vfs_vnops.c revision 1.45
1/* $NetBSD: vfs_vnops.c,v 1.45 2000/11/27 08:39:44 chs Exp $ */ 2 3/* 4 * Copyright (c) 1982, 1986, 1989, 1993 5 * The Regents of the University of California. All rights reserved. 6 * (c) UNIX System Laboratories, Inc. 7 * All or some portions of this file are derived from material licensed 8 * to the University of California by American Telephone and Telegraph 9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 10 * the permission of UNIX System Laboratories, Inc. 11 * 12 * Redistribution and use in source and binary forms, with or without 13 * modification, are permitted provided that the following conditions 14 * are met: 15 * 1. Redistributions of source code must retain the above copyright 16 * notice, this list of conditions and the following disclaimer. 17 * 2. Redistributions in binary form must reproduce the above copyright 18 * notice, this list of conditions and the following disclaimer in the 19 * documentation and/or other materials provided with the distribution. 20 * 3. All advertising materials mentioning features or use of this software 21 * must display the following acknowledgement: 22 * This product includes software developed by the University of 23 * California, Berkeley and its contributors. 24 * 4. Neither the name of the University nor the names of its contributors 25 * may be used to endorse or promote products derived from this software 26 * without specific prior written permission. 27 * 28 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 29 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 30 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 31 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 32 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 33 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 34 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 35 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 36 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 37 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 38 * SUCH DAMAGE. 39 * 40 * @(#)vfs_vnops.c 8.14 (Berkeley) 6/15/95 41 */ 42 43#include "fs_union.h" 44 45#include <sys/param.h> 46#include <sys/systm.h> 47#include <sys/kernel.h> 48#include <sys/file.h> 49#include <sys/stat.h> 50#include <sys/buf.h> 51#include <sys/proc.h> 52#include <sys/mount.h> 53#include <sys/namei.h> 54#include <sys/vnode.h> 55#include <sys/ioctl.h> 56#include <sys/tty.h> 57#include <sys/poll.h> 58 59#include <uvm/uvm_extern.h> 60 61#ifdef UNION 62#include <miscfs/union/union.h> 63#endif 64 65struct fileops vnops = 66 { vn_read, vn_write, vn_ioctl, vn_fcntl, vn_poll, vn_closefile }; 67 68/* 69 * Common code for vnode open operations. 70 * Check permissions, and call the VOP_OPEN or VOP_CREATE routine. 71 */ 72int 73vn_open(ndp, fmode, cmode) 74 struct nameidata *ndp; 75 int fmode, cmode; 76{ 77 struct vnode *vp; 78 struct proc *p = ndp->ni_cnd.cn_proc; 79 struct ucred *cred = p->p_ucred; 80 struct vattr va; 81 int error; 82 83 if (fmode & O_CREAT) { 84 ndp->ni_cnd.cn_nameiop = CREATE; 85 ndp->ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 86 if ((fmode & O_EXCL) == 0 && 87 ((fmode & FNOSYMLINK) == 0)) 88 ndp->ni_cnd.cn_flags |= FOLLOW; 89 if ((error = namei(ndp)) != 0) 90 return (error); 91 if (ndp->ni_vp == NULL) { 92 VATTR_NULL(&va); 93 va.va_type = VREG; 94 va.va_mode = cmode; 95 if (fmode & O_EXCL) 96 va.va_vaflags |= VA_EXCLUSIVE; 97 VOP_LEASE(ndp->ni_dvp, p, cred, LEASE_WRITE); 98 error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp, 99 &ndp->ni_cnd, &va); 100 if (error) 101 return (error); 102 fmode &= ~O_TRUNC; 103 vp = ndp->ni_vp; 104 } else { 105 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 106 if (ndp->ni_dvp == ndp->ni_vp) 107 vrele(ndp->ni_dvp); 108 else 109 vput(ndp->ni_dvp); 110 ndp->ni_dvp = NULL; 111 vp = ndp->ni_vp; 112 if (fmode & O_EXCL) { 113 error = EEXIST; 114 goto bad; 115 } 116 if (ndp->ni_vp->v_type == VLNK) { 117 error = EFTYPE; 118 goto bad; 119 } 120 fmode &= ~O_CREAT; 121 } 122 } else { 123 ndp->ni_cnd.cn_nameiop = LOOKUP; 124 ndp->ni_cnd.cn_flags = FOLLOW | LOCKLEAF; 125 if ((error = namei(ndp)) != 0) 126 return (error); 127 vp = ndp->ni_vp; 128 } 129 if (vp->v_type == VSOCK) { 130 error = EOPNOTSUPP; 131 goto bad; 132 } 133 if ((fmode & O_CREAT) == 0) { 134 if (fmode & FREAD) { 135 if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0) 136 goto bad; 137 } 138 if (fmode & (FWRITE | O_TRUNC)) { 139 if (vp->v_type == VDIR) { 140 error = EISDIR; 141 goto bad; 142 } 143 if ((error = vn_writechk(vp)) != 0 || 144 (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0) 145 goto bad; 146 } 147 } 148 if (fmode & O_TRUNC) { 149 VOP_UNLOCK(vp, 0); /* XXX */ 150 VOP_LEASE(vp, p, cred, LEASE_WRITE); 151 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 152 VATTR_NULL(&va); 153 va.va_size = 0; 154 if ((error = VOP_SETATTR(vp, &va, cred, p)) != 0) 155 goto bad; 156 } 157 if ((error = VOP_OPEN(vp, fmode, cred, p)) != 0) 158 goto bad; 159 if (vp->v_type == VREG && 160 uvn_attach(vp, fmode & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 161 error = EIO; 162 goto bad; 163 } 164 if (fmode & FWRITE) 165 vp->v_writecount++; 166 167 return (0); 168bad: 169 vput(vp); 170 return (error); 171} 172 173/* 174 * Check for write permissions on the specified vnode. 175 * Prototype text segments cannot be written. 176 */ 177int 178vn_writechk(vp) 179 struct vnode *vp; 180{ 181 182 /* 183 * If the vnode is in use as a process's text, 184 * we can't allow writing. 185 */ 186 if (vp->v_flag & VTEXT) 187 return (ETXTBSY); 188 return (0); 189} 190 191/* 192 * Mark a vnode as being the text image of a running process. 193 */ 194void 195vn_marktext(vp) 196 struct vnode *vp; 197{ 198 vp->v_flag |= VTEXT; 199} 200 201/* 202 * Vnode close call 203 * 204 * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node. 205 */ 206int 207vn_close(vp, flags, cred, p) 208 struct vnode *vp; 209 int flags; 210 struct ucred *cred; 211 struct proc *p; 212{ 213 int error; 214 215 if (flags & FWRITE) 216 vp->v_writecount--; 217 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 218 error = VOP_CLOSE(vp, flags, cred, p); 219 vput(vp); 220 return (error); 221} 222 223/* 224 * Package up an I/O request on a vnode into a uio and do it. 225 */ 226int 227vn_rdwr(rw, vp, base, len, offset, segflg, ioflg, cred, aresid, p) 228 enum uio_rw rw; 229 struct vnode *vp; 230 caddr_t base; 231 int len; 232 off_t offset; 233 enum uio_seg segflg; 234 int ioflg; 235 struct ucred *cred; 236 size_t *aresid; 237 struct proc *p; 238{ 239 struct uio auio; 240 struct iovec aiov; 241 int error; 242 243 if ((ioflg & IO_NODELOCKED) == 0) 244 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 245 auio.uio_iov = &aiov; 246 auio.uio_iovcnt = 1; 247 aiov.iov_base = base; 248 aiov.iov_len = len; 249 auio.uio_resid = len; 250 auio.uio_offset = offset; 251 auio.uio_segflg = segflg; 252 auio.uio_rw = rw; 253 auio.uio_procp = p; 254 if (rw == UIO_READ) { 255 error = VOP_READ(vp, &auio, ioflg, cred); 256 } else { 257 error = VOP_WRITE(vp, &auio, ioflg, cred); 258 } 259 if (aresid) 260 *aresid = auio.uio_resid; 261 else 262 if (auio.uio_resid && error == 0) 263 error = EIO; 264 if ((ioflg & IO_NODELOCKED) == 0) 265 VOP_UNLOCK(vp, 0); 266 return (error); 267} 268 269int 270vn_readdir(fp, buf, segflg, count, done, p, cookies, ncookies) 271 struct file *fp; 272 char *buf; 273 int segflg, *done, *ncookies; 274 u_int count; 275 struct proc *p; 276 off_t **cookies; 277{ 278 struct vnode *vp = (struct vnode *)fp->f_data; 279 struct iovec aiov; 280 struct uio auio; 281 int error, eofflag; 282 283unionread: 284 if (vp->v_type != VDIR) 285 return (EINVAL); 286 aiov.iov_base = buf; 287 aiov.iov_len = count; 288 auio.uio_iov = &aiov; 289 auio.uio_iovcnt = 1; 290 auio.uio_rw = UIO_READ; 291 auio.uio_segflg = segflg; 292 auio.uio_procp = p; 293 auio.uio_resid = count; 294 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 295 auio.uio_offset = fp->f_offset; 296 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies, 297 ncookies); 298 fp->f_offset = auio.uio_offset; 299 VOP_UNLOCK(vp, 0); 300 if (error) 301 return (error); 302 303#ifdef UNION 304{ 305 extern struct vnode *union_dircache __P((struct vnode *)); 306 307 if (count == auio.uio_resid && (vp->v_op == union_vnodeop_p)) { 308 struct vnode *lvp; 309 310 lvp = union_dircache(vp); 311 if (lvp != NULLVP) { 312 struct vattr va; 313 314 /* 315 * If the directory is opaque, 316 * then don't show lower entries 317 */ 318 error = VOP_GETATTR(vp, &va, fp->f_cred, p); 319 if (va.va_flags & OPAQUE) { 320 vput(lvp); 321 lvp = NULL; 322 } 323 } 324 325 if (lvp != NULLVP) { 326 error = VOP_OPEN(lvp, FREAD, fp->f_cred, p); 327 if (error) { 328 vput(lvp); 329 return (error); 330 } 331 VOP_UNLOCK(lvp, 0); 332 fp->f_data = (caddr_t) lvp; 333 fp->f_offset = 0; 334 error = vn_close(vp, FREAD, fp->f_cred, p); 335 if (error) 336 return (error); 337 vp = lvp; 338 goto unionread; 339 } 340 } 341} 342#endif /* UNION */ 343 344 if (count == auio.uio_resid && (vp->v_flag & VROOT) && 345 (vp->v_mount->mnt_flag & MNT_UNION)) { 346 struct vnode *tvp = vp; 347 vp = vp->v_mount->mnt_vnodecovered; 348 VREF(vp); 349 fp->f_data = (caddr_t) vp; 350 fp->f_offset = 0; 351 vrele(tvp); 352 goto unionread; 353 } 354 *done = count - auio.uio_resid; 355 return error; 356} 357 358/* 359 * File table vnode read routine. 360 */ 361int 362vn_read(fp, offset, uio, cred, flags) 363 struct file *fp; 364 off_t *offset; 365 struct uio *uio; 366 struct ucred *cred; 367 int flags; 368{ 369 struct vnode *vp = (struct vnode *)fp->f_data; 370 int count, error, ioflag = 0; 371 372 VOP_LEASE(vp, uio->uio_procp, cred, LEASE_READ); 373 if (fp->f_flag & FNONBLOCK) 374 ioflag |= IO_NDELAY; 375 if ((fp->f_flag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC)) 376 ioflag |= IO_SYNC; 377 if (fp->f_flag & FALTIO) 378 ioflag |= IO_ALTSEMANTICS; 379 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 380 uio->uio_offset = *offset; 381 count = uio->uio_resid; 382 error = VOP_READ(vp, uio, ioflag, cred); 383 if (flags & FOF_UPDATE_OFFSET) 384 *offset += count - uio->uio_resid; 385 VOP_UNLOCK(vp, 0); 386 return (error); 387} 388 389/* 390 * File table vnode write routine. 391 */ 392int 393vn_write(fp, offset, uio, cred, flags) 394 struct file *fp; 395 off_t *offset; 396 struct uio *uio; 397 struct ucred *cred; 398 int flags; 399{ 400 struct vnode *vp = (struct vnode *)fp->f_data; 401 int count, error, ioflag = IO_UNIT; 402 403 if (vp->v_type == VREG && (fp->f_flag & O_APPEND)) 404 ioflag |= IO_APPEND; 405 if (fp->f_flag & FNONBLOCK) 406 ioflag |= IO_NDELAY; 407 if (fp->f_flag & FFSYNC || 408 (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS))) 409 ioflag |= IO_SYNC; 410 else if (fp->f_flag & FDSYNC) 411 ioflag |= IO_DSYNC; 412 if (fp->f_flag & FALTIO) 413 ioflag |= IO_ALTSEMANTICS; 414 VOP_LEASE(vp, uio->uio_procp, cred, LEASE_WRITE); 415 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 416 uio->uio_offset = *offset; 417 count = uio->uio_resid; 418 error = VOP_WRITE(vp, uio, ioflag, cred); 419 if (flags & FOF_UPDATE_OFFSET) { 420 if (ioflag & IO_APPEND) 421 *offset = uio->uio_offset; 422 else 423 *offset += count - uio->uio_resid; 424 } 425 VOP_UNLOCK(vp, 0); 426 return (error); 427} 428 429/* 430 * File table vnode stat routine. 431 */ 432int 433vn_stat(vp, sb, p) 434 struct vnode *vp; 435 struct stat *sb; 436 struct proc *p; 437{ 438 struct vattr va; 439 int error; 440 mode_t mode; 441 442 error = VOP_GETATTR(vp, &va, p->p_ucred, p); 443 if (error) 444 return (error); 445 /* 446 * Copy from vattr table 447 */ 448 sb->st_dev = va.va_fsid; 449 sb->st_ino = va.va_fileid; 450 mode = va.va_mode; 451 switch (vp->v_type) { 452 case VREG: 453 mode |= S_IFREG; 454 break; 455 case VDIR: 456 mode |= S_IFDIR; 457 break; 458 case VBLK: 459 mode |= S_IFBLK; 460 break; 461 case VCHR: 462 mode |= S_IFCHR; 463 break; 464 case VLNK: 465 mode |= S_IFLNK; 466 break; 467 case VSOCK: 468 mode |= S_IFSOCK; 469 break; 470 case VFIFO: 471 mode |= S_IFIFO; 472 break; 473 default: 474 return (EBADF); 475 }; 476 sb->st_mode = mode; 477 sb->st_nlink = va.va_nlink; 478 sb->st_uid = va.va_uid; 479 sb->st_gid = va.va_gid; 480 sb->st_rdev = va.va_rdev; 481 sb->st_size = va.va_size; 482 sb->st_atimespec = va.va_atime; 483 sb->st_mtimespec = va.va_mtime; 484 sb->st_ctimespec = va.va_ctime; 485 sb->st_blksize = va.va_blocksize; 486 sb->st_flags = va.va_flags; 487 sb->st_gen = 0; 488 sb->st_blocks = va.va_bytes / S_BLKSIZE; 489 return (0); 490} 491 492/* 493 * File table vnode fcntl routine. 494 */ 495int 496vn_fcntl(fp, com, data, p) 497 struct file *fp; 498 u_int com; 499 caddr_t data; 500 struct proc *p; 501{ 502 struct vnode *vp = ((struct vnode *)fp->f_data); 503 int error; 504 505 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 506 error = VOP_FCNTL(vp, com, data, fp->f_flag, p->p_ucred, p); 507 VOP_UNLOCK(vp, 0); 508 return (error); 509} 510 511/* 512 * File table vnode ioctl routine. 513 */ 514int 515vn_ioctl(fp, com, data, p) 516 struct file *fp; 517 u_long com; 518 caddr_t data; 519 struct proc *p; 520{ 521 struct vnode *vp = ((struct vnode *)fp->f_data); 522 struct vattr vattr; 523 int error; 524 525 switch (vp->v_type) { 526 527 case VREG: 528 case VDIR: 529 if (com == FIONREAD) { 530 error = VOP_GETATTR(vp, &vattr, p->p_ucred, p); 531 if (error) 532 return (error); 533 *(int *)data = vattr.va_size - fp->f_offset; 534 return (0); 535 } 536 if (com == FIONBIO || com == FIOASYNC) /* XXX */ 537 return (0); /* XXX */ 538 /* fall into ... */ 539 540 default: 541 return (ENOTTY); 542 543 case VFIFO: 544 case VCHR: 545 case VBLK: 546 error = VOP_IOCTL(vp, com, data, fp->f_flag, p->p_ucred, p); 547 if (error == 0 && com == TIOCSCTTY) { 548 if (p->p_session->s_ttyvp) 549 vrele(p->p_session->s_ttyvp); 550 p->p_session->s_ttyvp = vp; 551 VREF(vp); 552 } 553 return (error); 554 } 555} 556 557/* 558 * File table vnode poll routine. 559 */ 560int 561vn_poll(fp, events, p) 562 struct file *fp; 563 int events; 564 struct proc *p; 565{ 566 567 return (VOP_POLL(((struct vnode *)fp->f_data), events, p)); 568} 569 570/* 571 * Check that the vnode is still valid, and if so 572 * acquire requested lock. 573 */ 574int 575vn_lock(vp, flags) 576 struct vnode *vp; 577 int flags; 578{ 579 int error; 580 581 do { 582 if ((flags & LK_INTERLOCK) == 0) 583 simple_lock(&vp->v_interlock); 584 if (vp->v_flag & VXLOCK) { 585 vp->v_flag |= VXWANT; 586 ltsleep((caddr_t)vp, PINOD | PNORELOCK, 587 "vn_lock", 0, &vp->v_interlock); 588 error = ENOENT; 589 } else { 590 error = VOP_LOCK(vp, flags | LK_INTERLOCK); 591 if (error == 0 || error == EDEADLK) 592 return (error); 593 } 594 flags &= ~LK_INTERLOCK; 595 } while (flags & LK_RETRY); 596 return (error); 597} 598 599/* 600 * File table vnode close routine. 601 */ 602int 603vn_closefile(fp, p) 604 struct file *fp; 605 struct proc *p; 606{ 607 608 return (vn_close(((struct vnode *)fp->f_data), fp->f_flag, 609 fp->f_cred, p)); 610} 611 612/* 613 * Enable LK_CANRECURSE on lock. Return prior status. 614 */ 615u_int 616vn_setrecurse(vp) 617 struct vnode *vp; 618{ 619 struct lock *lkp = &vp->v_lock; 620 u_int retval = lkp->lk_flags & LK_CANRECURSE; 621 622 lkp->lk_flags |= LK_CANRECURSE; 623 return retval; 624} 625 626/* 627 * Called when done with locksetrecurse. 628 */ 629void 630vn_restorerecurse(vp, flags) 631 struct vnode *vp; 632 u_int flags; 633{ 634 struct lock *lkp = &vp->v_lock; 635 636 lkp->lk_flags &= ~LK_CANRECURSE; 637 lkp->lk_flags |= flags; 638} 639