vfs_vnops.c revision 1.86.2.11
1 /* $NetBSD: vfs_vnops.c,v 1.86.2.11 2006/05/31 21:18:59 tron 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. Neither the name of the University nor the names of its contributors 21 * may be used to endorse or promote products derived from this software 22 * without specific prior written permission. 23 * 24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 34 * SUCH DAMAGE. 35 * 36 * @(#)vfs_vnops.c 8.14 (Berkeley) 6/15/95 37 */ 38 39 #include <sys/cdefs.h> 40 __KERNEL_RCSID(0, "$NetBSD: vfs_vnops.c,v 1.86.2.11 2006/05/31 21:18:59 tron Exp $"); 41 42 #include "opt_verified_exec.h" 43 44 #include "fs_union.h" 45 46 #include <sys/param.h> 47 #include <sys/systm.h> 48 #include <sys/kernel.h> 49 #include <sys/file.h> 50 #include <sys/stat.h> 51 #include <sys/buf.h> 52 #include <sys/proc.h> 53 #include <sys/malloc.h> 54 #include <sys/mount.h> 55 #include <sys/namei.h> 56 #include <sys/vnode.h> 57 #include <sys/ioctl.h> 58 #include <sys/tty.h> 59 #include <sys/poll.h> 60 61 #include <miscfs/specfs/specdev.h> 62 63 #include <uvm/uvm_extern.h> 64 65 #ifdef UNION 66 #include <fs/union/union.h> 67 #endif 68 69 #if defined(LKM) || defined(UNION) 70 int (*vn_union_readdir_hook) (struct vnode **, struct file *, struct proc *); 71 #endif 72 73 #ifdef VERIFIED_EXEC 74 #include <sys/verified_exec.h> 75 #endif 76 77 static int vn_read(struct file *fp, off_t *offset, struct uio *uio, 78 struct ucred *cred, int flags); 79 static int vn_write(struct file *fp, off_t *offset, struct uio *uio, 80 struct ucred *cred, int flags); 81 static int vn_closefile(struct file *fp, struct proc *p); 82 static int vn_poll(struct file *fp, int events, struct proc *p); 83 static int vn_fcntl(struct file *fp, u_int com, void *data, struct proc *p); 84 static int vn_statfile(struct file *fp, struct stat *sb, struct proc *p); 85 static int vn_ioctl(struct file *fp, u_long com, void *data, struct proc *p); 86 87 const struct fileops vnops = { 88 vn_read, vn_write, vn_ioctl, vn_fcntl, vn_poll, 89 vn_statfile, vn_closefile, vn_kqfilter 90 }; 91 92 /* 93 * Common code for vnode open operations. 94 * Check permissions, and call the VOP_OPEN or VOP_CREATE routine. 95 */ 96 int 97 vn_open(ndp, fmode, cmode) 98 struct nameidata *ndp; 99 int fmode, cmode; 100 { 101 struct vnode *vp; 102 struct mount *mp; 103 struct proc *p = ndp->ni_cnd.cn_proc; 104 struct ucred *cred = p->p_ucred; 105 struct vattr va; 106 int error; 107 #ifdef VERIFIED_EXEC 108 struct veriexec_hash_entry *vhe = NULL; 109 char pathbuf[MAXPATHLEN]; 110 size_t pathlen; 111 int (*copyfun)(const void *, void *, size_t, size_t *) = 112 ndp->ni_segflg == UIO_SYSSPACE ? copystr : copyinstr; 113 #endif /* VERIFIED_EXEC */ 114 115 #ifdef VERIFIED_EXEC 116 error = (*copyfun)(ndp->ni_dirp, pathbuf, sizeof(pathbuf), &pathlen); 117 if (error) { 118 if (veriexec_verbose >= 1) 119 printf("veriexec: Can't copy path. (error=%d)\n", 120 error); 121 122 return (error); 123 } 124 #endif /* VERIFIED_EXEC */ 125 126 restart: 127 if (fmode & O_CREAT) { 128 ndp->ni_cnd.cn_nameiop = CREATE; 129 ndp->ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 130 if ((fmode & O_EXCL) == 0 && 131 ((fmode & O_NOFOLLOW) == 0)) 132 ndp->ni_cnd.cn_flags |= FOLLOW; 133 if ((error = namei(ndp)) != 0) 134 return (error); 135 if (ndp->ni_vp == NULL) { 136 #ifdef VERIFIED_EXEC 137 /* Lockdown mode: Prevent creation of new files. */ 138 if (veriexec_strict >= 3) { 139 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 140 141 printf("Veriexec: vn_open: Preventing " 142 "new file creation in %s.\n", 143 pathbuf); 144 145 vp = ndp->ni_dvp; 146 error = EPERM; 147 goto bad; 148 } 149 #endif /* VERIFIED_EXEC */ 150 151 VATTR_NULL(&va); 152 va.va_type = VREG; 153 va.va_mode = cmode; 154 if (fmode & O_EXCL) 155 va.va_vaflags |= VA_EXCLUSIVE; 156 if (vn_start_write(ndp->ni_dvp, &mp, V_NOWAIT) != 0) { 157 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 158 vput(ndp->ni_dvp); 159 if ((error = vn_start_write(NULL, &mp, 160 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 161 return (error); 162 goto restart; 163 } 164 VOP_LEASE(ndp->ni_dvp, p, cred, LEASE_WRITE); 165 error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp, 166 &ndp->ni_cnd, &va); 167 vn_finished_write(mp, 0); 168 if (error) 169 return (error); 170 fmode &= ~O_TRUNC; 171 vp = ndp->ni_vp; 172 } else { 173 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 174 if (ndp->ni_dvp == ndp->ni_vp) 175 vrele(ndp->ni_dvp); 176 else 177 vput(ndp->ni_dvp); 178 ndp->ni_dvp = NULL; 179 vp = ndp->ni_vp; 180 if (fmode & O_EXCL) { 181 error = EEXIST; 182 goto bad; 183 } 184 fmode &= ~O_CREAT; 185 } 186 } else { 187 ndp->ni_cnd.cn_nameiop = LOOKUP; 188 ndp->ni_cnd.cn_flags = LOCKLEAF; 189 if ((fmode & O_NOFOLLOW) == 0) 190 ndp->ni_cnd.cn_flags |= FOLLOW; 191 if ((error = namei(ndp)) != 0) 192 return (error); 193 vp = ndp->ni_vp; 194 } 195 if (vp->v_type == VSOCK) { 196 error = EOPNOTSUPP; 197 goto bad; 198 } 199 if (ndp->ni_vp->v_type == VLNK) { 200 error = EFTYPE; 201 goto bad; 202 } 203 204 #ifdef VERIFIED_EXEC 205 if ((error = VOP_GETATTR(vp, &va, cred, p)) != 0) 206 goto bad; 207 #endif 208 209 if ((fmode & O_CREAT) == 0) { 210 #ifdef VERIFIED_EXEC 211 if ((error = veriexec_verify(p, vp, &va, pathbuf, 212 VERIEXEC_FILE, &vhe)) != 0) 213 goto bad; 214 #endif 215 216 if (fmode & FREAD) { 217 if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0) 218 goto bad; 219 } 220 221 if (fmode & (FWRITE | O_TRUNC)) { 222 if (vp->v_type == VDIR) { 223 error = EISDIR; 224 goto bad; 225 } 226 if ((error = vn_writechk(vp)) != 0 || 227 (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0) 228 goto bad; 229 #ifdef VERIFIED_EXEC 230 if (vhe != NULL) { 231 veriexec_report("Write access request.", 232 pathbuf, &va, p, 233 REPORT_NOVERBOSE, 234 REPORT_ALARM, 235 REPORT_NOPANIC); 236 237 /* IPS mode: Deny writing to monitored files. */ 238 if (veriexec_strict >= 2) { 239 error = EPERM; 240 goto bad; 241 } else { 242 vhe->status = FINGERPRINT_NOTEVAL; 243 } 244 } 245 #endif 246 } 247 } 248 249 if (fmode & O_TRUNC) { 250 VOP_UNLOCK(vp, 0); /* XXX */ 251 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 252 vrele(vp); 253 return (error); 254 } 255 VOP_LEASE(vp, p, cred, LEASE_WRITE); 256 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 257 VATTR_NULL(&va); 258 va.va_size = 0; 259 error = VOP_SETATTR(vp, &va, cred, p); 260 vn_finished_write(mp, 0); 261 if (error != 0) 262 goto bad; 263 } 264 if ((error = VOP_OPEN(vp, fmode, cred, p)) != 0) 265 goto bad; 266 if (vp->v_type == VREG && 267 uvn_attach(vp, fmode & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 268 error = EIO; 269 goto bad; 270 } 271 if (fmode & FWRITE) 272 vp->v_writecount++; 273 274 return (0); 275 bad: 276 vput(vp); 277 return (error); 278 } 279 280 /* 281 * Check for write permissions on the specified vnode. 282 * Prototype text segments cannot be written. 283 */ 284 int 285 vn_writechk(vp) 286 struct vnode *vp; 287 { 288 289 /* 290 * If the vnode is in use as a process's text, 291 * we can't allow writing. 292 */ 293 if (vp->v_flag & VTEXT) 294 return (ETXTBSY); 295 return (0); 296 } 297 298 /* 299 * Mark a vnode as having executable mappings. 300 */ 301 void 302 vn_markexec(vp) 303 struct vnode *vp; 304 { 305 if ((vp->v_flag & VEXECMAP) == 0) { 306 uvmexp.filepages -= vp->v_uobj.uo_npages; 307 uvmexp.execpages += vp->v_uobj.uo_npages; 308 } 309 vp->v_flag |= VEXECMAP; 310 } 311 312 /* 313 * Mark a vnode as being the text of a process. 314 * Fail if the vnode is currently writable. 315 */ 316 int 317 vn_marktext(vp) 318 struct vnode *vp; 319 { 320 321 if (vp->v_writecount != 0) { 322 KASSERT((vp->v_flag & VTEXT) == 0); 323 return (ETXTBSY); 324 } 325 vp->v_flag |= VTEXT; 326 vn_markexec(vp); 327 return (0); 328 } 329 330 /* 331 * Vnode close call 332 * 333 * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node. 334 */ 335 int 336 vn_close(vp, flags, cred, p) 337 struct vnode *vp; 338 int flags; 339 struct ucred *cred; 340 struct proc *p; 341 { 342 int error; 343 344 if (flags & FWRITE) 345 vp->v_writecount--; 346 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 347 error = VOP_CLOSE(vp, flags, cred, p); 348 vput(vp); 349 return (error); 350 } 351 352 /* 353 * Package up an I/O request on a vnode into a uio and do it. 354 */ 355 int 356 vn_rdwr(rw, vp, base, len, offset, segflg, ioflg, cred, aresid, p) 357 enum uio_rw rw; 358 struct vnode *vp; 359 caddr_t base; 360 int len; 361 off_t offset; 362 enum uio_seg segflg; 363 int ioflg; 364 struct ucred *cred; 365 size_t *aresid; 366 struct proc *p; 367 { 368 struct uio auio; 369 struct iovec aiov; 370 struct mount *mp; 371 int error; 372 373 if ((ioflg & IO_NODELOCKED) == 0) { 374 if (rw == UIO_READ) { 375 vn_lock(vp, LK_SHARED | LK_RETRY); 376 } else /* UIO_WRITE */ { 377 if (vp->v_type != VCHR && 378 (error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) 379 != 0) 380 return (error); 381 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 382 } 383 } 384 auio.uio_iov = &aiov; 385 auio.uio_iovcnt = 1; 386 aiov.iov_base = base; 387 aiov.iov_len = len; 388 auio.uio_resid = len; 389 auio.uio_offset = offset; 390 auio.uio_segflg = segflg; 391 auio.uio_rw = rw; 392 auio.uio_procp = p; 393 if (rw == UIO_READ) { 394 error = VOP_READ(vp, &auio, ioflg, cred); 395 } else { 396 error = VOP_WRITE(vp, &auio, ioflg, cred); 397 } 398 if (aresid) 399 *aresid = auio.uio_resid; 400 else 401 if (auio.uio_resid && error == 0) 402 error = EIO; 403 if ((ioflg & IO_NODELOCKED) == 0) { 404 if (rw == UIO_WRITE) 405 vn_finished_write(mp, 0); 406 VOP_UNLOCK(vp, 0); 407 } 408 return (error); 409 } 410 411 int 412 vn_readdir(fp, buf, segflg, count, done, p, cookies, ncookies) 413 struct file *fp; 414 char *buf; 415 int segflg, *done, *ncookies; 416 u_int count; 417 struct proc *p; 418 off_t **cookies; 419 { 420 struct vnode *vp = (struct vnode *)fp->f_data; 421 struct iovec aiov; 422 struct uio auio; 423 int error, eofflag; 424 425 /* Limit the size on any kernel buffers used by VOP_READDIR */ 426 count = min(MAXBSIZE, count); 427 428 unionread: 429 if (vp->v_type != VDIR) 430 return (EINVAL); 431 aiov.iov_base = buf; 432 aiov.iov_len = count; 433 auio.uio_iov = &aiov; 434 auio.uio_iovcnt = 1; 435 auio.uio_rw = UIO_READ; 436 auio.uio_segflg = segflg; 437 auio.uio_procp = p; 438 auio.uio_resid = count; 439 vn_lock(vp, LK_SHARED | LK_RETRY); 440 auio.uio_offset = fp->f_offset; 441 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies, 442 ncookies); 443 fp->f_offset = auio.uio_offset; 444 VOP_UNLOCK(vp, 0); 445 if (error) 446 return (error); 447 448 #if defined(UNION) || defined(LKM) 449 if (count == auio.uio_resid && vn_union_readdir_hook) { 450 struct vnode *ovp = vp; 451 452 error = (*vn_union_readdir_hook)(&vp, fp, p); 453 if (error) 454 return (error); 455 if (vp != ovp) 456 goto unionread; 457 } 458 #endif /* UNION || LKM */ 459 460 if (count == auio.uio_resid && (vp->v_flag & VROOT) && 461 (vp->v_mount->mnt_flag & MNT_UNION)) { 462 struct vnode *tvp = vp; 463 vp = vp->v_mount->mnt_vnodecovered; 464 VREF(vp); 465 fp->f_data = vp; 466 fp->f_offset = 0; 467 vrele(tvp); 468 goto unionread; 469 } 470 *done = count - auio.uio_resid; 471 return error; 472 } 473 474 /* 475 * File table vnode read routine. 476 */ 477 static int 478 vn_read(fp, offset, uio, cred, flags) 479 struct file *fp; 480 off_t *offset; 481 struct uio *uio; 482 struct ucred *cred; 483 int flags; 484 { 485 struct vnode *vp = (struct vnode *)fp->f_data; 486 int count, error, ioflag = 0; 487 488 VOP_LEASE(vp, uio->uio_procp, cred, LEASE_READ); 489 if (fp->f_flag & FNONBLOCK) 490 ioflag |= IO_NDELAY; 491 if ((fp->f_flag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC)) 492 ioflag |= IO_SYNC; 493 if (fp->f_flag & FALTIO) 494 ioflag |= IO_ALTSEMANTICS; 495 vn_lock(vp, LK_SHARED | LK_RETRY); 496 uio->uio_offset = *offset; 497 count = uio->uio_resid; 498 error = VOP_READ(vp, uio, ioflag, cred); 499 if (flags & FOF_UPDATE_OFFSET) 500 *offset += count - uio->uio_resid; 501 VOP_UNLOCK(vp, 0); 502 return (error); 503 } 504 505 /* 506 * File table vnode write routine. 507 */ 508 static int 509 vn_write(fp, offset, uio, cred, flags) 510 struct file *fp; 511 off_t *offset; 512 struct uio *uio; 513 struct ucred *cred; 514 int flags; 515 { 516 struct vnode *vp = (struct vnode *)fp->f_data; 517 struct mount *mp; 518 int count, error, ioflag = IO_UNIT; 519 520 if (vp->v_type == VREG && (fp->f_flag & O_APPEND)) 521 ioflag |= IO_APPEND; 522 if (fp->f_flag & FNONBLOCK) 523 ioflag |= IO_NDELAY; 524 if (fp->f_flag & FFSYNC || 525 (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS))) 526 ioflag |= IO_SYNC; 527 else if (fp->f_flag & FDSYNC) 528 ioflag |= IO_DSYNC; 529 if (fp->f_flag & FALTIO) 530 ioflag |= IO_ALTSEMANTICS; 531 mp = NULL; 532 if (vp->v_type != VCHR && 533 (error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 534 return (error); 535 VOP_LEASE(vp, uio->uio_procp, cred, LEASE_WRITE); 536 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 537 uio->uio_offset = *offset; 538 count = uio->uio_resid; 539 error = VOP_WRITE(vp, uio, ioflag, cred); 540 if (flags & FOF_UPDATE_OFFSET) { 541 if (ioflag & IO_APPEND) 542 *offset = uio->uio_offset; 543 else 544 *offset += count - uio->uio_resid; 545 } 546 VOP_UNLOCK(vp, 0); 547 vn_finished_write(mp, 0); 548 return (error); 549 } 550 551 /* 552 * File table vnode stat routine. 553 */ 554 static int 555 vn_statfile(fp, sb, p) 556 struct file *fp; 557 struct stat *sb; 558 struct proc *p; 559 { 560 struct vnode *vp = (struct vnode *)fp->f_data; 561 562 return vn_stat(vp, sb, p); 563 } 564 565 int 566 vn_stat(vp, sb, p) 567 struct vnode *vp; 568 struct stat *sb; 569 struct proc *p; 570 { 571 struct vattr va; 572 int error; 573 mode_t mode; 574 575 error = VOP_GETATTR(vp, &va, p->p_ucred, p); 576 if (error) 577 return (error); 578 /* 579 * Copy from vattr table 580 */ 581 sb->st_dev = va.va_fsid; 582 sb->st_ino = va.va_fileid; 583 mode = va.va_mode; 584 switch (vp->v_type) { 585 case VREG: 586 mode |= S_IFREG; 587 break; 588 case VDIR: 589 mode |= S_IFDIR; 590 break; 591 case VBLK: 592 mode |= S_IFBLK; 593 break; 594 case VCHR: 595 mode |= S_IFCHR; 596 break; 597 case VLNK: 598 mode |= S_IFLNK; 599 break; 600 case VSOCK: 601 mode |= S_IFSOCK; 602 break; 603 case VFIFO: 604 mode |= S_IFIFO; 605 break; 606 default: 607 return (EBADF); 608 }; 609 sb->st_mode = mode; 610 sb->st_nlink = va.va_nlink; 611 sb->st_uid = va.va_uid; 612 sb->st_gid = va.va_gid; 613 sb->st_rdev = va.va_rdev; 614 sb->st_size = va.va_size; 615 sb->st_atimespec = va.va_atime; 616 sb->st_mtimespec = va.va_mtime; 617 sb->st_ctimespec = va.va_ctime; 618 sb->st_birthtimespec = va.va_birthtime; 619 sb->st_blksize = va.va_blocksize; 620 sb->st_flags = va.va_flags; 621 sb->st_gen = 0; 622 sb->st_blocks = va.va_bytes / S_BLKSIZE; 623 return (0); 624 } 625 626 /* 627 * File table vnode fcntl routine. 628 */ 629 static int 630 vn_fcntl(fp, com, data, p) 631 struct file *fp; 632 u_int com; 633 void *data; 634 struct proc *p; 635 { 636 struct vnode *vp = ((struct vnode *)fp->f_data); 637 int error; 638 639 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 640 error = VOP_FCNTL(vp, com, data, fp->f_flag, p->p_ucred, p); 641 VOP_UNLOCK(vp, 0); 642 return (error); 643 } 644 645 /* 646 * File table vnode ioctl routine. 647 */ 648 static int 649 vn_ioctl(fp, com, data, p) 650 struct file *fp; 651 u_long com; 652 void *data; 653 struct proc *p; 654 { 655 struct vnode *vp = ((struct vnode *)fp->f_data); 656 struct vattr vattr; 657 int error; 658 659 switch (vp->v_type) { 660 661 case VREG: 662 case VDIR: 663 if (com == FIONREAD) { 664 error = VOP_GETATTR(vp, &vattr, p->p_ucred, p); 665 if (error) 666 return (error); 667 *(int *)data = vattr.va_size - fp->f_offset; 668 return (0); 669 } 670 if ((com == FIONWRITE) || (com == FIONSPACE)) { 671 /* 672 * Files don't have send queues, so there never 673 * are any bytes in them, nor is there any 674 * open space in them. 675 */ 676 *(int *)data = 0; 677 return (0); 678 } 679 if (com == FIOGETBMAP) { 680 daddr_t *block; 681 682 if (*(daddr_t *)data < 0) 683 return (EINVAL); 684 block = (daddr_t *)data; 685 return (VOP_BMAP(vp, *block, NULL, block, NULL)); 686 } 687 if (com == OFIOGETBMAP) { 688 daddr_t ibn, obn; 689 690 if (*(int32_t *)data < 0) 691 return (EINVAL); 692 ibn = (daddr_t)*(int32_t *)data; 693 error = VOP_BMAP(vp, ibn, NULL, &obn, NULL); 694 *(int32_t *)data = (int32_t)obn; 695 return error; 696 } 697 if (com == FIONBIO || com == FIOASYNC) /* XXX */ 698 return (0); /* XXX */ 699 /* fall into ... */ 700 case VFIFO: 701 case VCHR: 702 case VBLK: 703 error = VOP_IOCTL(vp, com, data, fp->f_flag, p->p_ucred, p); 704 if (error == 0 && com == TIOCSCTTY) { 705 if (p->p_session->s_ttyvp) 706 vrele(p->p_session->s_ttyvp); 707 p->p_session->s_ttyvp = vp; 708 VREF(vp); 709 } 710 return (error); 711 712 default: 713 return (EPASSTHROUGH); 714 } 715 } 716 717 /* 718 * File table vnode poll routine. 719 */ 720 static int 721 vn_poll(fp, events, p) 722 struct file *fp; 723 int events; 724 struct proc *p; 725 { 726 727 return (VOP_POLL(((struct vnode *)fp->f_data), events, p)); 728 } 729 730 /* 731 * File table vnode kqfilter routine. 732 */ 733 int 734 vn_kqfilter(fp, kn) 735 struct file *fp; 736 struct knote *kn; 737 { 738 739 return (VOP_KQFILTER((struct vnode *)fp->f_data, kn)); 740 } 741 742 /* 743 * Check that the vnode is still valid, and if so 744 * acquire requested lock. 745 */ 746 int 747 vn_lock(vp, flags) 748 struct vnode *vp; 749 int flags; 750 { 751 int error; 752 753 #if 0 754 KASSERT(vp->v_usecount > 0 || (flags & LK_INTERLOCK) != 0 755 || (vp->v_flag & VONWORKLST) != 0); 756 #endif 757 758 do { 759 if ((flags & LK_INTERLOCK) == 0) 760 simple_lock(&vp->v_interlock); 761 if (vp->v_flag & VXLOCK) { 762 if (flags & LK_NOWAIT) { 763 simple_unlock(&vp->v_interlock); 764 return EBUSY; 765 } 766 vp->v_flag |= VXWANT; 767 ltsleep(vp, PINOD | PNORELOCK, 768 "vn_lock", 0, &vp->v_interlock); 769 error = ENOENT; 770 } else { 771 error = VOP_LOCK(vp, 772 (flags & ~LK_RETRY) | LK_INTERLOCK); 773 if (error == 0 || error == EDEADLK || error == EBUSY) 774 return (error); 775 } 776 flags &= ~LK_INTERLOCK; 777 } while (flags & LK_RETRY); 778 return (error); 779 } 780 781 /* 782 * File table vnode close routine. 783 */ 784 static int 785 vn_closefile(fp, p) 786 struct file *fp; 787 struct proc *p; 788 { 789 790 return (vn_close(((struct vnode *)fp->f_data), fp->f_flag, 791 fp->f_cred, p)); 792 } 793 794 /* 795 * Enable LK_CANRECURSE on lock. Return prior status. 796 */ 797 u_int 798 vn_setrecurse(vp) 799 struct vnode *vp; 800 { 801 struct lock *lkp = &vp->v_lock; 802 u_int retval = lkp->lk_flags & LK_CANRECURSE; 803 804 lkp->lk_flags |= LK_CANRECURSE; 805 return retval; 806 } 807 808 /* 809 * Called when done with locksetrecurse. 810 */ 811 void 812 vn_restorerecurse(vp, flags) 813 struct vnode *vp; 814 u_int flags; 815 { 816 struct lock *lkp = &vp->v_lock; 817 818 lkp->lk_flags &= ~LK_CANRECURSE; 819 lkp->lk_flags |= flags; 820 } 821 822 int 823 vn_cow_establish(struct vnode *vp, 824 int (*func)(void *, struct buf *), void *cookie) 825 { 826 int s; 827 struct spec_cow_entry *e; 828 829 MALLOC(e, struct spec_cow_entry *, sizeof(struct spec_cow_entry), 830 M_DEVBUF, M_WAITOK); 831 e->ce_func = func; 832 e->ce_cookie = cookie; 833 834 SPEC_COW_LOCK(vp->v_specinfo, s); 835 vp->v_spec_cow_req++; 836 while (vp->v_spec_cow_count > 0) 837 ltsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 838 &vp->v_spec_cow_slock); 839 840 SLIST_INSERT_HEAD(&vp->v_spec_cow_head, e, ce_list); 841 842 vp->v_spec_cow_req--; 843 if (vp->v_spec_cow_req == 0) 844 wakeup(&vp->v_spec_cow_req); 845 SPEC_COW_UNLOCK(vp->v_specinfo, s); 846 847 return 0; 848 } 849 850 int 851 vn_cow_disestablish(struct vnode *vp, 852 int (*func)(void *, struct buf *), void *cookie) 853 { 854 int s; 855 struct spec_cow_entry *e; 856 857 SPEC_COW_LOCK(vp->v_specinfo, s); 858 vp->v_spec_cow_req++; 859 while (vp->v_spec_cow_count > 0) 860 ltsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 861 &vp->v_spec_cow_slock); 862 863 SLIST_FOREACH(e, &vp->v_spec_cow_head, ce_list) 864 if (e->ce_func == func && e->ce_cookie == cookie) { 865 SLIST_REMOVE(&vp->v_spec_cow_head, e, 866 spec_cow_entry, ce_list); 867 FREE(e, M_DEVBUF); 868 break; 869 } 870 871 vp->v_spec_cow_req--; 872 if (vp->v_spec_cow_req == 0) 873 wakeup(&vp->v_spec_cow_req); 874 SPEC_COW_UNLOCK(vp->v_specinfo, s); 875 876 return e ? 0 : EINVAL; 877 } 878 879 /* 880 * Simplified in-kernel wrapper calls for extended attribute access. 881 * Both calls pass in a NULL credential, authorizing a "kernel" access. 882 * Set IO_NODELOCKED in ioflg if the vnode is already locked. 883 */ 884 int 885 vn_extattr_get(struct vnode *vp, int ioflg, int attrnamespace, 886 const char *attrname, size_t *buflen, void *buf, struct proc *p) 887 { 888 struct uio auio; 889 struct iovec aiov; 890 int error; 891 892 aiov.iov_len = *buflen; 893 aiov.iov_base = buf; 894 895 auio.uio_iov = &aiov; 896 auio.uio_iovcnt = 1; 897 auio.uio_rw = UIO_READ; 898 auio.uio_segflg = UIO_SYSSPACE; 899 auio.uio_procp = p; 900 auio.uio_offset = 0; 901 auio.uio_resid = *buflen; 902 903 if ((ioflg & IO_NODELOCKED) == 0) 904 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 905 906 error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, NULL, 907 p); 908 909 if ((ioflg & IO_NODELOCKED) == 0) 910 VOP_UNLOCK(vp, 0); 911 912 if (error == 0) 913 *buflen = *buflen - auio.uio_resid; 914 915 return (error); 916 } 917 918 /* 919 * XXX Failure mode if partially written? 920 */ 921 int 922 vn_extattr_set(struct vnode *vp, int ioflg, int attrnamespace, 923 const char *attrname, size_t buflen, const void *buf, struct proc *p) 924 { 925 struct uio auio; 926 struct iovec aiov; 927 struct mount *mp; 928 int error; 929 930 aiov.iov_len = buflen; 931 aiov.iov_base = (caddr_t) buf; /* XXX kills const */ 932 933 auio.uio_iov = &aiov; 934 auio.uio_iovcnt = 1; 935 auio.uio_rw = UIO_WRITE; 936 auio.uio_segflg = UIO_SYSSPACE; 937 auio.uio_procp = p; 938 auio.uio_offset = 0; 939 auio.uio_resid = buflen; 940 941 if ((ioflg & IO_NODELOCKED) == 0) { 942 if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0) 943 return (error); 944 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 945 } 946 947 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, p); 948 949 if ((ioflg & IO_NODELOCKED) == 0) { 950 vn_finished_write(mp, 0); 951 VOP_UNLOCK(vp, 0); 952 } 953 954 return (error); 955 } 956 957 int 958 vn_extattr_rm(struct vnode *vp, int ioflg, int attrnamespace, 959 const char *attrname, struct proc *p) 960 { 961 struct mount *mp; 962 int error; 963 964 if ((ioflg & IO_NODELOCKED) == 0) { 965 if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0) 966 return (error); 967 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 968 } 969 970 error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, NULL, p); 971 if (error == EOPNOTSUPP) 972 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL, 973 NULL, p); 974 975 if ((ioflg & IO_NODELOCKED) == 0) { 976 vn_finished_write(mp, 0); 977 VOP_UNLOCK(vp, 0); 978 } 979 980 return (error); 981 } 982
Indexes created Sat Sep 20 22:09:52 GMT 2025