1 /* $NetBSD: vfs_vnops.c,v 1.54 2002/03/17 19:41:08 atatat 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 <sys/cdefs.h> 44 __KERNEL_RCSID(0, "$NetBSD: vfs_vnops.c,v 1.54 2002/03/17 19:41:08 atatat Exp $"); 45 46 #include "fs_union.h" 47 48 #include <sys/param.h> 49 #include <sys/systm.h> 50 #include <sys/kernel.h> 51 #include <sys/file.h> 52 #include <sys/stat.h> 53 #include <sys/buf.h> 54 #include <sys/proc.h> 55 #include <sys/mount.h> 56 #include <sys/namei.h> 57 #include <sys/vnode.h> 58 #include <sys/ioctl.h> 59 #include <sys/tty.h> 60 #include <sys/poll.h> 61 62 #include <uvm/uvm_extern.h> 63 64 #ifdef UNION 65 #include <miscfs/union/union.h> 66 #endif 67 68 static int vn_statfile __P((struct file *fp, struct stat *sb, struct proc *p)); 69 70 struct fileops vnops = { 71 vn_read, vn_write, vn_ioctl, vn_fcntl, vn_poll, 72 vn_statfile, vn_closefile 73 }; 74 75 /* 76 * Common code for vnode open operations. 77 * Check permissions, and call the VOP_OPEN or VOP_CREATE routine. 78 */ 79 int 80 vn_open(ndp, fmode, cmode) 81 struct nameidata *ndp; 82 int fmode, cmode; 83 { 84 struct vnode *vp; 85 struct proc *p = ndp->ni_cnd.cn_proc; 86 struct ucred *cred = p->p_ucred; 87 struct vattr va; 88 int error; 89 90 if (fmode & O_CREAT) { 91 ndp->ni_cnd.cn_nameiop = CREATE; 92 ndp->ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 93 if ((fmode & O_EXCL) == 0 && 94 ((fmode & FNOSYMLINK) == 0)) 95 ndp->ni_cnd.cn_flags |= FOLLOW; 96 if ((error = namei(ndp)) != 0) 97 return (error); 98 if (ndp->ni_vp == NULL) { 99 VATTR_NULL(&va); 100 va.va_type = VREG; 101 va.va_mode = cmode; 102 if (fmode & O_EXCL) 103 va.va_vaflags |= VA_EXCLUSIVE; 104 VOP_LEASE(ndp->ni_dvp, p, cred, LEASE_WRITE); 105 error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp, 106 &ndp->ni_cnd, &va); 107 if (error) 108 return (error); 109 fmode &= ~O_TRUNC; 110 vp = ndp->ni_vp; 111 } else { 112 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 113 if (ndp->ni_dvp == ndp->ni_vp) 114 vrele(ndp->ni_dvp); 115 else 116 vput(ndp->ni_dvp); 117 ndp->ni_dvp = NULL; 118 vp = ndp->ni_vp; 119 if (fmode & O_EXCL) { 120 error = EEXIST; 121 goto bad; 122 } 123 if (ndp->ni_vp->v_type == VLNK) { 124 error = EFTYPE; 125 goto bad; 126 } 127 fmode &= ~O_CREAT; 128 } 129 } else { 130 ndp->ni_cnd.cn_nameiop = LOOKUP; 131 ndp->ni_cnd.cn_flags = FOLLOW | LOCKLEAF; 132 if ((error = namei(ndp)) != 0) 133 return (error); 134 vp = ndp->ni_vp; 135 } 136 if (vp->v_type == VSOCK) { 137 error = EOPNOTSUPP; 138 goto bad; 139 } 140 if ((fmode & O_CREAT) == 0) { 141 if (fmode & FREAD) { 142 if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0) 143 goto bad; 144 } 145 if (fmode & (FWRITE | O_TRUNC)) { 146 if (vp->v_type == VDIR) { 147 error = EISDIR; 148 goto bad; 149 } 150 if ((error = vn_writechk(vp)) != 0 || 151 (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0) 152 goto bad; 153 } 154 } 155 if (fmode & O_TRUNC) { 156 VOP_UNLOCK(vp, 0); /* XXX */ 157 VOP_LEASE(vp, p, cred, LEASE_WRITE); 158 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 159 VATTR_NULL(&va); 160 va.va_size = 0; 161 if ((error = VOP_SETATTR(vp, &va, cred, p)) != 0) 162 goto bad; 163 } 164 if ((error = VOP_OPEN(vp, fmode, cred, p)) != 0) 165 goto bad; 166 if (vp->v_type == VREG && 167 uvn_attach(vp, fmode & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 168 error = EIO; 169 goto bad; 170 } 171 if (fmode & FWRITE) 172 vp->v_writecount++; 173 174 return (0); 175 bad: 176 vput(vp); 177 return (error); 178 } 179 180 /* 181 * Check for write permissions on the specified vnode. 182 * Prototype text segments cannot be written. 183 */ 184 int 185 vn_writechk(vp) 186 struct vnode *vp; 187 { 188 189 /* 190 * If the vnode is in use as a process's text, 191 * we can't allow writing. 192 */ 193 if (vp->v_flag & VTEXT) 194 return (ETXTBSY); 195 return (0); 196 } 197 198 /* 199 * Mark a vnode as having executable mappings. 200 */ 201 void 202 vn_markexec(vp) 203 struct vnode *vp; 204 { 205 if ((vp->v_flag & VEXECMAP) == 0) { 206 uvmexp.filepages -= vp->v_uobj.uo_npages; 207 uvmexp.execpages += vp->v_uobj.uo_npages; 208 } 209 vp->v_flag |= VEXECMAP; 210 } 211 212 /* 213 * Vnode close call 214 * 215 * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node. 216 */ 217 int 218 vn_close(vp, flags, cred, p) 219 struct vnode *vp; 220 int flags; 221 struct ucred *cred; 222 struct proc *p; 223 { 224 int error; 225 226 if (flags & FWRITE) 227 vp->v_writecount--; 228 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 229 error = VOP_CLOSE(vp, flags, cred, p); 230 vput(vp); 231 return (error); 232 } 233 234 /* 235 * Package up an I/O request on a vnode into a uio and do it. 236 */ 237 int 238 vn_rdwr(rw, vp, base, len, offset, segflg, ioflg, cred, aresid, p) 239 enum uio_rw rw; 240 struct vnode *vp; 241 caddr_t base; 242 int len; 243 off_t offset; 244 enum uio_seg segflg; 245 int ioflg; 246 struct ucred *cred; 247 size_t *aresid; 248 struct proc *p; 249 { 250 struct uio auio; 251 struct iovec aiov; 252 int error; 253 254 if ((ioflg & IO_NODELOCKED) == 0) { 255 if (rw == UIO_READ) { 256 vn_lock(vp, LK_SHARED | LK_RETRY); 257 } else { 258 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 259 } 260 } 261 auio.uio_iov = &aiov; 262 auio.uio_iovcnt = 1; 263 aiov.iov_base = base; 264 aiov.iov_len = len; 265 auio.uio_resid = len; 266 auio.uio_offset = offset; 267 auio.uio_segflg = segflg; 268 auio.uio_rw = rw; 269 auio.uio_procp = p; 270 if (rw == UIO_READ) { 271 error = VOP_READ(vp, &auio, ioflg, cred); 272 } else { 273 error = VOP_WRITE(vp, &auio, ioflg, cred); 274 } 275 if (aresid) 276 *aresid = auio.uio_resid; 277 else 278 if (auio.uio_resid && error == 0) 279 error = EIO; 280 if ((ioflg & IO_NODELOCKED) == 0) 281 VOP_UNLOCK(vp, 0); 282 return (error); 283 } 284 285 int 286 vn_readdir(fp, buf, segflg, count, done, p, cookies, ncookies) 287 struct file *fp; 288 char *buf; 289 int segflg, *done, *ncookies; 290 u_int count; 291 struct proc *p; 292 off_t **cookies; 293 { 294 struct vnode *vp = (struct vnode *)fp->f_data; 295 struct iovec aiov; 296 struct uio auio; 297 int error, eofflag; 298 299 unionread: 300 if (vp->v_type != VDIR) 301 return (EINVAL); 302 aiov.iov_base = buf; 303 aiov.iov_len = count; 304 auio.uio_iov = &aiov; 305 auio.uio_iovcnt = 1; 306 auio.uio_rw = UIO_READ; 307 auio.uio_segflg = segflg; 308 auio.uio_procp = p; 309 auio.uio_resid = count; 310 vn_lock(vp, LK_SHARED | LK_RETRY); 311 auio.uio_offset = fp->f_offset; 312 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies, 313 ncookies); 314 fp->f_offset = auio.uio_offset; 315 VOP_UNLOCK(vp, 0); 316 if (error) 317 return (error); 318 319 #ifdef UNION 320 { 321 extern struct vnode *union_dircache __P((struct vnode *)); 322 323 if (count == auio.uio_resid && (vp->v_op == union_vnodeop_p)) { 324 struct vnode *lvp; 325 326 lvp = union_dircache(vp); 327 if (lvp != NULLVP) { 328 struct vattr va; 329 330 /* 331 * If the directory is opaque, 332 * then don't show lower entries 333 */ 334 error = VOP_GETATTR(vp, &va, fp->f_cred, p); 335 if (va.va_flags & OPAQUE) { 336 vput(lvp); 337 lvp = NULL; 338 } 339 } 340 341 if (lvp != NULLVP) { 342 error = VOP_OPEN(lvp, FREAD, fp->f_cred, p); 343 if (error) { 344 vput(lvp); 345 return (error); 346 } 347 VOP_UNLOCK(lvp, 0); 348 fp->f_data = (caddr_t) lvp; 349 fp->f_offset = 0; 350 error = vn_close(vp, FREAD, fp->f_cred, p); 351 if (error) 352 return (error); 353 vp = lvp; 354 goto unionread; 355 } 356 } 357 } 358 #endif /* UNION */ 359 360 if (count == auio.uio_resid && (vp->v_flag & VROOT) && 361 (vp->v_mount->mnt_flag & MNT_UNION)) { 362 struct vnode *tvp = vp; 363 vp = vp->v_mount->mnt_vnodecovered; 364 VREF(vp); 365 fp->f_data = (caddr_t) vp; 366 fp->f_offset = 0; 367 vrele(tvp); 368 goto unionread; 369 } 370 *done = count - auio.uio_resid; 371 return error; 372 } 373 374 /* 375 * File table vnode read routine. 376 */ 377 int 378 vn_read(fp, offset, uio, cred, flags) 379 struct file *fp; 380 off_t *offset; 381 struct uio *uio; 382 struct ucred *cred; 383 int flags; 384 { 385 struct vnode *vp = (struct vnode *)fp->f_data; 386 int count, error, ioflag = 0; 387 388 VOP_LEASE(vp, uio->uio_procp, cred, LEASE_READ); 389 if (fp->f_flag & FNONBLOCK) 390 ioflag |= IO_NDELAY; 391 if ((fp->f_flag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC)) 392 ioflag |= IO_SYNC; 393 if (fp->f_flag & FALTIO) 394 ioflag |= IO_ALTSEMANTICS; 395 vn_lock(vp, LK_SHARED | LK_RETRY); 396 uio->uio_offset = *offset; 397 count = uio->uio_resid; 398 error = VOP_READ(vp, uio, ioflag, cred); 399 if (flags & FOF_UPDATE_OFFSET) 400 *offset += count - uio->uio_resid; 401 VOP_UNLOCK(vp, 0); 402 return (error); 403 } 404 405 /* 406 * File table vnode write routine. 407 */ 408 int 409 vn_write(fp, offset, uio, cred, flags) 410 struct file *fp; 411 off_t *offset; 412 struct uio *uio; 413 struct ucred *cred; 414 int flags; 415 { 416 struct vnode *vp = (struct vnode *)fp->f_data; 417 int count, error, ioflag = IO_UNIT; 418 419 if (vp->v_type == VREG && (fp->f_flag & O_APPEND)) 420 ioflag |= IO_APPEND; 421 if (fp->f_flag & FNONBLOCK) 422 ioflag |= IO_NDELAY; 423 if (fp->f_flag & FFSYNC || 424 (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS))) 425 ioflag |= IO_SYNC; 426 else if (fp->f_flag & FDSYNC) 427 ioflag |= IO_DSYNC; 428 if (fp->f_flag & FALTIO) 429 ioflag |= IO_ALTSEMANTICS; 430 VOP_LEASE(vp, uio->uio_procp, cred, LEASE_WRITE); 431 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 432 uio->uio_offset = *offset; 433 count = uio->uio_resid; 434 error = VOP_WRITE(vp, uio, ioflag, cred); 435 if (flags & FOF_UPDATE_OFFSET) { 436 if (ioflag & IO_APPEND) 437 *offset = uio->uio_offset; 438 else 439 *offset += count - uio->uio_resid; 440 } 441 VOP_UNLOCK(vp, 0); 442 return (error); 443 } 444 445 /* 446 * File table vnode stat routine. 447 */ 448 static int 449 vn_statfile(fp, sb, p) 450 struct file *fp; 451 struct stat *sb; 452 struct proc *p; 453 { 454 struct vnode *vp = (struct vnode *)fp->f_data; 455 456 return vn_stat(vp, sb, p); 457 } 458 459 int 460 vn_stat(fdata, sb, p) 461 void *fdata; 462 struct stat *sb; 463 struct proc *p; 464 { 465 struct vnode *vp = fdata; 466 struct vattr va; 467 int error; 468 mode_t mode; 469 470 error = VOP_GETATTR(vp, &va, p->p_ucred, p); 471 if (error) 472 return (error); 473 /* 474 * Copy from vattr table 475 */ 476 sb->st_dev = va.va_fsid; 477 sb->st_ino = va.va_fileid; 478 mode = va.va_mode; 479 switch (vp->v_type) { 480 case VREG: 481 mode |= S_IFREG; 482 break; 483 case VDIR: 484 mode |= S_IFDIR; 485 break; 486 case VBLK: 487 mode |= S_IFBLK; 488 break; 489 case VCHR: 490 mode |= S_IFCHR; 491 break; 492 case VLNK: 493 mode |= S_IFLNK; 494 break; 495 case VSOCK: 496 mode |= S_IFSOCK; 497 break; 498 case VFIFO: 499 mode |= S_IFIFO; 500 break; 501 default: 502 return (EBADF); 503 }; 504 sb->st_mode = mode; 505 sb->st_nlink = va.va_nlink; 506 sb->st_uid = va.va_uid; 507 sb->st_gid = va.va_gid; 508 sb->st_rdev = va.va_rdev; 509 sb->st_size = va.va_size; 510 sb->st_atimespec = va.va_atime; 511 sb->st_mtimespec = va.va_mtime; 512 sb->st_ctimespec = va.va_ctime; 513 sb->st_blksize = va.va_blocksize; 514 sb->st_flags = va.va_flags; 515 sb->st_gen = 0; 516 sb->st_blocks = va.va_bytes / S_BLKSIZE; 517 return (0); 518 } 519 520 /* 521 * File table vnode fcntl routine. 522 */ 523 int 524 vn_fcntl(fp, com, data, p) 525 struct file *fp; 526 u_int com; 527 caddr_t data; 528 struct proc *p; 529 { 530 struct vnode *vp = ((struct vnode *)fp->f_data); 531 int error; 532 533 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 534 error = VOP_FCNTL(vp, com, data, fp->f_flag, p->p_ucred, p); 535 VOP_UNLOCK(vp, 0); 536 return (error); 537 } 538 539 /* 540 * File table vnode ioctl routine. 541 */ 542 int 543 vn_ioctl(fp, com, data, p) 544 struct file *fp; 545 u_long com; 546 caddr_t data; 547 struct proc *p; 548 { 549 struct vnode *vp = ((struct vnode *)fp->f_data); 550 struct vattr vattr; 551 int error; 552 553 switch (vp->v_type) { 554 555 case VREG: 556 case VDIR: 557 if (com == FIONREAD) { 558 error = VOP_GETATTR(vp, &vattr, p->p_ucred, p); 559 if (error) 560 return (error); 561 *(int *)data = vattr.va_size - fp->f_offset; 562 return (0); 563 } 564 if (com == FIONBIO || com == FIOASYNC) /* XXX */ 565 return (0); /* XXX */ 566 /* fall into ... */ 567 568 default: 569 return (EPASSTHROUGH); 570 571 case VFIFO: 572 case VCHR: 573 case VBLK: 574 error = VOP_IOCTL(vp, com, data, fp->f_flag, p->p_ucred, p); 575 if (error == 0 && com == TIOCSCTTY) { 576 if (p->p_session->s_ttyvp) 577 vrele(p->p_session->s_ttyvp); 578 p->p_session->s_ttyvp = vp; 579 VREF(vp); 580 } 581 return (error); 582 } 583 } 584 585 /* 586 * File table vnode poll routine. 587 */ 588 int 589 vn_poll(fp, events, p) 590 struct file *fp; 591 int events; 592 struct proc *p; 593 { 594 595 return (VOP_POLL(((struct vnode *)fp->f_data), events, p)); 596 } 597 598 /* 599 * Check that the vnode is still valid, and if so 600 * acquire requested lock. 601 */ 602 int 603 vn_lock(vp, flags) 604 struct vnode *vp; 605 int flags; 606 { 607 int error; 608 609 do { 610 if ((flags & LK_INTERLOCK) == 0) 611 simple_lock(&vp->v_interlock); 612 if (vp->v_flag & VXLOCK) { 613 if (flags & LK_NOWAIT) { 614 simple_unlock(&vp->v_interlock); 615 return EBUSY; 616 } 617 vp->v_flag |= VXWANT; 618 ltsleep(vp, PINOD | PNORELOCK, 619 "vn_lock", 0, &vp->v_interlock); 620 error = ENOENT; 621 } else { 622 error = VOP_LOCK(vp, flags | LK_INTERLOCK); 623 if (error == 0 || error == EDEADLK || error == EBUSY) 624 return (error); 625 } 626 flags &= ~LK_INTERLOCK; 627 } while (flags & LK_RETRY); 628 return (error); 629 } 630 631 /* 632 * File table vnode close routine. 633 */ 634 int 635 vn_closefile(fp, p) 636 struct file *fp; 637 struct proc *p; 638 { 639 640 return (vn_close(((struct vnode *)fp->f_data), fp->f_flag, 641 fp->f_cred, p)); 642 } 643 644 /* 645 * Enable LK_CANRECURSE on lock. Return prior status. 646 */ 647 u_int 648 vn_setrecurse(vp) 649 struct vnode *vp; 650 { 651 struct lock *lkp = &vp->v_lock; 652 u_int retval = lkp->lk_flags & LK_CANRECURSE; 653 654 lkp->lk_flags |= LK_CANRECURSE; 655 return retval; 656 } 657 658 /* 659 * Called when done with locksetrecurse. 660 */ 661 void 662 vn_restorerecurse(vp, flags) 663 struct vnode *vp; 664 u_int flags; 665 { 666 struct lock *lkp = &vp->v_lock; 667 668 lkp->lk_flags &= ~LK_CANRECURSE; 669 lkp->lk_flags |= flags; 670 } 671