9Sstevel@tonic-gate goto out; 54107c478bd9Sstevel@tonic-gate } 54117c478bd9Sstevel@tonic-gate /* 54127c478bd9Sstevel@tonic-gate * If the pager is trying to push a page in the bad range 54137c478bd9Sstevel@tonic-gate * just tell him to try again later when things are better. 54147c478bd9Sstevel@tonic-gate */ 54157c478bd9Sstevel@tonic-gate if (flags & B_ASYNC) { 54167c478bd9Sstevel@tonic-gate err = EAGAIN; 54177c478bd9Sstevel@tonic-gate goto out; 54187c478bd9Sstevel@tonic-gate } 54197c478bd9Sstevel@tonic-gate err = ufs_fault(ITOV(ip), "ufs_putapage: bn == UFS_HOLE"); 54207c478bd9Sstevel@tonic-gate goto out; 54217c478bd9Sstevel@tonic-gate } 54227c478bd9Sstevel@tonic-gate 5423303bf60bSsdebnath /* 5424303bf60bSsdebnath * If it is an fallocate'd block, reverse the negativity since 5425303bf60bSsdebnath * we are now writing to it 5426303bf60bSsdebnath */ 5427303bf60bSsdebnath if (ISFALLOCBLK(ip, bn)) { 5428303bf60bSsdebnath err = bmap_set_bn(vp, off, dbtofsb(fs, -bn)); 5429303bf60bSsdebnath if (err) 5430303bf60bSsdebnath goto out; 5431303bf60bSsdebnath 5432303bf60bSsdebnath bn = -bn; 5433303bf60bSsdebnath } 5434303bf60bSsdebnath 54357c478bd9Sstevel@tonic-gate /* 54367c478bd9Sstevel@tonic-gate * Take the length (of contiguous bytes) passed back from bmap() 54377c478bd9Sstevel@tonic-gate * and _try_ and get a set of pages covering that extent. 54387c478bd9Sstevel@tonic-gate */ 54397c478bd9Sstevel@tonic-gate pp = pvn_write_kluster(vp, pp, &io_off, &io_len, off, contig, flags); 54407c478bd9Sstevel@tonic-gate 54417c478bd9Sstevel@tonic-gate /* 54427c478bd9Sstevel@tonic-gate * May have run out of memory and not clustered backwards. 54437c478bd9Sstevel@tonic-gate * off p_offset 54447c478bd9Sstevel@tonic-gate * [ pp - 1 ][ pp ] 54457c478bd9Sstevel@tonic-gate * [ block ] 54467c478bd9Sstevel@tonic-gate * We told bmap off, so we have to adjust the bn accordingly. 54477c478bd9Sstevel@tonic-gate */ 54487c478bd9Sstevel@tonic-gate if (io_off > off) { 54497c478bd9Sstevel@tonic-gate bn += btod(io_off - off); 54507c478bd9Sstevel@tonic-gate contig -= (io_off - off); 54517c478bd9Sstevel@tonic-gate } 54527c478bd9Sstevel@tonic-gate 54537c478bd9Sstevel@tonic-gate /* 54547c478bd9Sstevel@tonic-gate * bmap was carefull to tell us the right size so use that. 54557c478bd9Sstevel@tonic-gate * There might be unallocated frags at the end. 54567c478bd9Sstevel@tonic-gate * LMXXX - bzero the end of the page? We must be writing after EOF. 54577c478bd9Sstevel@tonic-gate */ 54587c478bd9Sstevel@tonic-gate if (io_len > contig) { 54597c478bd9Sstevel@tonic-gate ASSERT(io_len - contig < fs->fs_bsize); 54607c478bd9Sstevel@tonic-gate io_len -= (io_len - contig); 54617c478bd9Sstevel@tonic-gate } 54627c478bd9Sstevel@tonic-gate 54637c478bd9Sstevel@tonic-gate /* 54647c478bd9Sstevel@tonic-gate * Handle the case where we are writing the last page after EOF. 54657c478bd9Sstevel@tonic-gate * 54667c478bd9Sstevel@tonic-gate * XXX - just a patch for i-mt3. 54677c478bd9Sstevel@tonic-gate */ 54687c478bd9Sstevel@tonic-gate if (io_len == 0) { 54697c478bd9Sstevel@tonic-gate ASSERT(pp->p_offset >= (u_offset_t)(roundup(ip->i_size, 54707c478bd9Sstevel@tonic-gate PAGESIZE))); 54717c478bd9Sstevel@tonic-gate io_len = PAGESIZE; 54727c478bd9Sstevel@tonic-gate } 54737c478bd9Sstevel@tonic-gate 54747c478bd9Sstevel@tonic-gate bp = pageio_setup(pp, io_len, ip->i_devvp, B_WRITE | flags); 54757c478bd9Sstevel@tonic-gate 54767c478bd9Sstevel@tonic-gate ULOCKFS_SET_MOD(ITOUL(ip)); 54777c478bd9Sstevel@tonic-gate 54787c478bd9Sstevel@tonic-gate bp->b_edev = ip->i_dev; 54797c478bd9Sstevel@tonic-gate bp->b_dev = cmpdev(ip->i_dev); 54807c478bd9Sstevel@tonic-gate bp->b_blkno = bn; 54817c478bd9Sstevel@tonic-gate bp->b_un.b_addr = (caddr_t)0; 54827c478bd9Sstevel@tonic-gate bp->b_file = ip->i_vnode; 54837c478bd9Sstevel@tonic-gate 54847c478bd9Sstevel@tonic-gate if (TRANS_ISTRANS(ufsvfsp)) { 54857c478bd9Sstevel@tonic-gate if ((ip->i_mode & IFMT) == IFSHAD) { 54867c478bd9Sstevel@tonic-gate TRANS_BUF(ufsvfsp, 0, io_len, bp, DT_SHAD); 54877c478bd9Sstevel@tonic-gate } else if (ufsvfsp->vfs_qinod == ip) { 54887c478bd9Sstevel@tonic-gate TRANS_DELTA(ufsvfsp, ldbtob(bn), bp->b_bcount, DT_QR, 54897c478bd9Sstevel@tonic-gate 0, 0); 54907c478bd9Sstevel@tonic-gate } 54917c478bd9Sstevel@tonic-gate } 54927c478bd9Sstevel@tonic-gate 54937c478bd9Sstevel@tonic-gate /* write throttle */ 54947c478bd9Sstevel@tonic-gate 54957c478bd9Sstevel@tonic-gate ASSERT(bp->b_iodone == NULL); 54967c478bd9Sstevel@tonic-gate bp->b_iodone = (int (*)())ufs_iodone; 54977c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 54987c478bd9Sstevel@tonic-gate ip->i_writes += bp->b_bcount; 54997c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 55007c478bd9Sstevel@tonic-gate 55017c478bd9Sstevel@tonic-gate if (bp->b_flags & B_ASYNC) { 55027c478bd9Sstevel@tonic-gate if (ufsvfsp->vfs_log) { 55037c478bd9Sstevel@tonic-gate lufs_write_strategy(ufsvfsp->vfs_log, bp); 55047c478bd9Sstevel@tonic-gate } else if (ufsvfsp->vfs_snapshot) { 55057c478bd9Sstevel@tonic-gate fssnap_strategy(&ufsvfsp->vfs_snapshot, bp); 55067c478bd9Sstevel@tonic-gate } else { 55077c478bd9Sstevel@tonic-gate ufsvfsp->vfs_iotstamp = lbolt; 55087c478bd9Sstevel@tonic-gate ub.ub_putasyncs.value.ul++; 55097c478bd9Sstevel@tonic-gate (void) bdev_strategy(bp); 55107c478bd9Sstevel@tonic-gate lwp_stat_update(LWP_STAT_OUBLK, 1); 55117c478bd9Sstevel@tonic-gate } 55127c478bd9Sstevel@tonic-gate } else { 55137c478bd9Sstevel@tonic-gate if (ufsvfsp->vfs_log) { 55147c478bd9Sstevel@tonic-gate lufs_write_strategy(ufsvfsp->vfs_log, bp); 55157c478bd9Sstevel@tonic-gate } else if (ufsvfsp->vfs_snapshot) { 55167c478bd9Sstevel@tonic-gate fssnap_strategy(&ufsvfsp->vfs_snapshot, bp); 55177c478bd9Sstevel@tonic-gate } else { 55187c478bd9Sstevel@tonic-gate ufsvfsp->vfs_iotstamp = lbolt; 55197c478bd9Sstevel@tonic-gate ub.ub_putsyncs.value.ul++; 55207c478bd9Sstevel@tonic-gate (void) bdev_strategy(bp); 55217c478bd9Sstevel@tonic-gate lwp_stat_update(LWP_STAT_OUBLK, 1); 55227c478bd9Sstevel@tonic-gate } 55237c478bd9Sstevel@tonic-gate err = biowait(bp); 55247c478bd9Sstevel@tonic-gate pageio_done(bp); 55257c478bd9Sstevel@tonic-gate pvn_write_done(pp, ((err) ? B_ERROR : 0) | B_WRITE | flags); 55267c478bd9Sstevel@tonic-gate } 55277c478bd9Sstevel@tonic-gate 55287c478bd9Sstevel@tonic-gate pp = NULL; 55297c478bd9Sstevel@tonic-gate 55307c478bd9Sstevel@tonic-gate out: 55317c478bd9Sstevel@tonic-gate if (err != 0 && pp != NULL) 55327c478bd9Sstevel@tonic-gate pvn_write_done(pp, B_ERROR | B_WRITE | flags); 55337c478bd9Sstevel@tonic-gate 55347c478bd9Sstevel@tonic-gate if (offp) 55357c478bd9Sstevel@tonic-gate *offp = io_off; 55367c478bd9Sstevel@tonic-gate if (lenp) 55377c478bd9Sstevel@tonic-gate *lenp = io_len; 55387c478bd9Sstevel@tonic-gate out_trace: 55397c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_PUTAPAGE_END, 55407c478bd9Sstevel@tonic-gate "ufs_putapage_end:vp %p error %d", vp, err); 55417c478bd9Sstevel@tonic-gate return (err); 55427c478bd9Sstevel@tonic-gate } 55437c478bd9Sstevel@tonic-gate 55447c478bd9Sstevel@tonic-gate /* ARGSUSED */ 55457c478bd9Sstevel@tonic-gate static int 55467c478bd9Sstevel@tonic-gate ufs_map(struct vnode *vp, 55477c478bd9Sstevel@tonic-gate offset_t off, 55487c478bd9Sstevel@tonic-gate struct as *as, 55497c478bd9Sstevel@tonic-gate caddr_t *addrp, 55507c478bd9Sstevel@tonic-gate size_t len, 55517c478bd9Sstevel@tonic-gate uchar_t prot, 55527c478bd9Sstevel@tonic-gate uchar_t maxprot, 55537c478bd9Sstevel@tonic-gate uint_t flags, 55547c478bd9Sstevel@tonic-gate struct cred *cr) 55557c478bd9Sstevel@tonic-gate { 55567c478bd9Sstevel@tonic-gate struct segvn_crargs vn_a; 55577c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp = VTOI(vp)->i_ufsvfs; 55587c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 55597c478bd9Sstevel@tonic-gate int error; 55607c478bd9Sstevel@tonic-gate 55617c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_MAP_START, 55627c478bd9Sstevel@tonic-gate "ufs_map_start:vp %p", vp); 55637c478bd9Sstevel@tonic-gate 5564*02ff05a9Svsakar retry_map: 55657c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_MAP_MASK); 55667c478bd9Sstevel@tonic-gate if (error) 55677c478bd9Sstevel@tonic-gate goto out; 55687c478bd9Sstevel@tonic-gate 55697c478bd9Sstevel@tonic-gate if (vp->v_flag & VNOMAP) { 55707c478bd9Sstevel@tonic-gate error = ENOSYS; 55717c478bd9Sstevel@tonic-gate goto unlock; 55727c478bd9Sstevel@tonic-gate } 55737c478bd9Sstevel@tonic-gate 55747c478bd9Sstevel@tonic-gate if (off < (offset_t)0 || (offset_t)(off + len) < (offset_t)0) { 55757c478bd9Sstevel@tonic-gate error = ENXIO; 55767c478bd9Sstevel@tonic-gate goto unlock; 55777c478bd9Sstevel@tonic-gate } 55787c478bd9Sstevel@tonic-gate 55797c478bd9Sstevel@tonic-gate if (vp->v_type != VREG) { 55807c478bd9Sstevel@tonic-gate error = ENODEV; 55817c478bd9Sstevel@tonic-gate goto unlock; 55827c478bd9Sstevel@tonic-gate } 55837c478bd9Sstevel@tonic-gate 55847c478bd9Sstevel@tonic-gate /* 55857c478bd9Sstevel@tonic-gate * If file is being locked, disallow mapping. 55867c478bd9Sstevel@tonic-gate */ 55877c478bd9Sstevel@tonic-gate if (vn_has_mandatory_locks(vp, VTOI(vp)->i_mode)) { 55887c478bd9Sstevel@tonic-gate error = EAGAIN; 55897c478bd9Sstevel@tonic-gate goto unlock; 55907c478bd9Sstevel@tonic-gate } 55917c478bd9Sstevel@tonic-gate 55927c478bd9Sstevel@tonic-gate as_rangelock(as); 55937c478bd9Sstevel@tonic-gate if ((flags & MAP_FIXED) == 0) { 55947c478bd9Sstevel@tonic-gate map_addr(addrp, len, off, 1, flags); 55957c478bd9Sstevel@tonic-gate if (*addrp == NULL) { 55967c478bd9Sstevel@tonic-gate as_rangeunlock(as); 55977c478bd9Sstevel@tonic-gate error = ENOMEM; 55987c478bd9Sstevel@tonic-gate goto unlock; 55997c478bd9Sstevel@tonic-gate } 56007c478bd9Sstevel@tonic-gate } else { 56017c478bd9Sstevel@tonic-gate /* 56027c478bd9Sstevel@tonic-gate * User specified address - blow away any previous mappings 56037c478bd9Sstevel@tonic-gate */ 56047c478bd9Sstevel@tonic-gate (void) as_unmap(as, *addrp, len); 56057c478bd9Sstevel@tonic-gate } 56067c478bd9Sstevel@tonic-gate 56077c478bd9Sstevel@tonic-gate vn_a.vp = vp; 56087c478bd9Sstevel@tonic-gate vn_a.offset = (u_offset_t)off; 56097c478bd9Sstevel@tonic-gate vn_a.type = flags & MAP_TYPE; 56107c478bd9Sstevel@tonic-gate vn_a.prot = prot; 56117c478bd9Sstevel@tonic-gate vn_a.maxprot = maxprot; 56127c478bd9Sstevel@tonic-gate vn_a.cred = cr; 56137c478bd9Sstevel@tonic-gate vn_a.amp = NULL; 56147c478bd9Sstevel@tonic-gate vn_a.flags = flags & ~MAP_TYPE; 56157c478bd9Sstevel@tonic-gate vn_a.szc = 0; 56167c478bd9Sstevel@tonic-gate vn_a.lgrp_mem_policy_flags = 0; 56177c478bd9Sstevel@tonic-gate 5618*02ff05a9Svsakar retry_lock: 5619*02ff05a9Svsakar if (!AS_LOCK_TRYENTER(ias, &as->a_lock, RW_WRITER)) { 5620*02ff05a9Svsakar /* 5621*02ff05a9Svsakar * We didn't get the lock. Check if the SLOCK is set in the 5622*02ff05a9Svsakar * ufsvfs. If yes, we might be in a deadlock. Safer to give up 5623*02ff05a9Svsakar * and wait for SLOCK to be cleared. 5624*02ff05a9Svsakar */ 5625*02ff05a9Svsakar 5626*02ff05a9Svsakar if (ulp && ULOCKFS_IS_SLOCK(ulp)) { 5627*02ff05a9Svsakar as_rangeunlock(as); 5628*02ff05a9Svsakar ufs_lockfs_end(ulp); 5629*02ff05a9Svsakar goto retry_map; 5630*02ff05a9Svsakar } else { 5631*02ff05a9Svsakar /* 5632*02ff05a9Svsakar * SLOCK isn't set so this is a genuine synchronization 5633*02ff05a9Svsakar * case. Let's try again after giving them a breather. 5634*02ff05a9Svsakar */ 5635*02ff05a9Svsakar delay(RETRY_LOCK_DELAY); 5636*02ff05a9Svsakar goto retry_lock; 5637*02ff05a9Svsakar } 5638*02ff05a9Svsakar } 5639*02ff05a9Svsakar error = as_map_locked(as, *addrp, len, segvn_create, &vn_a); 56407c478bd9Sstevel@tonic-gate as_rangeunlock(as); 56417c478bd9Sstevel@tonic-gate 56427c478bd9Sstevel@tonic-gate unlock: 56437c478bd9Sstevel@tonic-gate if (ulp) { 56447c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 56457c478bd9Sstevel@tonic-gate } 56467c478bd9Sstevel@tonic-gate out: 56477c478bd9Sstevel@tonic-gate TRACE_2(TR_FAC_UFS, TR_UFS_MAP_END, 56487c478bd9Sstevel@tonic-gate "ufs_map_end:vp %p error %d", vp, error); 56497c478bd9Sstevel@tonic-gate return (error); 56507c478bd9Sstevel@tonic-gate } 56517c478bd9Sstevel@tonic-gate 56527c478bd9Sstevel@tonic-gate /* ARGSUSED */ 56537c478bd9Sstevel@tonic-gate static int 56547c478bd9Sstevel@tonic-gate ufs_addmap(struct vnode *vp, 56557c478bd9Sstevel@tonic-gate offset_t off, 56567c478bd9Sstevel@tonic-gate struct as *as, 56577c478bd9Sstevel@tonic-gate caddr_t addr, 56587c478bd9Sstevel@tonic-gate size_t len, 56597c478bd9Sstevel@tonic-gate uchar_t prot, 56607c478bd9Sstevel@tonic-gate uchar_t maxprot, 56617c478bd9Sstevel@tonic-gate uint_t flags, 56627c478bd9Sstevel@tonic-gate struct cred *cr) 56637c478bd9Sstevel@tonic-gate { 56647c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 56657c478bd9Sstevel@tonic-gate 56667c478bd9Sstevel@tonic-gate if (vp->v_flag & VNOMAP) { 56677c478bd9Sstevel@tonic-gate return (ENOSYS); 56687c478bd9Sstevel@tonic-gate } 56697c478bd9Sstevel@tonic-gate 56707c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 56717c478bd9Sstevel@tonic-gate ip->i_mapcnt += btopr(len); 56727c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 56737c478bd9Sstevel@tonic-gate return (0); 56747c478bd9Sstevel@tonic-gate } 56757c478bd9Sstevel@tonic-gate 56767c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 56777c478bd9Sstevel@tonic-gate static int 56787c478bd9Sstevel@tonic-gate ufs_delmap(struct vnode *vp, offset_t off, struct as *as, caddr_t addr, 56797c478bd9Sstevel@tonic-gate size_t len, uint_t prot, uint_t maxprot, uint_t flags, 56807c478bd9Sstevel@tonic-gate struct cred *cr) 56817c478bd9Sstevel@tonic-gate { 56827c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 56837c478bd9Sstevel@tonic-gate 56847c478bd9Sstevel@tonic-gate if (vp->v_flag & VNOMAP) { 56857c478bd9Sstevel@tonic-gate return (ENOSYS); 56867c478bd9Sstevel@tonic-gate } 56877c478bd9Sstevel@tonic-gate 56887c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 56897c478bd9Sstevel@tonic-gate ip->i_mapcnt -= btopr(len); /* Count released mappings */ 56907c478bd9Sstevel@tonic-gate ASSERT(ip->i_mapcnt >= 0); 56917c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 56927c478bd9Sstevel@tonic-gate return (0); 56937c478bd9Sstevel@tonic-gate } 56947c478bd9Sstevel@tonic-gate /* 56957c478bd9Sstevel@tonic-gate * Return the answer requested to poll() for non-device files 56967c478bd9Sstevel@tonic-gate */ 56977c478bd9Sstevel@tonic-gate struct pollhead ufs_pollhd; 56987c478bd9Sstevel@tonic-gate 56997c478bd9Sstevel@tonic-gate /* ARGSUSED */ 57007c478bd9Sstevel@tonic-gate int 57017c478bd9Sstevel@tonic-gate ufs_poll(vnode_t *vp, short ev, int any, short *revp, struct pollhead **phpp) 57027c478bd9Sstevel@tonic-gate { 57037c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp; 57047c478bd9Sstevel@tonic-gate 57057c478bd9Sstevel@tonic-gate *revp = 0; 57067c478bd9Sstevel@tonic-gate ufsvfsp = VTOI(vp)->i_ufsvfs; 57077c478bd9Sstevel@tonic-gate 57087c478bd9Sstevel@tonic-gate if (!ufsvfsp) { 57097c478bd9Sstevel@tonic-gate *revp = POLLHUP; 57107c478bd9Sstevel@tonic-gate goto out; 57117c478bd9Sstevel@tonic-gate } 57127c478bd9Sstevel@tonic-gate 57137c478bd9Sstevel@tonic-gate if (ULOCKFS_IS_HLOCK(&ufsvfsp->vfs_ulockfs) || 57147c478bd9Sstevel@tonic-gate ULOCKFS_IS_ELOCK(&ufsvfsp->vfs_ulockfs)) { 57157c478bd9Sstevel@tonic-gate *revp |= POLLERR; 57167c478bd9Sstevel@tonic-gate 57177c478bd9Sstevel@tonic-gate } else { 57187c478bd9Sstevel@tonic-gate if ((ev & POLLOUT) && !ufsvfsp->vfs_fs->fs_ronly && 57197c478bd9Sstevel@tonic-gate !ULOCKFS_IS_WLOCK(&ufsvfsp->vfs_ulockfs)) 57207c478bd9Sstevel@tonic-gate *revp |= POLLOUT; 57217c478bd9Sstevel@tonic-gate 57227c478bd9Sstevel@tonic-gate if ((ev & POLLWRBAND) && !ufsvfsp->vfs_fs->fs_ronly && 57237c478bd9Sstevel@tonic-gate !ULOCKFS_IS_WLOCK(&ufsvfsp->vfs_ulockfs)) 57247c478bd9Sstevel@tonic-gate *revp |= POLLWRBAND; 57257c478bd9Sstevel@tonic-gate 57267c478bd9Sstevel@tonic-gate if (ev & POLLIN) 57277c478bd9Sstevel@tonic-gate *revp |= POLLIN; 57287c478bd9Sstevel@tonic-gate 57297c478bd9Sstevel@tonic-gate if (ev & POLLRDNORM) 57307c478bd9Sstevel@tonic-gate *revp |= POLLRDNORM; 57317c478bd9Sstevel@tonic-gate 57327c478bd9Sstevel@tonic-gate if (ev & POLLRDBAND) 57337c478bd9Sstevel@tonic-gate *revp |= POLLRDBAND; 57347c478bd9Sstevel@tonic-gate } 57357c478bd9Sstevel@tonic-gate 57367c478bd9Sstevel@tonic-gate if ((ev & POLLPRI) && (*revp & (POLLERR|POLLHUP))) 57377c478bd9Sstevel@tonic-gate *revp |= POLLPRI; 57387c478bd9Sstevel@tonic-gate out: 57397c478bd9Sstevel@tonic-gate *phpp = !any && !*revp ? &ufs_pollhd : (struct pollhead *)NULL; 57407c478bd9Sstevel@tonic-gate 57417c478bd9Sstevel@tonic-gate return (0); 57427c478bd9Sstevel@tonic-gate } 57437c478bd9Sstevel@tonic-gate 57447c478bd9Sstevel@tonic-gate /* ARGSUSED */ 57457c478bd9Sstevel@tonic-gate static int 57467c478bd9Sstevel@tonic-gate ufs_l_pathconf(struct vnode *vp, int cmd, ulong_t *valp, struct cred *cr) 57477c478bd9Sstevel@tonic-gate { 57487c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp = VTOI(vp)->i_ufsvfs; 57497c478bd9Sstevel@tonic-gate struct ulockfs *ulp = NULL; 57507c478bd9Sstevel@tonic-gate struct inode *sip = NULL; 57517c478bd9Sstevel@tonic-gate int error; 57527c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 57537c478bd9Sstevel@tonic-gate int issync; 57547c478bd9Sstevel@tonic-gate 57557c478bd9Sstevel@tonic-gate error = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_PATHCONF_MASK); 57567c478bd9Sstevel@tonic-gate if (error) 57577c478bd9Sstevel@tonic-gate return (error); 57587c478bd9Sstevel@tonic-gate 57597c478bd9Sstevel@tonic-gate switch (cmd) { 57607c478bd9Sstevel@tonic-gate /* 57617c478bd9Sstevel@tonic-gate * Have to handle _PC_NAME_MAX here, because the normal way 57627c478bd9Sstevel@tonic-gate * [fs_pathconf() -> VOP_STATVFS() -> ufs_statvfs()] 57637c478bd9Sstevel@tonic-gate * results in a lock ordering reversal between 57647c478bd9Sstevel@tonic-gate * ufs_lockfs_{begin,end}() and 57657c478bd9Sstevel@tonic-gate * ufs_thread_{suspend,continue}(). 57667c478bd9Sstevel@tonic-gate * 57677c478bd9Sstevel@tonic-gate * Keep in sync with ufs_statvfs(). 57687c478bd9Sstevel@tonic-gate */ 57697c478bd9Sstevel@tonic-gate case _PC_NAME_MAX: 57707c478bd9Sstevel@tonic-gate *valp = MAXNAMLEN; 57717c478bd9Sstevel@tonic-gate break; 57727c478bd9Sstevel@tonic-gate 57737c478bd9Sstevel@tonic-gate case _PC_FILESIZEBITS: 57747c478bd9Sstevel@tonic-gate if (ufsvfsp->vfs_lfflags & UFS_LARGEFILES) 57757c478bd9Sstevel@tonic-gate *valp = UFS_FILESIZE_BITS; 57767c478bd9Sstevel@tonic-gate else 57777c478bd9Sstevel@tonic-gate *valp = 32; 57787c478bd9Sstevel@tonic-gate break; 57797c478bd9Sstevel@tonic-gate 57807c478bd9Sstevel@tonic-gate case _PC_XATTR_EXISTS: 57817c478bd9Sstevel@tonic-gate if (vp->v_vfsp->vfs_flag & VFS_XATTR) { 57827c478bd9Sstevel@tonic-gate 57837c478bd9Sstevel@tonic-gate error = ufs_xattr_getattrdir(vp, &sip, LOOKUP_XATTR, 57847c478bd9Sstevel@tonic-gate cr); 57857c478bd9Sstevel@tonic-gate if (error == 0 && sip != NULL) { 57867c478bd9Sstevel@tonic-gate /* Start transaction */ 57877c478bd9Sstevel@tonic-gate if (ulp) { 57887c478bd9Sstevel@tonic-gate TRANS_BEGIN_CSYNC(ufsvfsp, issync, 57897c478bd9Sstevel@tonic-gate TOP_RMDIR, TOP_RMDIR_SIZE); 57907c478bd9Sstevel@tonic-gate } 57917c478bd9Sstevel@tonic-gate /* 57927c478bd9Sstevel@tonic-gate * Is directory empty 57937c478bd9Sstevel@tonic-gate */ 57947c478bd9Sstevel@tonic-gate rw_enter(&sip->i_rwlock, RW_WRITER); 57957c478bd9Sstevel@tonic-gate rw_enter(&sip->i_contents, RW_WRITER); 57967c478bd9Sstevel@tonic-gate if (ufs_xattrdirempty(sip, 57977c478bd9Sstevel@tonic-gate sip->i_number, CRED())) { 57987c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_WRITER); 57997c478bd9Sstevel@tonic-gate ufs_unhook_shadow(ip, sip); 58007c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 58017c478bd9Sstevel@tonic-gate 58027c478bd9Sstevel@tonic-gate *valp = 0; 58037c478bd9Sstevel@tonic-gate 58047c478bd9Sstevel@tonic-gate } else 58057c478bd9Sstevel@tonic-gate *valp = 1; 58067c478bd9Sstevel@tonic-gate rw_exit(&sip->i_contents); 58077c478bd9Sstevel@tonic-gate rw_exit(&sip->i_rwlock); 58087c478bd9Sstevel@tonic-gate if (ulp) { 58097c478bd9Sstevel@tonic-gate TRANS_END_CSYNC(ufsvfsp, error, issync, 58107c478bd9Sstevel@tonic-gate TOP_RMDIR, TOP_RMDIR_SIZE); 58117c478bd9Sstevel@tonic-gate } 58127c478bd9Sstevel@tonic-gate VN_RELE(ITOV(sip)); 58137c478bd9Sstevel@tonic-gate } else if (error == ENOENT) { 58147c478bd9Sstevel@tonic-gate *valp = 0; 58157c478bd9Sstevel@tonic-gate error = 0; 58167c478bd9Sstevel@tonic-gate } 58177c478bd9Sstevel@tonic-gate } else { 58187c478bd9Sstevel@tonic-gate error = fs_pathconf(vp, cmd, valp, cr); 58197c478bd9Sstevel@tonic-gate } 58207c478bd9Sstevel@tonic-gate break; 58217c478bd9Sstevel@tonic-gate 58227c478bd9Sstevel@tonic-gate case _PC_ACL_ENABLED: 58237c478bd9Sstevel@tonic-gate *valp = _ACL_ACLENT_ENABLED; 58247c478bd9Sstevel@tonic-gate break; 58257c478bd9Sstevel@tonic-gate 58267c478bd9Sstevel@tonic-gate case _PC_MIN_HOLE_SIZE: 58277c478bd9Sstevel@tonic-gate *valp = (ulong_t)ip->i_fs->fs_bsize; 58287c478bd9Sstevel@tonic-gate break; 58297c478bd9Sstevel@tonic-gate 58307c478bd9Sstevel@tonic-gate default: 58317c478bd9Sstevel@tonic-gate error = fs_pathconf(vp, cmd, valp, cr); 58327c478bd9Sstevel@tonic-gate } 58337c478bd9Sstevel@tonic-gate 58347c478bd9Sstevel@tonic-gate if (ulp != NULL) { 58357c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 58367c478bd9Sstevel@tonic-gate } 58377c478bd9Sstevel@tonic-gate return (error); 58387c478bd9Sstevel@tonic-gate } 58397c478bd9Sstevel@tonic-gate 58407c478bd9Sstevel@tonic-gate int ufs_pageio_writes, ufs_pageio_reads; 58417c478bd9Sstevel@tonic-gate 58427c478bd9Sstevel@tonic-gate /*ARGSUSED*/ 58437c478bd9Sstevel@tonic-gate static int 58447c478bd9Sstevel@tonic-gate ufs_pageio(struct vnode *vp, page_t *pp, u_offset_t io_off, size_t io_len, 58457c478bd9Sstevel@tonic-gate int flags, struct cred *cr) 58467c478bd9Sstevel@tonic-gate { 58477c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 58487c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp; 58497c478bd9Sstevel@tonic-gate page_t *npp = NULL, *opp = NULL, *cpp = pp; 58507c478bd9Sstevel@tonic-gate struct buf *bp; 58517c478bd9Sstevel@tonic-gate daddr_t bn; 58527c478bd9Sstevel@tonic-gate size_t done_len = 0, cur_len = 0; 58537c478bd9Sstevel@tonic-gate int err = 0; 58547c478bd9Sstevel@tonic-gate int contig = 0; 58557c478bd9Sstevel@tonic-gate int dolock; 58567c478bd9Sstevel@tonic-gate int vmpss = 0; 5857bc69f433Saguzovsk struct ulockfs *ulp; 58587c478bd9Sstevel@tonic-gate 58597c478bd9Sstevel@tonic-gate if ((flags & B_READ) && pp != NULL && pp->p_vnode == vp && 58607c478bd9Sstevel@tonic-gate vp->v_mpssdata != NULL) { 58617c478bd9Sstevel@tonic-gate vmpss = 1; 58627c478bd9Sstevel@tonic-gate } 58637c478bd9Sstevel@tonic-gate 58647c478bd9Sstevel@tonic-gate dolock = (rw_owner(&ip->i_contents) != curthread); 58657c478bd9Sstevel@tonic-gate /* 58667c478bd9Sstevel@tonic-gate * We need a better check. Ideally, we would use another 58677c478bd9Sstevel@tonic-gate * vnodeops so that hlocked and forcibly unmounted file 58687c478bd9Sstevel@tonic-gate * systems would return EIO where appropriate and w/o the 58697c478bd9Sstevel@tonic-gate * need for these checks. 58707c478bd9Sstevel@tonic-gate */ 58717c478bd9Sstevel@tonic-gate if ((ufsvfsp = ip->i_ufsvfs) == NULL) 58727c478bd9Sstevel@tonic-gate return (EIO); 58737c478bd9Sstevel@tonic-gate 5874bc69f433Saguzovsk /* 5875bc69f433Saguzovsk * For vmpss (pp can be NULL) case respect the quiesce protocol. 5876bc69f433Saguzovsk * ul_lock must be taken before locking pages so we can't use it here 5877bc69f433Saguzovsk * if pp is non NULL because segvn already locked pages 5878bc69f433Saguzovsk * SE_EXCL. Instead we rely on the fact that a forced umount or 5879bc69f433Saguzovsk * applying a filesystem lock via ufs_fiolfs() will block in the 5880bc69f433Saguzovsk * implicit call to ufs_flush() until we unlock the pages after the 5881bc69f433Saguzovsk * return to segvn. Other ufs_quiesce() callers keep ufs_quiesce_pend 5882bc69f433Saguzovsk * above 0 until they are done. We have to be careful not to increment 5883bc69f433Saguzovsk * ul_vnops_cnt here after forceful unmount hlocks the file system. 5884bc69f433Saguzovsk * 5885bc69f433Saguzovsk * If pp is NULL use ul_lock to make sure we don't increment 5886bc69f433Saguzovsk * ul_vnops_cnt after forceful unmount hlocks the file system. 5887bc69f433Saguzovsk */ 5888bc69f433Saguzovsk if (vmpss || pp == NULL) { 5889bc69f433Saguzovsk ulp = &ufsvfsp->vfs_ulockfs; 5890bc69f433Saguzovsk if (pp == NULL) 5891bc69f433Saguzovsk mutex_enter(&ulp->ul_lock); 5892bc69f433Saguzovsk if (ulp->ul_fs_lock & ULOCKFS_GETREAD_MASK) { 5893bc69f433Saguzovsk if (pp == NULL) { 5894bc69f433Saguzovsk mutex_exit(&ulp->ul_lock); 5895bc69f433Saguzovsk } 5896bc69f433Saguzovsk return (vmpss ? EIO : EINVAL); 5897bc69f433Saguzovsk } 5898bc69f433Saguzovsk atomic_add_long(&ulp->ul_vnops_cnt, 1); 5899bc69f433Saguzovsk if (pp == NULL) 5900bc69f433Saguzovsk mutex_exit(&ulp->ul_lock); 5901bc69f433Saguzovsk if (ufs_quiesce_pend) { 5902bc69f433Saguzovsk if (!atomic_add_long_nv(&ulp->ul_vnops_cnt, -1)) 5903bc69f433Saguzovsk cv_broadcast(&ulp->ul_cv); 5904bc69f433Saguzovsk return (vmpss ? EIO : EINVAL); 5905bc69f433Saguzovsk } 5906bc69f433Saguzovsk } 5907bc69f433Saguzovsk 59087c478bd9Sstevel@tonic-gate if (dolock) { 59097c478bd9Sstevel@tonic-gate /* 59107c478bd9Sstevel@tonic-gate * segvn may call VOP_PAGEIO() instead of VOP_GETPAGE() to 59117c478bd9Sstevel@tonic-gate * handle a fault against a segment that maps vnode pages with 59127c478bd9Sstevel@tonic-gate * large mappings. Segvn creates pages and holds them locked 59137c478bd9Sstevel@tonic-gate * SE_EXCL during VOP_PAGEIO() call. In this case we have to 59147c478bd9Sstevel@tonic-gate * use rw_tryenter() to avoid a potential deadlock since in 59157c478bd9Sstevel@tonic-gate * lock order i_contents needs to be taken first. 59167c478bd9Sstevel@tonic-gate * Segvn will retry via VOP_GETPAGE() if VOP_PAGEIO() fails. 59177c478bd9Sstevel@tonic-gate */ 59187c478bd9Sstevel@tonic-gate if (!vmpss) { 59197c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_READER); 59207c478bd9Sstevel@tonic-gate } else if (!rw_tryenter(&ip->i_contents, RW_READER)) { 5921bc69f433Saguzovsk if (!atomic_add_long_nv(&ulp->ul_vnops_cnt, -1)) 5922bc69f433Saguzovsk cv_broadcast(&ulp->ul_cv); 59237c478bd9Sstevel@tonic-gate return (EDEADLK); 59247c478bd9Sstevel@tonic-gate } 59257c478bd9Sstevel@tonic-gate } 59267c478bd9Sstevel@tonic-gate 592737fbc076Saguzovsk /* 592837fbc076Saguzovsk * Return an error to segvn because the pagefault request is beyond 592937fbc076Saguzovsk * PAGESIZE rounded EOF. 593037fbc076Saguzovsk */ 593137fbc076Saguzovsk if (vmpss && btopr(io_off + io_len) > btopr(ip->i_size)) { 593237fbc076Saguzovsk if (dolock) 593337fbc076Saguzovsk rw_exit(&ip->i_contents); 5934bc69f433Saguzovsk if (!atomic_add_long_nv(&ulp->ul_vnops_cnt, -1)) 5935bc69f433Saguzovsk cv_broadcast(&ulp->ul_cv); 593637fbc076Saguzovsk return (EFAULT); 593737fbc076Saguzovsk } 593837fbc076Saguzovsk 59397c478bd9Sstevel@tonic-gate if (pp == NULL) { 59407c478bd9Sstevel@tonic-gate if (bmap_has_holes(ip)) { 59417c478bd9Sstevel@tonic-gate err = ENOSYS; 59427c478bd9Sstevel@tonic-gate } else { 59437c478bd9Sstevel@tonic-gate err = EINVAL; 59447c478bd9Sstevel@tonic-gate } 59457c478bd9Sstevel@tonic-gate if (dolock) 59467c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 5947bc69f433Saguzovsk if (!atomic_add_long_nv(&ulp->ul_vnops_cnt, -1)) 5948bc69f433Saguzovsk cv_broadcast(&ulp->ul_cv); 59497c478bd9Sstevel@tonic-gate return (err); 59507c478bd9Sstevel@tonic-gate } 59517c478bd9Sstevel@tonic-gate 59527c478bd9Sstevel@tonic-gate /* 59537c478bd9Sstevel@tonic-gate * Break the io request into chunks, one for each contiguous 59547c478bd9Sstevel@tonic-gate * stretch of disk blocks in the target file. 59557c478bd9Sstevel@tonic-gate */ 59567c478bd9Sstevel@tonic-gate while (done_len < io_len) { 59577c478bd9Sstevel@tonic-gate ASSERT(cpp); 59587c478bd9Sstevel@tonic-gate contig = 0; 59597c478bd9Sstevel@tonic-gate if (err = bmap_read(ip, (u_offset_t)(io_off + done_len), 59607c478bd9Sstevel@tonic-gate &bn, &contig)) 59617c478bd9Sstevel@tonic-gate break; 59627c478bd9Sstevel@tonic-gate 59637c478bd9Sstevel@tonic-gate if (bn == UFS_HOLE) { /* No holey swapfiles */ 59647c478bd9Sstevel@tonic-gate if (vmpss) { 59657c478bd9Sstevel@tonic-gate err = EFAULT; 59667c478bd9Sstevel@tonic-gate break; 59677c478bd9Sstevel@tonic-gate } 59687c478bd9Sstevel@tonic-gate err = ufs_fault(ITOV(ip), "ufs_pageio: bn == UFS_HOLE"); 59697c478bd9Sstevel@tonic-gate break; 59707c478bd9Sstevel@tonic-gate } 59717c478bd9Sstevel@tonic-gate 59727c478bd9Sstevel@tonic-gate cur_len = MIN(io_len - done_len, contig); 59737c478bd9Sstevel@tonic-gate /* 59747c478bd9Sstevel@tonic-gate * Zero out a page beyond EOF, when the last block of 59757c478bd9Sstevel@tonic-gate * a file is a UFS fragment so that ufs_pageio() can be used 59767c478bd9Sstevel@tonic-gate * instead of ufs_getpage() to handle faults against 59777c478bd9Sstevel@tonic-gate * segvn segments that use large pages. 59787c478bd9Sstevel@tonic-gate */ 59797c478bd9Sstevel@tonic-gate page_list_break(&cpp, &npp, btopr(cur_len)); 59807c478bd9Sstevel@tonic-gate if ((flags & B_READ) && (cur_len & PAGEOFFSET)) { 59817c478bd9Sstevel@tonic-gate size_t xlen = cur_len & PAGEOFFSET; 59827c478bd9Sstevel@tonic-gate pagezero(cpp->p_prev, xlen, PAGESIZE - xlen); 59837c478bd9Sstevel@tonic-gate } 59847c478bd9Sstevel@tonic-gate 59857c478bd9Sstevel@tonic-gate bp = pageio_setup(cpp, cur_len, ip->i_devvp, flags); 59867c478bd9Sstevel@tonic-gate ASSERT(bp != NULL); 59877c478bd9Sstevel@tonic-gate 59887c478bd9Sstevel@tonic-gate bp->b_edev = ip->i_dev; 59897c478bd9Sstevel@tonic-gate bp->b_dev = cmpdev(ip->i_dev); 59907c478bd9Sstevel@tonic-gate bp->b_blkno = bn; 59917c478bd9Sstevel@tonic-gate bp->b_un.b_addr = (caddr_t)0; 59927c478bd9Sstevel@tonic-gate bp->b_file = ip->i_vnode; 59937c478bd9Sstevel@tonic-gate 59947c478bd9Sstevel@tonic-gate ufsvfsp->vfs_iotstamp = lbolt; 59957c478bd9Sstevel@tonic-gate ub.ub_pageios.value.ul++; 59967c478bd9Sstevel@tonic-gate if (ufsvfsp->vfs_snapshot) 59977c478bd9Sstevel@tonic-gate fssnap_strategy(&(ufsvfsp->vfs_snapshot), bp); 59987c478bd9Sstevel@tonic-gate else 59997c478bd9Sstevel@tonic-gate (void) bdev_strategy(bp); 60007c478bd9Sstevel@tonic-gate 60017c478bd9Sstevel@tonic-gate if (flags & B_READ) 60027c478bd9Sstevel@tonic-gate ufs_pageio_reads++; 60037c478bd9Sstevel@tonic-gate else 60047c478bd9Sstevel@tonic-gate ufs_pageio_writes++; 60057c478bd9Sstevel@tonic-gate if (flags & B_READ) 60067c478bd9Sstevel@tonic-gate lwp_stat_update(LWP_STAT_INBLK, 1); 60077c478bd9Sstevel@tonic-gate else 60087c478bd9Sstevel@tonic-gate lwp_stat_update(LWP_STAT_OUBLK, 1); 60097c478bd9Sstevel@tonic-gate /* 60107c478bd9Sstevel@tonic-gate * If the request is not B_ASYNC, wait for i/o to complete 60117c478bd9Sstevel@tonic-gate * and re-assemble the page list to return to the caller. 60127c478bd9Sstevel@tonic-gate * If it is B_ASYNC we leave the page list in pieces and 60137c478bd9Sstevel@tonic-gate * cleanup() will dispose of them. 60147c478bd9Sstevel@tonic-gate */ 60157c478bd9Sstevel@tonic-gate if ((flags & B_ASYNC) == 0) { 60167c478bd9Sstevel@tonic-gate err = biowait(bp); 60177c478bd9Sstevel@tonic-gate pageio_done(bp); 60187c478bd9Sstevel@tonic-gate if (err) 60197c478bd9Sstevel@tonic-gate break; 60207c478bd9Sstevel@tonic-gate page_list_concat(&opp, &cpp); 60217c478bd9Sstevel@tonic-gate } 60227c478bd9Sstevel@tonic-gate cpp = npp; 60237c478bd9Sstevel@tonic-gate npp = NULL; 60247c478bd9Sstevel@tonic-gate if (flags & B_READ) 60257c478bd9Sstevel@tonic-gate cur_len = P2ROUNDUP_TYPED(cur_len, PAGESIZE, size_t); 60267c478bd9Sstevel@tonic-gate done_len += cur_len; 60277c478bd9Sstevel@tonic-gate } 60287c478bd9Sstevel@tonic-gate ASSERT(err || (cpp == NULL && npp == NULL && done_len == io_len)); 60297c478bd9Sstevel@tonic-gate if (err) { 60307c478bd9Sstevel@tonic-gate if (flags & B_ASYNC) { 60317c478bd9Sstevel@tonic-gate /* Cleanup unprocessed parts of list */ 60327c478bd9Sstevel@tonic-gate page_list_concat(&cpp, &npp); 60337c478bd9Sstevel@tonic-gate if (flags & B_READ) 60347c478bd9Sstevel@tonic-gate pvn_read_done(cpp, B_ERROR); 60357c478bd9Sstevel@tonic-gate else 60367c478bd9Sstevel@tonic-gate pvn_write_done(cpp, B_ERROR); 60377c478bd9Sstevel@tonic-gate } else { 60387c478bd9Sstevel@tonic-gate /* Re-assemble list and let caller clean up */ 60397c478bd9Sstevel@tonic-gate page_list_concat(&opp, &cpp); 60407c478bd9Sstevel@tonic-gate page_list_concat(&opp, &npp); 60417c478bd9Sstevel@tonic-gate } 60427c478bd9Sstevel@tonic-gate } 6043bc69f433Saguzovsk 6044bc69f433Saguzovsk if (vmpss && !(ip->i_flag & IACC) && !ULOCKFS_IS_NOIACC(ulp) && 6045bc69f433Saguzovsk ufsvfsp->vfs_fs->fs_ronly == 0 && !ufsvfsp->vfs_noatime) { 6046bc69f433Saguzovsk mutex_enter(&ip->i_tlock); 6047bc69f433Saguzovsk ip->i_flag |= IACC; 6048bc69f433Saguzovsk ITIMES_NOLOCK(ip); 6049bc69f433Saguzovsk mutex_exit(&ip->i_tlock); 6050bc69f433Saguzovsk } 6051bc69f433Saguzovsk 60527c478bd9Sstevel@tonic-gate if (dolock) 60537c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 6054bc69f433Saguzovsk if (vmpss && !atomic_add_long_nv(&ulp->ul_vnops_cnt, -1)) 6055bc69f433Saguzovsk cv_broadcast(&ulp->ul_cv); 60567c478bd9Sstevel@tonic-gate return (err); 60577c478bd9Sstevel@tonic-gate } 60587c478bd9Sstevel@tonic-gate 60597c478bd9Sstevel@tonic-gate /* 60607c478bd9Sstevel@tonic-gate * Called when the kernel is in a frozen state to dump data 60617c478bd9Sstevel@tonic-gate * directly to the device. It uses a private dump data structure, 60627c478bd9Sstevel@tonic-gate * set up by dump_ctl, to locate the correct disk block to which to dump. 60637c478bd9Sstevel@tonic-gate */ 60647c478bd9Sstevel@tonic-gate static int 60657c478bd9Sstevel@tonic-gate ufs_dump(vnode_t *vp, caddr_t addr, int ldbn, int dblks) 60667c478bd9Sstevel@tonic-gate { 60677c478bd9Sstevel@tonic-gate u_offset_t file_size; 60687c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 60697c478bd9Sstevel@tonic-gate struct fs *fs = ip->i_fs; 60707c478bd9Sstevel@tonic-gate daddr_t dbn, lfsbn; 60717c478bd9Sstevel@tonic-gate int disk_blks = fs->fs_bsize >> DEV_BSHIFT; 60727c478bd9Sstevel@tonic-gate int error = 0; 60737c478bd9Sstevel@tonic-gate int ndbs, nfsbs; 60747c478bd9Sstevel@tonic-gate 60757c478bd9Sstevel@tonic-gate /* 60767c478bd9Sstevel@tonic-gate * forced unmount case 60777c478bd9Sstevel@tonic-gate */ 60787c478bd9Sstevel@tonic-gate if (ip->i_ufsvfs == NULL) 60797c478bd9Sstevel@tonic-gate return (EIO); 60807c478bd9Sstevel@tonic-gate /* 60817c478bd9Sstevel@tonic-gate * Validate the inode that it has not been modified since 60827c478bd9Sstevel@tonic-gate * the dump structure is allocated. 60837c478bd9Sstevel@tonic-gate */ 60847c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 60857c478bd9Sstevel@tonic-gate if ((dump_info == NULL) || 60867c478bd9Sstevel@tonic-gate (dump_info->ip != ip) || 60877c478bd9Sstevel@tonic-gate (dump_info->time.tv_sec != ip->i_mtime.tv_sec) || 60887c478bd9Sstevel@tonic-gate (dump_info->time.tv_usec != ip->i_mtime.tv_usec)) { 60897c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 60907c478bd9Sstevel@tonic-gate return (-1); 60917c478bd9Sstevel@tonic-gate } 60927c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 60937c478bd9Sstevel@tonic-gate 60947c478bd9Sstevel@tonic-gate /* 60957c478bd9Sstevel@tonic-gate * See that the file has room for this write 60967c478bd9Sstevel@tonic-gate */ 60977c478bd9Sstevel@tonic-gate UFS_GET_ISIZE(&file_size, ip); 60987c478bd9Sstevel@tonic-gate 60997c478bd9Sstevel@tonic-gate if (ldbtob((offset_t)(ldbn + dblks)) > file_size) 61007c478bd9Sstevel@tonic-gate return (ENOSPC); 61017c478bd9Sstevel@tonic-gate 61027c478bd9Sstevel@tonic-gate /* 61037c478bd9Sstevel@tonic-gate * Find the physical disk block numbers from the dump 61047c478bd9Sstevel@tonic-gate * private data structure directly and write out the data 61057c478bd9Sstevel@tonic-gate * in contiguous block lumps 61067c478bd9Sstevel@tonic-gate */ 61077c478bd9Sstevel@tonic-gate while (dblks > 0 && !error) { 61087c478bd9Sstevel@tonic-gate lfsbn = (daddr_t)lblkno(fs, ldbtob((offset_t)ldbn)); 61097c478bd9Sstevel@tonic-gate dbn = fsbtodb(fs, dump_info->dblk[lfsbn]) + ldbn % disk_blks; 61107c478bd9Sstevel@tonic-gate nfsbs = 1; 61117c478bd9Sstevel@tonic-gate ndbs = disk_blks - ldbn % disk_blks; 61127c478bd9Sstevel@tonic-gate while (ndbs < dblks && fsbtodb(fs, dump_info->dblk[lfsbn + 61137c478bd9Sstevel@tonic-gate nfsbs]) == dbn + ndbs) { 61147c478bd9Sstevel@tonic-gate nfsbs++; 61157c478bd9Sstevel@tonic-gate ndbs += disk_blks; 61167c478bd9Sstevel@tonic-gate } 61177c478bd9Sstevel@tonic-gate if (ndbs > dblks) 61187c478bd9Sstevel@tonic-gate ndbs = dblks; 61197c478bd9Sstevel@tonic-gate error = bdev_dump(ip->i_dev, addr, dbn, ndbs); 61207c478bd9Sstevel@tonic-gate addr += ldbtob((offset_t)ndbs); 61217c478bd9Sstevel@tonic-gate dblks -= ndbs; 61227c478bd9Sstevel@tonic-gate ldbn += ndbs; 61237c478bd9Sstevel@tonic-gate } 61247c478bd9Sstevel@tonic-gate return (error); 61257c478bd9Sstevel@tonic-gate 61267c478bd9Sstevel@tonic-gate } 61277c478bd9Sstevel@tonic-gate 61287c478bd9Sstevel@tonic-gate /* 61297c478bd9Sstevel@tonic-gate * Prepare the file system before and after the dump operation. 61307c478bd9Sstevel@tonic-gate * 61317c478bd9Sstevel@tonic-gate * action = DUMP_ALLOC: 61327c478bd9Sstevel@tonic-gate * Preparation before dump, allocate dump private data structure 61337c478bd9Sstevel@tonic-gate * to hold all the direct and indirect block info for dump. 61347c478bd9Sstevel@tonic-gate * 61357c478bd9Sstevel@tonic-gate * action = DUMP_FREE: 61367c478bd9Sstevel@tonic-gate * Clean up after dump, deallocate the dump private data structure. 61377c478bd9Sstevel@tonic-gate * 61387c478bd9Sstevel@tonic-gate * action = DUMP_SCAN: 61397c478bd9Sstevel@tonic-gate * Scan dump_info for *blkp DEV_BSIZE blocks of contig fs space; 61407c478bd9Sstevel@tonic-gate * if found, the starting file-relative DEV_BSIZE lbn is written 61417c478bd9Sstevel@tonic-gate * to *bklp; that lbn is intended for use with VOP_DUMP() 61427c478bd9Sstevel@tonic-gate */ 61437c478bd9Sstevel@tonic-gate static int 61447c478bd9Sstevel@tonic-gate ufs_dumpctl(vnode_t *vp, int action, int *blkp) 61457c478bd9Sstevel@tonic-gate { 61467c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 61477c478bd9Sstevel@tonic-gate ufsvfs_t *ufsvfsp = ip->i_ufsvfs; 61487c478bd9Sstevel@tonic-gate struct fs *fs; 61497c478bd9Sstevel@tonic-gate daddr32_t *dblk, *storeblk; 61507c478bd9Sstevel@tonic-gate daddr32_t *nextblk, *endblk; 61517c478bd9Sstevel@tonic-gate struct buf *bp; 61527c478bd9Sstevel@tonic-gate int i, entry, entries; 61537c478bd9Sstevel@tonic-gate int n, ncontig; 61547c478bd9Sstevel@tonic-gate 61557c478bd9Sstevel@tonic-gate /* 61567c478bd9Sstevel@tonic-gate * check for forced unmount 61577c478bd9Sstevel@tonic-gate */ 61587c478bd9Sstevel@tonic-gate if (ufsvfsp == NULL) 61597c478bd9Sstevel@tonic-gate return (EIO); 61607c478bd9Sstevel@tonic-gate 61617c478bd9Sstevel@tonic-gate if (action == DUMP_ALLOC) { 61627c478bd9Sstevel@tonic-gate /* 61637c478bd9Sstevel@tonic-gate * alloc and record dump_info 61647c478bd9Sstevel@tonic-gate */ 61657c478bd9Sstevel@tonic-gate if (dump_info != NULL) 61667c478bd9Sstevel@tonic-gate return (EINVAL); 61677c478bd9Sstevel@tonic-gate 61687c478bd9Sstevel@tonic-gate ASSERT(vp->v_type == VREG); 61697c478bd9Sstevel@tonic-gate fs = ufsvfsp->vfs_fs; 61707c478bd9Sstevel@tonic-gate 61717c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_READER); 61727c478bd9Sstevel@tonic-gate 61737c478bd9Sstevel@tonic-gate if (bmap_has_holes(ip)) { 61747c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 61757c478bd9Sstevel@tonic-gate return (EFAULT); 61767c478bd9Sstevel@tonic-gate } 61777c478bd9Sstevel@tonic-gate 61787c478bd9Sstevel@tonic-gate /* 61797c478bd9Sstevel@tonic-gate * calculate and allocate space needed according to i_size 61807c478bd9Sstevel@tonic-gate */ 61817c478bd9Sstevel@tonic-gate entries = (int)lblkno(fs, blkroundup(fs, ip->i_size)); 61827c478bd9Sstevel@tonic-gate if ((dump_info = (struct dump *) 61837c478bd9Sstevel@tonic-gate kmem_alloc(sizeof (struct dump) + 61847c478bd9Sstevel@tonic-gate (entries - 1) * sizeof (daddr32_t), KM_NOSLEEP)) == NULL) { 61857c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 61867c478bd9Sstevel@tonic-gate return (ENOMEM); 61877c478bd9Sstevel@tonic-gate } 61887c478bd9Sstevel@tonic-gate 61897c478bd9Sstevel@tonic-gate /* Start saving the info */ 61907c478bd9Sstevel@tonic-gate dump_info->fsbs = entries; 61917c478bd9Sstevel@tonic-gate dump_info->ip = ip; 61927c478bd9Sstevel@tonic-gate storeblk = &dump_info->dblk[0]; 61937c478bd9Sstevel@tonic-gate 61947c478bd9Sstevel@tonic-gate /* Direct Blocks */ 61957c478bd9Sstevel@tonic-gate for (entry = 0; entry < NDADDR && entry < entries; entry++) 61967c478bd9Sstevel@tonic-gate *storeblk++ = ip->i_db[entry]; 61977c478bd9Sstevel@tonic-gate 61987c478bd9Sstevel@tonic-gate /* Indirect Blocks */ 61997c478bd9Sstevel@tonic-gate for (i = 0; i < NIADDR; i++) { 62007c478bd9Sstevel@tonic-gate int error = 0; 62017c478bd9Sstevel@tonic-gate 62027c478bd9Sstevel@tonic-gate bp = UFS_BREAD(ufsvfsp, 62037c478bd9Sstevel@tonic-gate ip->i_dev, fsbtodb(fs, ip->i_ib[i]), 62047c478bd9Sstevel@tonic-gate fs->fs_bsize); 62057c478bd9Sstevel@tonic-gate if (bp->b_flags & B_ERROR) 62067c478bd9Sstevel@tonic-gate error = EIO; 62077c478bd9Sstevel@tonic-gate else { 62087c478bd9Sstevel@tonic-gate dblk = bp->b_un.b_daddr; 62097c478bd9Sstevel@tonic-gate if ((storeblk = save_dblks(ip, ufsvfsp, 62107c478bd9Sstevel@tonic-gate storeblk, dblk, i, entries)) == NULL) 62117c478bd9Sstevel@tonic-gate error = EIO; 62127c478bd9Sstevel@tonic-gate } 62137c478bd9Sstevel@tonic-gate 62147c478bd9Sstevel@tonic-gate brelse(bp); 62157c478bd9Sstevel@tonic-gate 62167c478bd9Sstevel@tonic-gate if (error != 0) { 62177c478bd9Sstevel@tonic-gate kmem_free(dump_info, sizeof (struct dump) + 62187c478bd9Sstevel@tonic-gate (entries - 1) * sizeof (daddr32_t)); 62197c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 62207c478bd9Sstevel@tonic-gate dump_info = NULL; 62217c478bd9Sstevel@tonic-gate return (error); 62227c478bd9Sstevel@tonic-gate } 62237c478bd9Sstevel@tonic-gate } 62247c478bd9Sstevel@tonic-gate /* and time stamp the information */ 62257c478bd9Sstevel@tonic-gate mutex_enter(&ip->i_tlock); 62267c478bd9Sstevel@tonic-gate dump_info->time = ip->i_mtime; 62277c478bd9Sstevel@tonic-gate mutex_exit(&ip->i_tlock); 62287c478bd9Sstevel@tonic-gate 62297c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 62307c478bd9Sstevel@tonic-gate } else if (action == DUMP_FREE) { 62317c478bd9Sstevel@tonic-gate /* 62327c478bd9Sstevel@tonic-gate * free dump_info 62337c478bd9Sstevel@tonic-gate */ 62347c478bd9Sstevel@tonic-gate if (dump_info == NULL) 62357c478bd9Sstevel@tonic-gate return (EINVAL); 62367c478bd9Sstevel@tonic-gate entries = dump_info->fsbs - 1; 62377c478bd9Sstevel@tonic-gate kmem_free(dump_info, sizeof (struct dump) + 62387c478bd9Sstevel@tonic-gate entries * sizeof (daddr32_t)); 62397c478bd9Sstevel@tonic-gate dump_info = NULL; 62407c478bd9Sstevel@tonic-gate } else if (action == DUMP_SCAN) { 62417c478bd9Sstevel@tonic-gate /* 62427c478bd9Sstevel@tonic-gate * scan dump_info 62437c478bd9Sstevel@tonic-gate */ 62447c478bd9Sstevel@tonic-gate if (dump_info == NULL) 62457c478bd9Sstevel@tonic-gate return (EINVAL); 62467c478bd9Sstevel@tonic-gate 62477c478bd9Sstevel@tonic-gate dblk = dump_info->dblk; 62487c478bd9Sstevel@tonic-gate nextblk = dblk + 1; 62497c478bd9Sstevel@tonic-gate endblk = dblk + dump_info->fsbs - 1; 62507c478bd9Sstevel@tonic-gate fs = ufsvfsp->vfs_fs; 62517c478bd9Sstevel@tonic-gate ncontig = *blkp >> (fs->fs_bshift - DEV_BSHIFT); 62527c478bd9Sstevel@tonic-gate 62537c478bd9Sstevel@tonic-gate /* 62547c478bd9Sstevel@tonic-gate * scan dblk[] entries; contig fs space is found when: 62557c478bd9Sstevel@tonic-gate * ((current blkno + frags per block) == next blkno) 62567c478bd9Sstevel@tonic-gate */ 62577c478bd9Sstevel@tonic-gate n = 0; 62587c478bd9Sstevel@tonic-gate while (n < ncontig && dblk < endblk) { 62597c478bd9Sstevel@tonic-gate if ((*dblk + fs->fs_frag) == *nextblk) 62607c478bd9Sstevel@tonic-gate n++; 62617c478bd9Sstevel@tonic-gate else 62627c478bd9Sstevel@tonic-gate n = 0; 62637c478bd9Sstevel@tonic-gate dblk++; 62647c478bd9Sstevel@tonic-gate nextblk++; 62657c478bd9Sstevel@tonic-gate } 62667c478bd9Sstevel@tonic-gate 62677c478bd9Sstevel@tonic-gate /* 62687c478bd9Sstevel@tonic-gate * index is where size bytes of contig space begins; 62697c478bd9Sstevel@tonic-gate * conversion from index to the file's DEV_BSIZE lbn 62707c478bd9Sstevel@tonic-gate * is equivalent to: (index * fs_bsize) / DEV_BSIZE 62717c478bd9Sstevel@tonic-gate */ 62727c478bd9Sstevel@tonic-gate if (n == ncontig) { 62737c478bd9Sstevel@tonic-gate i = (dblk - dump_info->dblk) - ncontig; 62747c478bd9Sstevel@tonic-gate *blkp = i << (fs->fs_bshift - DEV_BSHIFT); 62757c478bd9Sstevel@tonic-gate } else 62767c478bd9Sstevel@tonic-gate return (EFAULT); 62777c478bd9Sstevel@tonic-gate } 62787c478bd9Sstevel@tonic-gate return (0); 62797c478bd9Sstevel@tonic-gate } 62807c478bd9Sstevel@tonic-gate 62817c478bd9Sstevel@tonic-gate /* 62827c478bd9Sstevel@tonic-gate * Recursive helper function for ufs_dumpctl(). It follows the indirect file 62837c478bd9Sstevel@tonic-gate * system blocks until it reaches the the disk block addresses, which are 62847c478bd9Sstevel@tonic-gate * then stored into the given buffer, storeblk. 62857c478bd9Sstevel@tonic-gate */ 62867c478bd9Sstevel@tonic-gate static daddr32_t * 62877c478bd9Sstevel@tonic-gate save_dblks(struct inode *ip, struct ufsvfs *ufsvfsp, daddr32_t *storeblk, 62887c478bd9Sstevel@tonic-gate daddr32_t *dblk, int level, int entries) 62897c478bd9Sstevel@tonic-gate { 62907c478bd9Sstevel@tonic-gate struct fs *fs = ufsvfsp->vfs_fs; 62917c478bd9Sstevel@tonic-gate struct buf *bp; 62927c478bd9Sstevel@tonic-gate int i; 62937c478bd9Sstevel@tonic-gate 62947c478bd9Sstevel@tonic-gate if (level == 0) { 62957c478bd9Sstevel@tonic-gate for (i = 0; i < NINDIR(fs); i++) { 62967c478bd9Sstevel@tonic-gate if (storeblk - dump_info->dblk >= entries) 62977c478bd9Sstevel@tonic-gate break; 62987c478bd9Sstevel@tonic-gate *storeblk++ = dblk[i]; 62997c478bd9Sstevel@tonic-gate } 63007c478bd9Sstevel@tonic-gate return (storeblk); 63017c478bd9Sstevel@tonic-gate } 63027c478bd9Sstevel@tonic-gate for (i = 0; i < NINDIR(fs); i++) { 63037c478bd9Sstevel@tonic-gate if (storeblk - dump_info->dblk >= entries) 63047c478bd9Sstevel@tonic-gate break; 63057c478bd9Sstevel@tonic-gate bp = UFS_BREAD(ufsvfsp, 63067c478bd9Sstevel@tonic-gate ip->i_dev, fsbtodb(fs, dblk[i]), fs->fs_bsize); 63077c478bd9Sstevel@tonic-gate if (bp->b_flags & B_ERROR) { 63087c478bd9Sstevel@tonic-gate brelse(bp); 63097c478bd9Sstevel@tonic-gate return (NULL); 63107c478bd9Sstevel@tonic-gate } 63117c478bd9Sstevel@tonic-gate storeblk = save_dblks(ip, ufsvfsp, storeblk, bp->b_un.b_daddr, 63127c478bd9Sstevel@tonic-gate level - 1, entries); 63137c478bd9Sstevel@tonic-gate brelse(bp); 63147c478bd9Sstevel@tonic-gate 63157c478bd9Sstevel@tonic-gate if (storeblk == NULL) 63167c478bd9Sstevel@tonic-gate return (NULL); 63177c478bd9Sstevel@tonic-gate } 63187c478bd9Sstevel@tonic-gate return (storeblk); 63197c478bd9Sstevel@tonic-gate } 63207c478bd9Sstevel@tonic-gate 63217c478bd9Sstevel@tonic-gate /* ARGSUSED */ 63227c478bd9Sstevel@tonic-gate static int 63237c478bd9Sstevel@tonic-gate ufs_getsecattr(struct vnode *vp, vsecattr_t *vsap, int flag, 63247c478bd9Sstevel@tonic-gate struct cred *cr) 63257c478bd9Sstevel@tonic-gate { 63267c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 63277c478bd9Sstevel@tonic-gate struct ulockfs *ulp; 63287c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp = ip->i_ufsvfs; 63297c478bd9Sstevel@tonic-gate ulong_t vsa_mask = vsap->vsa_mask; 63307c478bd9Sstevel@tonic-gate int err = EINVAL; 63317c478bd9Sstevel@tonic-gate 63327c478bd9Sstevel@tonic-gate TRACE_3(TR_FAC_UFS, TR_UFS_GETSECATTR_START, 63337c478bd9Sstevel@tonic-gate "ufs_getsecattr_start:vp %p, vsap %p, flags %x", vp, vsap, flag); 63347c478bd9Sstevel@tonic-gate 63357c478bd9Sstevel@tonic-gate vsa_mask &= (VSA_ACL | VSA_ACLCNT | VSA_DFACL | VSA_DFACLCNT); 63367c478bd9Sstevel@tonic-gate 63377c478bd9Sstevel@tonic-gate /* 63387c478bd9Sstevel@tonic-gate * Only grab locks if needed - they're not needed to check vsa_mask 63397c478bd9Sstevel@tonic-gate * or if the mask contains no acl flags. 63407c478bd9Sstevel@tonic-gate */ 63417c478bd9Sstevel@tonic-gate if (vsa_mask != 0) { 63427c478bd9Sstevel@tonic-gate if (err = ufs_lockfs_begin(ufsvfsp, &ulp, 63437c478bd9Sstevel@tonic-gate ULOCKFS_GETATTR_MASK)) 63447c478bd9Sstevel@tonic-gate return (err); 63457c478bd9Sstevel@tonic-gate 63467c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_READER); 63477c478bd9Sstevel@tonic-gate err = ufs_acl_get(ip, vsap, flag, cr); 63487c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 63497c478bd9Sstevel@tonic-gate 63507c478bd9Sstevel@tonic-gate if (ulp) 63517c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 63527c478bd9Sstevel@tonic-gate } 63537c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_GETSECATTR_END, 63547c478bd9Sstevel@tonic-gate "ufs_getsecattr_end:vp %p", vp); 63557c478bd9Sstevel@tonic-gate return (err); 63567c478bd9Sstevel@tonic-gate } 63577c478bd9Sstevel@tonic-gate 63587c478bd9Sstevel@tonic-gate /* ARGSUSED */ 63597c478bd9Sstevel@tonic-gate static int 63607c478bd9Sstevel@tonic-gate ufs_setsecattr(struct vnode *vp, vsecattr_t *vsap, int flag, struct cred *cr) 63617c478bd9Sstevel@tonic-gate { 63627c478bd9Sstevel@tonic-gate struct inode *ip = VTOI(vp); 63637c478bd9Sstevel@tonic-gate struct ulockfs *ulp = NULL; 63647c478bd9Sstevel@tonic-gate struct ufsvfs *ufsvfsp = VTOI(vp)->i_ufsvfs; 63657c478bd9Sstevel@tonic-gate ulong_t vsa_mask = vsap->vsa_mask; 63667c478bd9Sstevel@tonic-gate int err; 63677c478bd9Sstevel@tonic-gate int haverwlock = 1; 63687c478bd9Sstevel@tonic-gate int trans_size; 63697c478bd9Sstevel@tonic-gate int donetrans = 0; 63707c478bd9Sstevel@tonic-gate int retry = 1; 63717c478bd9Sstevel@tonic-gate 63727c478bd9Sstevel@tonic-gate 63737c478bd9Sstevel@tonic-gate TRACE_3(TR_FAC_UFS, TR_UFS_SETSECATTR_START, 63747c478bd9Sstevel@tonic-gate "ufs_setsecattr_start:vp %p, vsap %p, flags %x", vp, vsap, flag); 63757c478bd9Sstevel@tonic-gate 63767c478bd9Sstevel@tonic-gate ASSERT(RW_LOCK_HELD(&ip->i_rwlock)); 63777c478bd9Sstevel@tonic-gate 63787c478bd9Sstevel@tonic-gate /* Abort now if the request is either empty or invalid. */ 63797c478bd9Sstevel@tonic-gate vsa_mask &= (VSA_ACL | VSA_ACLCNT | VSA_DFACL | VSA_DFACLCNT); 63807c478bd9Sstevel@tonic-gate if ((vsa_mask == 0) || 63817c478bd9Sstevel@tonic-gate ((vsap->vsa_aclentp == NULL) && 63827c478bd9Sstevel@tonic-gate (vsap->vsa_dfaclentp == NULL))) { 63837c478bd9Sstevel@tonic-gate err = EINVAL; 63847c478bd9Sstevel@tonic-gate goto out; 63857c478bd9Sstevel@tonic-gate } 63867c478bd9Sstevel@tonic-gate 63877c478bd9Sstevel@tonic-gate /* 63887c478bd9Sstevel@tonic-gate * Following convention, if this is a directory then we acquire the 63897c478bd9Sstevel@tonic-gate * inode's i_rwlock after starting a UFS logging transaction; 63907c478bd9Sstevel@tonic-gate * otherwise, we acquire it beforehand. Since we were called (and 63917c478bd9Sstevel@tonic-gate * must therefore return) with the lock held, we will have to drop it, 63927c478bd9Sstevel@tonic-gate * and later reacquire it, if operating on a directory. 63937c478bd9Sstevel@tonic-gate */ 63947c478bd9Sstevel@tonic-gate if (vp->v_type == VDIR) { 63957c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 63967c478bd9Sstevel@tonic-gate haverwlock = 0; 63977c478bd9Sstevel@tonic-gate } else { 63987c478bd9Sstevel@tonic-gate /* Upgrade the lock if required. */ 63997c478bd9Sstevel@tonic-gate if (!rw_write_held(&ip->i_rwlock)) { 64007c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 64017c478bd9Sstevel@tonic-gate rw_enter(&ip->i_rwlock, RW_WRITER); 64027c478bd9Sstevel@tonic-gate } 64037c478bd9Sstevel@tonic-gate } 64047c478bd9Sstevel@tonic-gate 64057c478bd9Sstevel@tonic-gate again: 64067c478bd9Sstevel@tonic-gate ASSERT(!(vp->v_type == VDIR && haverwlock)); 64077c478bd9Sstevel@tonic-gate if (err = ufs_lockfs_begin(ufsvfsp, &ulp, ULOCKFS_SETATTR_MASK)) { 64087c478bd9Sstevel@tonic-gate ulp = NULL; 64097c478bd9Sstevel@tonic-gate retry = 0; 64107c478bd9Sstevel@tonic-gate goto out; 64117c478bd9Sstevel@tonic-gate } 64127c478bd9Sstevel@tonic-gate 64137c478bd9Sstevel@tonic-gate /* 64147c478bd9Sstevel@tonic-gate * Check that the file system supports this operation. Note that 64157c478bd9Sstevel@tonic-gate * ufs_lockfs_begin() will have checked that the file system had 64167c478bd9Sstevel@tonic-gate * not been forcibly unmounted. 64177c478bd9Sstevel@tonic-gate */ 64187c478bd9Sstevel@tonic-gate if (ufsvfsp->vfs_fs->fs_ronly) { 64197c478bd9Sstevel@tonic-gate err = EROFS; 64207c478bd9Sstevel@tonic-gate goto out; 64217c478bd9Sstevel@tonic-gate } 64227c478bd9Sstevel@tonic-gate if (ufsvfsp->vfs_nosetsec) { 64237c478bd9Sstevel@tonic-gate err = ENOSYS; 64247c478bd9Sstevel@tonic-gate goto out; 64257c478bd9Sstevel@tonic-gate } 64267c478bd9Sstevel@tonic-gate 64277c478bd9Sstevel@tonic-gate if (ulp) { 64287c478bd9Sstevel@tonic-gate TRANS_BEGIN_ASYNC(ufsvfsp, TOP_SETSECATTR, 64297c478bd9Sstevel@tonic-gate trans_size = TOP_SETSECATTR_SIZE(VTOI(vp))); 64307c478bd9Sstevel@tonic-gate donetrans = 1; 64317c478bd9Sstevel@tonic-gate } 64327c478bd9Sstevel@tonic-gate 64337c478bd9Sstevel@tonic-gate if (vp->v_type == VDIR) { 64347c478bd9Sstevel@tonic-gate rw_enter(&ip->i_rwlock, RW_WRITER); 64357c478bd9Sstevel@tonic-gate haverwlock = 1; 64367c478bd9Sstevel@tonic-gate } 64377c478bd9Sstevel@tonic-gate 64387c478bd9Sstevel@tonic-gate ASSERT(haverwlock); 64397c478bd9Sstevel@tonic-gate 64407c478bd9Sstevel@tonic-gate /* Do the actual work. */ 64417c478bd9Sstevel@tonic-gate rw_enter(&ip->i_contents, RW_WRITER); 64427c478bd9Sstevel@tonic-gate /* 64437c478bd9Sstevel@tonic-gate * Suppress out of inodes messages if we will retry. 64447c478bd9Sstevel@tonic-gate */ 64457c478bd9Sstevel@tonic-gate if (retry) 64467c478bd9Sstevel@tonic-gate ip->i_flag |= IQUIET; 64477c478bd9Sstevel@tonic-gate err = ufs_acl_set(ip, vsap, flag, cr); 64487c478bd9Sstevel@tonic-gate ip->i_flag &= ~IQUIET; 64497c478bd9Sstevel@tonic-gate rw_exit(&ip->i_contents); 64507c478bd9Sstevel@tonic-gate 64517c478bd9Sstevel@tonic-gate out: 64527c478bd9Sstevel@tonic-gate if (ulp) { 64537c478bd9Sstevel@tonic-gate if (donetrans) { 64547c478bd9Sstevel@tonic-gate /* 64557c478bd9Sstevel@tonic-gate * top_end_async() can eventually call 64567c478bd9Sstevel@tonic-gate * top_end_sync(), which can block. We must 64577c478bd9Sstevel@tonic-gate * therefore observe the lock-ordering protocol 64587c478bd9Sstevel@tonic-gate * here as well. 64597c478bd9Sstevel@tonic-gate */ 64607c478bd9Sstevel@tonic-gate if (vp->v_type == VDIR) { 64617c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 64627c478bd9Sstevel@tonic-gate haverwlock = 0; 64637c478bd9Sstevel@tonic-gate } 64647c478bd9Sstevel@tonic-gate TRANS_END_ASYNC(ufsvfsp, TOP_SETSECATTR, trans_size); 64657c478bd9Sstevel@tonic-gate } 64667c478bd9Sstevel@tonic-gate ufs_lockfs_end(ulp); 64677c478bd9Sstevel@tonic-gate } 64687c478bd9Sstevel@tonic-gate /* 64697c478bd9Sstevel@tonic-gate * If no inodes available, try scaring a logically- 64707c478bd9Sstevel@tonic-gate * free one out of the delete queue to someplace 64717c478bd9Sstevel@tonic-gate * that we can find it. 64727c478bd9Sstevel@tonic-gate */ 64737c478bd9Sstevel@tonic-gate if ((err == ENOSPC) && retry && TRANS_ISTRANS(ufsvfsp)) { 64747c478bd9Sstevel@tonic-gate ufs_delete_drain_wait(ufsvfsp, 1); 64757c478bd9Sstevel@tonic-gate retry = 0; 64767c478bd9Sstevel@tonic-gate if (vp->v_type == VDIR && haverwlock) { 64777c478bd9Sstevel@tonic-gate rw_exit(&ip->i_rwlock); 64787c478bd9Sstevel@tonic-gate haverwlock = 0; 64797c478bd9Sstevel@tonic-gate } 64807c478bd9Sstevel@tonic-gate goto again; 64817c478bd9Sstevel@tonic-gate } 64827c478bd9Sstevel@tonic-gate /* 64837c478bd9Sstevel@tonic-gate * If we need to reacquire the lock then it is safe to do so 64847c478bd9Sstevel@tonic-gate * as a reader. This is because ufs_rwunlock(), which will be 64857c478bd9Sstevel@tonic-gate * called by our caller after we return, does not differentiate 64867c478bd9Sstevel@tonic-gate * between shared and exclusive locks. 64877c478bd9Sstevel@tonic-gate */ 64887c478bd9Sstevel@tonic-gate if (!haverwlock) { 64897c478bd9Sstevel@tonic-gate ASSERT(vp->v_type == VDIR); 64907c478bd9Sstevel@tonic-gate rw_enter(&ip->i_rwlock, RW_READER); 64917c478bd9Sstevel@tonic-gate } 64927c478bd9Sstevel@tonic-gate 64937c478bd9Sstevel@tonic-gate TRACE_1(TR_FAC_UFS, TR_UFS_SETSECATTR_END, 64947c478bd9Sstevel@tonic-gate "ufs_setsecattr_end:vp %p", vp); 64957c478bd9Sstevel@tonic-gate return (err); 64967c478bd9Sstevel@tonic-gate } 6497