1 /* 2 * Copyright (c) 1982, 1986, 1989, 1993 3 * The Regents of the University of California. All rights reserved. 4 * (c) UNIX System Laboratories, Inc. 5 * All or some portions of this file are derived from material licensed 6 * to the University of California by American Telephone and Telegraph 7 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 8 * the permission of UNIX System Laboratories, Inc. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. All advertising materials mentioning features or use of this software 19 * must display the following acknowledgement: 20 * This product includes software developed by the University of 21 * California, Berkeley and its contributors. 22 * 4. Neither the name of the University nor the names of its contributors 23 * may be used to endorse or promote products derived from this software 24 * without specific prior written permission. 25 * 26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 36 * SUCH DAMAGE. 37 * 38 * @(#)vfs_vnops.c 8.2 (Berkeley) 1/21/94 39 * $FreeBSD: src/sys/kern/vfs_vnops.c,v 1.87.2.13 2002/12/29 18:19:53 dillon Exp $ 40 * $DragonFly: src/sys/kern/vfs_vnops.c,v 1.15 2003/10/09 22:27:19 dillon Exp $ 41 */ 42 43 #include <sys/param.h> 44 #include <sys/systm.h> 45 #include <sys/fcntl.h> 46 #include <sys/file.h> 47 #include <sys/stat.h> 48 #include <sys/proc.h> 49 #include <sys/mount.h> 50 #include <sys/namei.h> 51 #include <sys/vnode.h> 52 #include <sys/buf.h> 53 #include <sys/filio.h> 54 #include <sys/ttycom.h> 55 #include <sys/conf.h> 56 #include <sys/syslog.h> 57 58 static int vn_closefile (struct file *fp, struct thread *td); 59 static int vn_ioctl (struct file *fp, u_long com, caddr_t data, 60 struct thread *td); 61 static int vn_read (struct file *fp, struct uio *uio, 62 struct ucred *cred, int flags, struct thread *td); 63 static int vn_poll (struct file *fp, int events, struct ucred *cred, 64 struct thread *td); 65 static int vn_kqfilter (struct file *fp, struct knote *kn); 66 static int vn_statfile (struct file *fp, struct stat *sb, struct thread *td); 67 static int vn_write (struct file *fp, struct uio *uio, 68 struct ucred *cred, int flags, struct thread *td); 69 70 struct fileops vnops = { 71 NULL, /* port */ 72 0, /* autoq */ 73 vn_read, vn_write, vn_ioctl, vn_poll, vn_kqfilter, 74 vn_statfile, vn_closefile 75 }; 76 77 /* 78 * Common code for vnode open operations. 79 * Check permissions, and call the VOP_OPEN or VOP_CREATE routine. 80 * 81 * Note that this does NOT free nameidata for the successful case, 82 * due to the NDINIT being done elsewhere. 83 */ 84 int 85 vn_open(ndp, fmode, cmode) 86 struct nameidata *ndp; 87 int fmode, cmode; 88 { 89 struct vnode *vp; 90 struct thread *td = ndp->ni_cnd.cn_td; 91 struct ucred *cred = ndp->ni_cnd.cn_cred; 92 struct vattr vat; 93 struct vattr *vap = &vat; 94 int mode, error; 95 96 KKASSERT(cred == td->td_proc->p_ucred); 97 98 if (fmode & O_CREAT) { 99 ndp->ni_cnd.cn_nameiop = NAMEI_CREATE; 100 ndp->ni_cnd.cn_flags = CNP_LOCKPARENT | CNP_LOCKLEAF; 101 if ((fmode & O_EXCL) == 0 && (fmode & O_NOFOLLOW) == 0) 102 ndp->ni_cnd.cn_flags |= CNP_FOLLOW; 103 bwillwrite(); 104 error = namei(ndp); 105 if (error) 106 return (error); 107 if (ndp->ni_vp == NULL) { 108 VATTR_NULL(vap); 109 vap->va_type = VREG; 110 vap->va_mode = cmode; 111 if (fmode & O_EXCL) 112 vap->va_vaflags |= VA_EXCLUSIVE; 113 VOP_LEASE(ndp->ni_dvp, td, cred, LEASE_WRITE); 114 error = VOP_CREATE(ndp->ni_dvp, NCPNULL, &ndp->ni_vp, 115 &ndp->ni_cnd, vap); 116 if (error) { 117 NDFREE(ndp, NDF_ONLY_PNBUF); 118 vput(ndp->ni_dvp); 119 return (error); 120 } 121 vput(ndp->ni_dvp); 122 ASSERT_VOP_UNLOCKED(ndp->ni_dvp, "create"); 123 ASSERT_VOP_LOCKED(ndp->ni_vp, "create"); 124 fmode &= ~O_TRUNC; 125 vp = ndp->ni_vp; 126 } else { 127 if (ndp->ni_dvp == ndp->ni_vp) 128 vrele(ndp->ni_dvp); 129 else 130 vput(ndp->ni_dvp); 131 ndp->ni_dvp = NULL; 132 vp = ndp->ni_vp; 133 if (fmode & O_EXCL) { 134 error = EEXIST; 135 goto bad; 136 } 137 fmode &= ~O_CREAT; 138 } 139 } else { 140 ndp->ni_cnd.cn_nameiop = NAMEI_LOOKUP; 141 ndp->ni_cnd.cn_flags = CNP_LOCKLEAF | 142 ((fmode & O_NOFOLLOW) ? 0 : CNP_FOLLOW); 143 error = namei(ndp); 144 if (error) 145 return (error); 146 vp = ndp->ni_vp; 147 } 148 if (vp->v_type == VLNK) { 149 error = EMLINK; 150 goto bad; 151 } 152 if (vp->v_type == VSOCK) { 153 error = EOPNOTSUPP; 154 goto bad; 155 } 156 if ((fmode & O_CREAT) == 0) { 157 mode = 0; 158 if (fmode & (FWRITE | O_TRUNC)) { 159 if (vp->v_type == VDIR) { 160 error = EISDIR; 161 goto bad; 162 } 163 error = vn_writechk(vp); 164 if (error) 165 goto bad; 166 mode |= VWRITE; 167 } 168 if (fmode & FREAD) 169 mode |= VREAD; 170 if (mode) { 171 error = VOP_ACCESS(vp, mode, cred, td); 172 if (error) 173 goto bad; 174 } 175 } 176 if (fmode & O_TRUNC) { 177 VOP_UNLOCK(vp, 0, td); /* XXX */ 178 VOP_LEASE(vp, td, cred, LEASE_WRITE); 179 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); /* XXX */ 180 VATTR_NULL(vap); 181 vap->va_size = 0; 182 error = VOP_SETATTR(vp, vap, cred, td); 183 if (error) 184 goto bad; 185 } 186 error = VOP_OPEN(vp, fmode, cred, td); 187 if (error) 188 goto bad; 189 /* 190 * Make sure that a VM object is created for VMIO support. 191 */ 192 if (vn_canvmio(vp) == TRUE) { 193 if ((error = vfs_object_create(vp, td)) != 0) 194 goto bad; 195 } 196 197 if (fmode & FWRITE) 198 vp->v_writecount++; 199 return (0); 200 bad: 201 NDFREE(ndp, NDF_ONLY_PNBUF); 202 vput(vp); 203 return (error); 204 } 205 206 /* 207 * Check for write permissions on the specified vnode. 208 * Prototype text segments cannot be written. 209 */ 210 int 211 vn_writechk(vp) 212 struct vnode *vp; 213 { 214 215 /* 216 * If there's shared text associated with 217 * the vnode, try to free it up once. If 218 * we fail, we can't allow writing. 219 */ 220 if (vp->v_flag & VTEXT) 221 return (ETXTBSY); 222 return (0); 223 } 224 225 /* 226 * Vnode close call 227 */ 228 int 229 vn_close(struct vnode *vp, int flags, struct thread *td) 230 { 231 int error; 232 233 if (flags & FWRITE) 234 vp->v_writecount--; 235 error = VOP_CLOSE(vp, flags, td); 236 vrele(vp); 237 return (error); 238 } 239 240 static __inline 241 int 242 sequential_heuristic(struct uio *uio, struct file *fp) 243 { 244 /* 245 * Sequential heuristic - detect sequential operation 246 */ 247 if ((uio->uio_offset == 0 && fp->f_seqcount > 0) || 248 uio->uio_offset == fp->f_nextoff) { 249 int tmpseq = fp->f_seqcount; 250 /* 251 * XXX we assume that the filesystem block size is 252 * the default. Not true, but still gives us a pretty 253 * good indicator of how sequential the read operations 254 * are. 255 */ 256 tmpseq += (uio->uio_resid + BKVASIZE - 1) / BKVASIZE; 257 if (tmpseq > IO_SEQMAX) 258 tmpseq = IO_SEQMAX; 259 fp->f_seqcount = tmpseq; 260 return(fp->f_seqcount << IO_SEQSHIFT); 261 } 262 263 /* 264 * Not sequential, quick draw-down of seqcount 265 */ 266 if (fp->f_seqcount > 1) 267 fp->f_seqcount = 1; 268 else 269 fp->f_seqcount = 0; 270 return(0); 271 } 272 273 /* 274 * Package up an I/O request on a vnode into a uio and do it. 275 */ 276 int 277 vn_rdwr(rw, vp, base, len, offset, segflg, ioflg, cred, aresid, td) 278 enum uio_rw rw; 279 struct vnode *vp; 280 caddr_t base; 281 int len; 282 off_t offset; 283 enum uio_seg segflg; 284 int ioflg; 285 struct ucred *cred; 286 int *aresid; 287 struct thread *td; 288 { 289 struct uio auio; 290 struct iovec aiov; 291 int error; 292 293 if ((ioflg & IO_NODELOCKED) == 0) 294 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); 295 auio.uio_iov = &aiov; 296 auio.uio_iovcnt = 1; 297 aiov.iov_base = base; 298 aiov.iov_len = len; 299 auio.uio_resid = len; 300 auio.uio_offset = offset; 301 auio.uio_segflg = segflg; 302 auio.uio_rw = rw; 303 auio.uio_td = td; 304 if (rw == UIO_READ) { 305 error = VOP_READ(vp, &auio, ioflg, cred); 306 } else { 307 error = VOP_WRITE(vp, &auio, ioflg, cred); 308 } 309 if (aresid) 310 *aresid = auio.uio_resid; 311 else 312 if (auio.uio_resid && error == 0) 313 error = EIO; 314 if ((ioflg & IO_NODELOCKED) == 0) 315 VOP_UNLOCK(vp, 0, td); 316 return (error); 317 } 318 319 /* 320 * Package up an I/O request on a vnode into a uio and do it. The I/O 321 * request is split up into smaller chunks and we try to avoid saturating 322 * the buffer cache while potentially holding a vnode locked, so we 323 * check bwillwrite() before calling vn_rdwr(). We also call uio_yield() 324 * to give other processes a chance to lock the vnode (either other processes 325 * core'ing the same binary, or unrelated processes scanning the directory). 326 */ 327 int 328 vn_rdwr_inchunks(rw, vp, base, len, offset, segflg, ioflg, cred, aresid, td) 329 enum uio_rw rw; 330 struct vnode *vp; 331 caddr_t base; 332 int len; 333 off_t offset; 334 enum uio_seg segflg; 335 int ioflg; 336 struct ucred *cred; 337 int *aresid; 338 struct thread *td; 339 { 340 int error = 0; 341 342 do { 343 int chunk = (len > MAXBSIZE) ? MAXBSIZE : len; 344 345 if (rw != UIO_READ && vp->v_type == VREG) 346 bwillwrite(); 347 error = vn_rdwr(rw, vp, base, chunk, offset, segflg, 348 ioflg, cred, aresid, td); 349 len -= chunk; /* aresid calc already includes length */ 350 if (error) 351 break; 352 offset += chunk; 353 base += chunk; 354 uio_yield(); 355 } while (len); 356 if (aresid) 357 *aresid += len; 358 return (error); 359 } 360 361 /* 362 * File table vnode read routine. 363 */ 364 static int 365 vn_read(fp, uio, cred, flags, td) 366 struct file *fp; 367 struct uio *uio; 368 struct ucred *cred; 369 struct thread *td; 370 int flags; 371 { 372 struct vnode *vp; 373 int error, ioflag; 374 375 KASSERT(uio->uio_td == td, ("uio_td %p is not td %p", uio->uio_td, td)); 376 vp = (struct vnode *)fp->f_data; 377 ioflag = 0; 378 if (fp->f_flag & FNONBLOCK) 379 ioflag |= IO_NDELAY; 380 if (fp->f_flag & O_DIRECT) 381 ioflag |= IO_DIRECT; 382 VOP_LEASE(vp, td, cred, LEASE_READ); 383 vn_lock(vp, LK_SHARED | LK_NOPAUSE | LK_RETRY, td); 384 if ((flags & FOF_OFFSET) == 0) 385 uio->uio_offset = fp->f_offset; 386 387 ioflag |= sequential_heuristic(uio, fp); 388 389 error = VOP_READ(vp, uio, ioflag, cred); 390 if ((flags & FOF_OFFSET) == 0) 391 fp->f_offset = uio->uio_offset; 392 fp->f_nextoff = uio->uio_offset; 393 VOP_UNLOCK(vp, 0, td); 394 return (error); 395 } 396 397 /* 398 * File table vnode write routine. 399 */ 400 static int 401 vn_write(fp, uio, cred, flags, td) 402 struct file *fp; 403 struct uio *uio; 404 struct ucred *cred; 405 struct thread *td; 406 int flags; 407 { 408 struct vnode *vp; 409 int error, ioflag; 410 411 KASSERT(uio->uio_td == td, ("uio_procp %p is not p %p", 412 uio->uio_td, td)); 413 vp = (struct vnode *)fp->f_data; 414 if (vp->v_type == VREG) 415 bwillwrite(); 416 vp = (struct vnode *)fp->f_data; /* XXX needed? */ 417 ioflag = IO_UNIT; 418 if (vp->v_type == VREG && (fp->f_flag & O_APPEND)) 419 ioflag |= IO_APPEND; 420 if (fp->f_flag & FNONBLOCK) 421 ioflag |= IO_NDELAY; 422 if (fp->f_flag & O_DIRECT) 423 ioflag |= IO_DIRECT; 424 if ((fp->f_flag & O_FSYNC) || 425 (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS))) 426 ioflag |= IO_SYNC; 427 VOP_LEASE(vp, td, cred, LEASE_WRITE); 428 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY, td); 429 if ((flags & FOF_OFFSET) == 0) 430 uio->uio_offset = fp->f_offset; 431 ioflag |= sequential_heuristic(uio, fp); 432 error = VOP_WRITE(vp, uio, ioflag, cred); 433 if ((flags & FOF_OFFSET) == 0) 434 fp->f_offset = uio->uio_offset; 435 fp->f_nextoff = uio->uio_offset; 436 VOP_UNLOCK(vp, 0, td); 437 return (error); 438 } 439 440 /* 441 * File table vnode stat routine. 442 */ 443 static int 444 vn_statfile(struct file *fp, struct stat *sb, struct thread *td) 445 { 446 struct vnode *vp = (struct vnode *)fp->f_data; 447 448 return vn_stat(vp, sb, td); 449 } 450 451 int 452 vn_stat(struct vnode *vp, struct stat *sb, struct thread *td) 453 { 454 struct vattr vattr; 455 struct vattr *vap; 456 int error; 457 u_short mode; 458 459 vap = &vattr; 460 error = VOP_GETATTR(vp, vap, td); 461 if (error) 462 return (error); 463 464 /* 465 * Zero the spare stat fields 466 */ 467 sb->st_lspare = 0; 468 sb->st_qspare[0] = 0; 469 sb->st_qspare[1] = 0; 470 471 /* 472 * Copy from vattr table 473 */ 474 if (vap->va_fsid != VNOVAL) 475 sb->st_dev = vap->va_fsid; 476 else 477 sb->st_dev = vp->v_mount->mnt_stat.f_fsid.val[0]; 478 sb->st_ino = vap->va_fileid; 479 mode = vap->va_mode; 480 switch (vap->va_type) { 481 case VREG: 482 mode |= S_IFREG; 483 break; 484 case VDIR: 485 mode |= S_IFDIR; 486 break; 487 case VBLK: 488 mode |= S_IFBLK; 489 break; 490 case VCHR: 491 mode |= S_IFCHR; 492 break; 493 case VLNK: 494 mode |= S_IFLNK; 495 /* This is a cosmetic change, symlinks do not have a mode. */ 496 if (vp->v_mount->mnt_flag & MNT_NOSYMFOLLOW) 497 sb->st_mode &= ~ACCESSPERMS; /* 0000 */ 498 else 499 sb->st_mode |= ACCESSPERMS; /* 0777 */ 500 break; 501 case VSOCK: 502 mode |= S_IFSOCK; 503 break; 504 case VFIFO: 505 mode |= S_IFIFO; 506 break; 507 default: 508 return (EBADF); 509 }; 510 sb->st_mode = mode; 511 sb->st_nlink = vap->va_nlink; 512 sb->st_uid = vap->va_uid; 513 sb->st_gid = vap->va_gid; 514 sb->st_rdev = vap->va_rdev; 515 sb->st_size = vap->va_size; 516 sb->st_atimespec = vap->va_atime; 517 sb->st_mtimespec = vap->va_mtime; 518 sb->st_ctimespec = vap->va_ctime; 519 520 /* 521 * According to www.opengroup.org, the meaning of st_blksize is 522 * "a filesystem-specific preferred I/O block size for this 523 * object. In some filesystem types, this may vary from file 524 * to file" 525 * Default to PAGE_SIZE after much discussion. 526 */ 527 528 if (vap->va_type == VREG) { 529 sb->st_blksize = vap->va_blocksize; 530 } else if (vn_isdisk(vp, NULL)) { 531 sb->st_blksize = vp->v_rdev->si_bsize_best; 532 if (sb->st_blksize < vp->v_rdev->si_bsize_phys) 533 sb->st_blksize = vp->v_rdev->si_bsize_phys; 534 if (sb->st_blksize < BLKDEV_IOSIZE) 535 sb->st_blksize = BLKDEV_IOSIZE; 536 } else { 537 sb->st_blksize = PAGE_SIZE; 538 } 539 540 sb->st_flags = vap->va_flags; 541 if (suser(td)) 542 sb->st_gen = 0; 543 else 544 sb->st_gen = vap->va_gen; 545 546 #if (S_BLKSIZE == 512) 547 /* Optimize this case */ 548 sb->st_blocks = vap->va_bytes >> 9; 549 #else 550 sb->st_blocks = vap->va_bytes / S_BLKSIZE; 551 #endif 552 return (0); 553 } 554 555 /* 556 * File table vnode ioctl routine. 557 */ 558 static int 559 vn_ioctl(struct file *fp, u_long com, caddr_t data, struct thread *td) 560 { 561 struct vnode *vp = ((struct vnode *)fp->f_data); 562 struct ucred *ucred; 563 struct vattr vattr; 564 int error; 565 566 KKASSERT(td->td_proc != NULL); 567 ucred = td->td_proc->p_ucred; 568 569 switch (vp->v_type) { 570 case VREG: 571 case VDIR: 572 if (com == FIONREAD) { 573 error = VOP_GETATTR(vp, &vattr, td); 574 if (error) 575 return (error); 576 *(int *)data = vattr.va_size - fp->f_offset; 577 return (0); 578 } 579 if (com == FIONBIO || com == FIOASYNC) /* XXX */ 580 return (0); /* XXX */ 581 /* fall into ... */ 582 default: 583 #if 0 584 return (ENOTTY); 585 #endif 586 case VFIFO: 587 case VCHR: 588 case VBLK: 589 if (com == FIODTYPE) { 590 if (vp->v_type != VCHR && vp->v_type != VBLK) 591 return (ENOTTY); 592 *(int *)data = dev_dflags(vp->v_rdev) & D_TYPEMASK; 593 return (0); 594 } 595 error = VOP_IOCTL(vp, com, data, fp->f_flag, ucred, td); 596 if (error == 0 && com == TIOCSCTTY) { 597 struct session *sess = td->td_proc->p_session; 598 599 /* Do nothing if reassigning same control tty */ 600 if (sess->s_ttyvp == vp) 601 return (0); 602 603 /* Get rid of reference to old control tty */ 604 if (sess->s_ttyvp) 605 vrele(sess->s_ttyvp); 606 607 sess->s_ttyvp = vp; 608 VREF(vp); 609 } 610 return (error); 611 } 612 } 613 614 /* 615 * File table vnode poll routine. 616 */ 617 static int 618 vn_poll(struct file *fp, int events, struct ucred *cred, struct thread *td) 619 { 620 return (VOP_POLL(((struct vnode *)fp->f_data), events, cred, td)); 621 } 622 623 /* 624 * Check that the vnode is still valid, and if so 625 * acquire requested lock. 626 */ 627 int 628 #ifndef DEBUG_LOCKS 629 vn_lock(struct vnode *vp, int flags, struct thread *td) 630 #else 631 debug_vn_lock(struct vnode *vp, int flags, struct thread *td, 632 const char *filename, int line) 633 #endif 634 { 635 int error; 636 637 do { 638 if ((flags & LK_INTERLOCK) == 0) 639 lwkt_gettoken(&vp->v_interlock); 640 if ((vp->v_flag & VXLOCK) && vp->v_vxproc != curproc) { 641 vp->v_flag |= VXWANT; 642 lwkt_reltoken(&vp->v_interlock); 643 tsleep((caddr_t)vp, 0, "vn_lock", 0); 644 error = ENOENT; 645 } else { 646 #if 0 647 /* this can now occur in normal operation */ 648 if (vp->v_vxproc != NULL) 649 log(LOG_INFO, "VXLOCK interlock avoided in vn_lock\n"); 650 #endif 651 #ifdef DEBUG_LOCKS 652 vp->filename = filename; 653 vp->line = line; 654 #endif 655 error = VOP_LOCK(vp, 656 flags | LK_NOPAUSE | LK_INTERLOCK, td); 657 if (error == 0) 658 return (error); 659 } 660 flags &= ~LK_INTERLOCK; 661 } while (flags & LK_RETRY); 662 return (error); 663 } 664 665 /* 666 * File table vnode close routine. 667 */ 668 static int 669 vn_closefile(struct file *fp, struct thread *td) 670 { 671 int err; 672 673 fp->f_ops = &badfileops; 674 err = vn_close(((struct vnode *)fp->f_data), fp->f_flag, td); 675 return(err); 676 } 677 678 static int 679 vn_kqfilter(struct file *fp, struct knote *kn) 680 { 681 682 return (VOP_KQFILTER(((struct vnode *)fp->f_data), kn)); 683 } 684