vfs_vnops.c revision 1.86.2.5
1 /* $NetBSD: vfs_vnops.c,v 1.86.2.5 2005/07/24 11:06:27 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 1.86.2.5 2005/07/24 11:06:27 tron Exp $"); 41 42 #include "fs_union.h" 43 44 #include <sys/param.h> 45 #include <sys/systm.h> 46 #include <sys/kernel.h> 47 #include <sys/file.h> 48 #include <sys/stat.h> 49 #include <sys/buf.h> 50 #include <sys/proc.h> 51 #include <sys/malloc.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 <miscfs/specfs/specdev.h> 60 61 #include <uvm/uvm_extern.h> 62 63 #ifdef UNION 64 #include <fs/union/union.h> 65 #endif 66 67 #if defined(LKM) || defined(UNION) 68 int (*vn_union_readdir_hook) (struct vnode **, struct file *, struct proc *); 69 #endif 70 71 #ifdef VERIFIED_EXEC 72 #include <sys/verified_exec.h> 73 #endif 74 75 static int vn_read(struct file *fp, off_t *offset, struct uio *uio, 76 struct ucred *cred, int flags); 77 static int vn_write(struct file *fp, off_t *offset, struct uio *uio, 78 struct ucred *cred, int flags); 79 static int vn_closefile(struct file *fp, struct proc *p); 80 static int vn_poll(struct file *fp, int events, struct proc *p); 81 static int vn_fcntl(struct file *fp, u_int com, void *data, struct proc *p); 82 static int vn_statfile(struct file *fp, struct stat *sb, struct proc *p); 83 static int vn_ioctl(struct file *fp, u_long com, void *data, struct proc *p); 84 85 const struct fileops vnops = { 86 vn_read, vn_write, vn_ioctl, vn_fcntl, vn_poll, 87 vn_statfile, vn_closefile, vn_kqfilter 88 }; 89 90 /* 91 * Common code for vnode open operations. 92 * Check permissions, and call the VOP_OPEN or VOP_CREATE routine. 93 */ 94 int 95 vn_open(ndp, fmode, cmode) 96 struct nameidata *ndp; 97 int fmode, cmode; 98 { 99 struct vnode *vp; 100 struct mount *mp; 101 struct proc *p = ndp->ni_cnd.cn_proc; 102 struct ucred *cred = p->p_ucred; 103 struct vattr va; 104 int error; 105 #ifdef VERIFIED_EXEC 106 struct veriexec_hash_entry *vhe = NULL; 107 #endif /* VERIFIED_EXEC */ 108 109 restart: 110 if (fmode & O_CREAT) { 111 ndp->ni_cnd.cn_nameiop = CREATE; 112 ndp->ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 113 if ((fmode & O_EXCL) == 0 && 114 ((fmode & O_NOFOLLOW) == 0)) 115 ndp->ni_cnd.cn_flags |= FOLLOW; 116 if ((error = namei(ndp)) != 0) 117 return (error); 118 if (ndp->ni_vp == NULL) { 119 #ifdef VERIFIED_EXEC 120 /* Lockdown mode: Prevent creation of new files. */ 121 if (veriexec_strict >= 3) { 122 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 123 124 printf("Veriexec: vn_open: Preventing " 125 "new file creation in %s.\n", 126 ndp->ni_dirp); 127 128 vp = ndp->ni_dvp; 129 error = EPERM; 130 goto bad; 131 } 132 #endif /* VERIFIED_EXEC */ 133 134 VATTR_NULL(&va); 135 va.va_type = VREG; 136 va.va_mode = cmode; 137 if (fmode & O_EXCL) 138 va.va_vaflags |= VA_EXCLUSIVE; 139 if (vn_start_write(ndp->ni_dvp, &mp, V_NOWAIT) != 0) { 140 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 141 vput(ndp->ni_dvp); 142 if ((error = vn_start_write(NULL, &mp, 143 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 144 return (error); 145 goto restart; 146 } 147 VOP_LEASE(ndp->ni_dvp, p, cred, LEASE_WRITE); 148 error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp, 149 &ndp->ni_cnd, &va); 150 vn_finished_write(mp, 0); 151 if (error) 152 return (error); 153 fmode &= ~O_TRUNC; 154 vp = ndp->ni_vp; 155 } else { 156 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 157 if (ndp->ni_dvp == ndp->ni_vp) 158 vrele(ndp->ni_dvp); 159 else 160 vput(ndp->ni_dvp); 161 ndp->ni_dvp = NULL; 162 vp = ndp->ni_vp; 163 if (fmode & O_EXCL) { 164 error = EEXIST; 165 goto bad; 166 } 167 fmode &= ~O_CREAT; 168 } 169 } else { 170 ndp->ni_cnd.cn_nameiop = LOOKUP; 171 ndp->ni_cnd.cn_flags = LOCKLEAF; 172 if ((fmode & O_NOFOLLOW) == 0) 173 ndp->ni_cnd.cn_flags |= FOLLOW; 174 if ((error = namei(ndp)) != 0) 175 return (error); 176 vp = ndp->ni_vp; 177 } 178 if (vp->v_type == VSOCK) { 179 error = EOPNOTSUPP; 180 goto bad; 181 } 182 if (ndp->ni_vp->v_type == VLNK) { 183 error = EFTYPE; 184 goto bad; 185 } 186 187 #ifdef VERIFIED_EXEC 188 if ((error = VOP_GETATTR(vp, &va, cred, p)) != 0) 189 goto bad; 190 #endif 191 192 if ((fmode & O_CREAT) == 0) { 193 #ifdef VERIFIED_EXEC 194 /* XXX may need pathbuf instead */ 195 if ((error = veriexec_verify(p, vp, &va, ndp->ni_dirp, 196 VERIEXEC_FILE, &vhe)) != 0) 197 goto bad; 198 #endif 199 200 if (fmode & FREAD) { 201 if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0) 202 goto bad; 203 } 204 205 if (fmode & (FWRITE | O_TRUNC)) { 206 if (vp->v_type == VDIR) { 207 error = EISDIR; 208 goto bad; 209 } 210 if ((error = vn_writechk(vp)) != 0 || 211 (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0) 212 goto bad; 213 #ifdef VERIFIED_EXEC 214 if (vhe != NULL) { 215 veriexec_report("Write access request.", 216 ndp->ni_dirp, &va, p, 217 REPORT_NOVERBOSE, 218 REPORT_ALARM, 219 REPORT_NOPANIC); 220 221 /* IPS mode: Deny writing to monitored files. */ 222 if (veriexec_strict >= 2) { 223 error = EPERM; 224 goto bad; 225 } else { 226 vhe->status = FINGERPRINT_NOTEVAL; 227 } 228 } 229 #endif 230 } 231 } 232 233 if (fmode & O_TRUNC) { 234 VOP_UNLOCK(vp, 0); /* XXX */ 235 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 236 vput(vp); 237 return (error); 238 } 239 VOP_LEASE(vp, p, cred, LEASE_WRITE); 240 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 241 VATTR_NULL(&va); 242 va.va_size = 0; 243 error = VOP_SETATTR(vp, &va, cred, p); 244 vn_finished_write(mp, 0); 245 if (error != 0) 246 goto bad; 247 } 248 if ((error = VOP_OPEN(vp, fmode, cred, p)) != 0) 249 goto bad; 250 if (vp->v_type == VREG && 251 uvn_attach(vp, fmode & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 252 error = EIO; 253 goto bad; 254 } 255 if (fmode & FWRITE) 256 vp->v_writecount++; 257 258 return (0); 259 bad: 260 vput(vp); 261 return (error); 262 } 263 264 /* 265 * Check for write permissions on the specified vnode. 266 * Prototype text segments cannot be written. 267 */ 268 int 269 vn_writechk(vp) 270 struct vnode *vp; 271 { 272 273 /* 274 * If the vnode is in use as a process's text, 275 * we can't allow writing. 276 */ 277 if (vp->v_flag & VTEXT) 278 return (ETXTBSY); 279 return (0); 280 } 281 282 /* 283 * Mark a vnode as having executable mappings. 284 */ 285 void 286 vn_markexec(vp) 287 struct vnode *vp; 288 { 289 if ((vp->v_flag & VEXECMAP) == 0) { 290 uvmexp.filepages -= vp->v_uobj.uo_npages; 291 uvmexp.execpages += vp->v_uobj.uo_npages; 292 } 293 vp->v_flag |= VEXECMAP; 294 } 295 296 /* 297 * Mark a vnode as being the text of a process. 298 * Fail if the vnode is currently writable. 299 */ 300 int 301 vn_marktext(vp) 302 struct vnode *vp; 303 { 304 305 if (vp->v_writecount != 0) { 306 KASSERT((vp->v_flag & VTEXT) == 0); 307 return (ETXTBSY); 308 } 309 vp->v_flag |= VTEXT; 310 vn_markexec(vp); 311 return (0); 312 } 313 314 /* 315 * Vnode close call 316 * 317 * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node. 318 */ 319 int 320 vn_close(vp, flags, cred, p) 321 struct vnode *vp; 322 int flags; 323 struct ucred *cred; 324 struct proc *p; 325 { 326 int error; 327 328 if (flags & FWRITE) 329 vp->v_writecount--; 330 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 331 error = VOP_CLOSE(vp, flags, cred, p); 332 vput(vp); 333 return (error); 334 } 335 336 /* 337 * Package up an I/O request on a vnode into a uio and do it. 338 */ 339 int 340 vn_rdwr(rw, vp, base, len, offset, segflg, ioflg, cred, aresid, p) 341 enum uio_rw rw; 342 struct vnode *vp; 343 caddr_t base; 344 int len; 345 off_t offset; 346 enum uio_seg segflg; 347 int ioflg; 348 struct ucred *cred; 349 size_t *aresid; 350 struct proc *p; 351 { 352 struct uio auio; 353 struct iovec aiov; 354 struct mount *mp; 355 int error; 356 357 if ((ioflg & IO_NODELOCKED) == 0) { 358 if (rw == UIO_READ) { 359 vn_lock(vp, LK_SHARED | LK_RETRY); 360 } else /* UIO_WRITE */ { 361 if (vp->v_type != VCHR && 362 (error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) 363 != 0) 364 return (error); 365 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 366 } 367 } 368 auio.uio_iov = &aiov; 369 auio.uio_iovcnt = 1; 370 aiov.iov_base = base; 371 aiov.iov_len = len; 372 auio.uio_resid = len; 373 auio.uio_offset = offset; 374 auio.uio_segflg = segflg; 375 auio.uio_rw = rw; 376 auio.uio_procp = p; 377 if (rw == UIO_READ) { 378 error = VOP_READ(vp, &auio, ioflg, cred); 379 } else { 380 error = VOP_WRITE(vp, &auio, ioflg, cred); 381 } 382 if (aresid) 383 *aresid = auio.uio_resid; 384 else 385 if (auio.uio_resid && error == 0) 386 error = EIO; 387 if ((ioflg & IO_NODELOCKED) == 0) { 388 if (rw == UIO_WRITE) 389 vn_finished_write(mp, 0); 390 VOP_UNLOCK(vp, 0); 391 } 392 return (error); 393 } 394 395 int 396 vn_readdir(fp, buf, segflg, count, done, p, cookies, ncookies) 397 struct file *fp; 398 char *buf; 399 int segflg, *done, *ncookies; 400 u_int count; 401 struct proc *p; 402 off_t **cookies; 403 { 404 struct vnode *vp = (struct vnode *)fp->f_data; 405 struct iovec aiov; 406 struct uio auio; 407 int error, eofflag; 408 409 unionread: 410 if (vp->v_type != VDIR) 411 return (EINVAL); 412 aiov.iov_base = buf; 413 aiov.iov_len = count; 414 auio.uio_iov = &aiov; 415 auio.uio_iovcnt = 1; 416 auio.uio_rw = UIO_READ; 417 auio.uio_segflg = segflg; 418 auio.uio_procp = p; 419 auio.uio_resid = count; 420 vn_lock(vp, LK_SHARED | LK_RETRY); 421 auio.uio_offset = fp->f_offset; 422 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies, 423 ncookies); 424 fp->f_offset = auio.uio_offset; 425 VOP_UNLOCK(vp, 0); 426 if (error) 427 return (error); 428 429 #if defined(UNION) || defined(LKM) 430 if (count == auio.uio_resid && vn_union_readdir_hook) { 431 struct vnode *ovp = vp; 432 433 error = (*vn_union_readdir_hook)(&vp, fp, p); 434 if (error) 435 return (error); 436 if (vp != ovp) 437 goto unionread; 438 } 439 #endif /* UNION || LKM */ 440 441 if (count == auio.uio_resid && (vp->v_flag & VROOT) && 442 (vp->v_mount->mnt_flag & MNT_UNION)) { 443 struct vnode *tvp = vp; 444 vp = vp->v_mount->mnt_vnodecovered; 445 VREF(vp); 446 fp->f_data = vp; 447 fp->f_offset = 0; 448 vrele(tvp); 449 goto unionread; 450 } 451 *done = count - auio.uio_resid; 452 return error; 453 } 454 455 /* 456 * File table vnode read routine. 457 */ 458 static int 459 vn_read(fp, offset, uio, cred, flags) 460 struct file *fp; 461 off_t *offset; 462 struct uio *uio; 463 struct ucred *cred; 464 int flags; 465 { 466 struct vnode *vp = (struct vnode *)fp->f_data; 467 int count, error, ioflag = 0; 468 469 VOP_LEASE(vp, uio->uio_procp, cred, LEASE_READ); 470 if (fp->f_flag & FNONBLOCK) 471 ioflag |= IO_NDELAY; 472 if ((fp->f_flag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC)) 473 ioflag |= IO_SYNC; 474 if (fp->f_flag & FALTIO) 475 ioflag |= IO_ALTSEMANTICS; 476 vn_lock(vp, LK_SHARED | LK_RETRY); 477 uio->uio_offset = *offset; 478 count = uio->uio_resid; 479 error = VOP_READ(vp, uio, ioflag, cred); 480 if (flags & FOF_UPDATE_OFFSET) 481 *offset += count - uio->uio_resid; 482 VOP_UNLOCK(vp, 0); 483 return (error); 484 } 485 486 /* 487 * File table vnode write routine. 488 */ 489 static int 490 vn_write(fp, offset, uio, cred, flags) 491 struct file *fp; 492 off_t *offset; 493 struct uio *uio; 494 struct ucred *cred; 495 int flags; 496 { 497 struct vnode *vp = (struct vnode *)fp->f_data; 498 struct mount *mp; 499 int count, error, ioflag = IO_UNIT; 500 501 if (vp->v_type == VREG && (fp->f_flag & O_APPEND)) 502 ioflag |= IO_APPEND; 503 if (fp->f_flag & FNONBLOCK) 504 ioflag |= IO_NDELAY; 505 if (fp->f_flag & FFSYNC || 506 (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS))) 507 ioflag |= IO_SYNC; 508 else if (fp->f_flag & FDSYNC) 509 ioflag |= IO_DSYNC; 510 if (fp->f_flag & FALTIO) 511 ioflag |= IO_ALTSEMANTICS; 512 mp = NULL; 513 if (vp->v_type != VCHR && 514 (error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 515 return (error); 516 VOP_LEASE(vp, uio->uio_procp, cred, LEASE_WRITE); 517 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 518 uio->uio_offset = *offset; 519 count = uio->uio_resid; 520 error = VOP_WRITE(vp, uio, ioflag, cred); 521 if (flags & FOF_UPDATE_OFFSET) { 522 if (ioflag & IO_APPEND) 523 *offset = uio->uio_offset; 524 else 525 *offset += count - uio->uio_resid; 526 } 527 VOP_UNLOCK(vp, 0); 528 vn_finished_write(mp, 0); 529 return (error); 530 } 531 532 /* 533 * File table vnode stat routine. 534 */ 535 static int 536 vn_statfile(fp, sb, p) 537 struct file *fp; 538 struct stat *sb; 539 struct proc *p; 540 { 541 struct vnode *vp = (struct vnode *)fp->f_data; 542 543 return vn_stat(vp, sb, p); 544 } 545 546 int 547 vn_stat(vp, sb, p) 548 struct vnode *vp; 549 struct stat *sb; 550 struct proc *p; 551 { 552 struct vattr va; 553 int error; 554 mode_t mode; 555 556 error = VOP_GETATTR(vp, &va, p->p_ucred, p); 557 if (error) 558 return (error); 559 /* 560 * Copy from vattr table 561 */ 562 sb->st_dev = va.va_fsid; 563 sb->st_ino = va.va_fileid; 564 mode = va.va_mode; 565 switch (vp->v_type) { 566 case VREG: 567 mode |= S_IFREG; 568 break; 569 case VDIR: 570 mode |= S_IFDIR; 571 break; 572 case VBLK: 573 mode |= S_IFBLK; 574 break; 575 case VCHR: 576 mode |= S_IFCHR; 577 break; 578 case VLNK: 579 mode |= S_IFLNK; 580 break; 581 case VSOCK: 582 mode |= S_IFSOCK; 583 break; 584 case VFIFO: 585 mode |= S_IFIFO; 586 break; 587 default: 588 return (EBADF); 589 }; 590 sb->st_mode = mode; 591 sb->st_nlink = va.va_nlink; 592 sb->st_uid = va.va_uid; 593 sb->st_gid = va.va_gid; 594 sb->st_rdev = va.va_rdev; 595 sb->st_size = va.va_size; 596 sb->st_atimespec = va.va_atime; 597 sb->st_mtimespec = va.va_mtime; 598 sb->st_ctimespec = va.va_ctime; 599 sb->st_birthtimespec = va.va_birthtime; 600 sb->st_blksize = va.va_blocksize; 601 sb->st_flags = va.va_flags; 602 sb->st_gen = 0; 603 sb->st_blocks = va.va_bytes / S_BLKSIZE; 604 return (0); 605 } 606 607 /* 608 * File table vnode fcntl routine. 609 */ 610 static int 611 vn_fcntl(fp, com, data, p) 612 struct file *fp; 613 u_int com; 614 void *data; 615 struct proc *p; 616 { 617 struct vnode *vp = ((struct vnode *)fp->f_data); 618 int error; 619 620 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 621 error = VOP_FCNTL(vp, com, data, fp->f_flag, p->p_ucred, p); 622 VOP_UNLOCK(vp, 0); 623 return (error); 624 } 625 626 /* 627 * File table vnode ioctl routine. 628 */ 629 static int 630 vn_ioctl(fp, com, data, p) 631 struct file *fp; 632 u_long com; 633 void *data; 634 struct proc *p; 635 { 636 struct vnode *vp = ((struct vnode *)fp->f_data); 637 struct vattr vattr; 638 int error; 639 640 switch (vp->v_type) { 641 642 case VREG: 643 case VDIR: 644 if (com == FIONREAD) { 645 error = VOP_GETATTR(vp, &vattr, p->p_ucred, p); 646 if (error) 647 return (error); 648 *(int *)data = vattr.va_size - fp->f_offset; 649 return (0); 650 } 651 if ((com == FIONWRITE) || (com == FIONSPACE)) { 652 /* 653 * Files don't have send queues, so there never 654 * are any bytes in them, nor is there any 655 * open space in them. 656 */ 657 *(int *)data = 0; 658 return (0); 659 } 660 if (com == FIOGETBMAP) { 661 daddr_t *block; 662 663 if (*(daddr_t *)data < 0) 664 return (EINVAL); 665 block = (daddr_t *)data; 666 return (VOP_BMAP(vp, *block, NULL, block, NULL)); 667 } 668 if (com == OFIOGETBMAP) { 669 daddr_t ibn, obn; 670 671 if (*(int32_t *)data < 0) 672 return (EINVAL); 673 ibn = (daddr_t)*(int32_t *)data; 674 error = VOP_BMAP(vp, ibn, NULL, &obn, NULL); 675 *(int32_t *)data = (int32_t)obn; 676 return error; 677 } 678 if (com == FIONBIO || com == FIOASYNC) /* XXX */ 679 return (0); /* XXX */ 680 /* fall into ... */ 681 case VFIFO: 682 case VCHR: 683 case VBLK: 684 error = VOP_IOCTL(vp, com, data, fp->f_flag, p->p_ucred, p); 685 if (error == 0 && com == TIOCSCTTY) { 686 if (p->p_session->s_ttyvp) 687 vrele(p->p_session->s_ttyvp); 688 p->p_session->s_ttyvp = vp; 689 VREF(vp); 690 } 691 return (error); 692 693 default: 694 return (EPASSTHROUGH); 695 } 696 } 697 698 /* 699 * File table vnode poll routine. 700 */ 701 static int 702 vn_poll(fp, events, p) 703 struct file *fp; 704 int events; 705 struct proc *p; 706 { 707 708 return (VOP_POLL(((struct vnode *)fp->f_data), events, p)); 709 } 710 711 /* 712 * File table vnode kqfilter routine. 713 */ 714 int 715 vn_kqfilter(fp, kn) 716 struct file *fp; 717 struct knote *kn; 718 { 719 720 return (VOP_KQFILTER((struct vnode *)fp->f_data, kn)); 721 } 722 723 /* 724 * Check that the vnode is still valid, and if so 725 * acquire requested lock. 726 */ 727 int 728 vn_lock(vp, flags) 729 struct vnode *vp; 730 int flags; 731 { 732 int error; 733 734 #if 0 735 KASSERT(vp->v_usecount > 0 || (flags & LK_INTERLOCK) != 0 736 || (vp->v_flag & VONWORKLST) != 0); 737 #endif 738 739 do { 740 if ((flags & LK_INTERLOCK) == 0) 741 simple_lock(&vp->v_interlock); 742 if (vp->v_flag & VXLOCK) { 743 if (flags & LK_NOWAIT) { 744 simple_unlock(&vp->v_interlock); 745 return EBUSY; 746 } 747 vp->v_flag |= VXWANT; 748 ltsleep(vp, PINOD | PNORELOCK, 749 "vn_lock", 0, &vp->v_interlock); 750 error = ENOENT; 751 } else { 752 error = VOP_LOCK(vp, 753 (flags & ~LK_RETRY) | LK_INTERLOCK); 754 if (error == 0 || error == EDEADLK || error == EBUSY) 755 return (error); 756 } 757 flags &= ~LK_INTERLOCK; 758 } while (flags & LK_RETRY); 759 return (error); 760 } 761 762 /* 763 * File table vnode close routine. 764 */ 765 static int 766 vn_closefile(fp, p) 767 struct file *fp; 768 struct proc *p; 769 { 770 771 return (vn_close(((struct vnode *)fp->f_data), fp->f_flag, 772 fp->f_cred, p)); 773 } 774 775 /* 776 * Enable LK_CANRECURSE on lock. Return prior status. 777 */ 778 u_int 779 vn_setrecurse(vp) 780 struct vnode *vp; 781 { 782 struct lock *lkp = &vp->v_lock; 783 u_int retval = lkp->lk_flags & LK_CANRECURSE; 784 785 lkp->lk_flags |= LK_CANRECURSE; 786 return retval; 787 } 788 789 /* 790 * Called when done with locksetrecurse. 791 */ 792 void 793 vn_restorerecurse(vp, flags) 794 struct vnode *vp; 795 u_int flags; 796 { 797 struct lock *lkp = &vp->v_lock; 798 799 lkp->lk_flags &= ~LK_CANRECURSE; 800 lkp->lk_flags |= flags; 801 } 802 803 int 804 vn_cow_establish(struct vnode *vp, 805 int (*func)(void *, struct buf *), void *cookie) 806 { 807 int s; 808 struct spec_cow_entry *e; 809 810 MALLOC(e, struct spec_cow_entry *, sizeof(struct spec_cow_entry), 811 M_DEVBUF, M_WAITOK); 812 e->ce_func = func; 813 e->ce_cookie = cookie; 814 815 SPEC_COW_LOCK(vp->v_specinfo, s); 816 vp->v_spec_cow_req++; 817 while (vp->v_spec_cow_count > 0) 818 ltsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 819 &vp->v_spec_cow_slock); 820 821 SLIST_INSERT_HEAD(&vp->v_spec_cow_head, e, ce_list); 822 823 vp->v_spec_cow_req--; 824 if (vp->v_spec_cow_req == 0) 825 wakeup(&vp->v_spec_cow_req); 826 SPEC_COW_UNLOCK(vp->v_specinfo, s); 827 828 return 0; 829 } 830 831 int 832 vn_cow_disestablish(struct vnode *vp, 833 int (*func)(void *, struct buf *), void *cookie) 834 { 835 int s; 836 struct spec_cow_entry *e; 837 838 SPEC_COW_LOCK(vp->v_specinfo, s); 839 vp->v_spec_cow_req++; 840 while (vp->v_spec_cow_count > 0) 841 ltsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 842 &vp->v_spec_cow_slock); 843 844 SLIST_FOREACH(e, &vp->v_spec_cow_head, ce_list) 845 if (e->ce_func == func && e->ce_cookie == cookie) { 846 SLIST_REMOVE(&vp->v_spec_cow_head, e, 847 spec_cow_entry, ce_list); 848 FREE(e, M_DEVBUF); 849 break; 850 } 851 852 vp->v_spec_cow_req--; 853 if (vp->v_spec_cow_req == 0) 854 wakeup(&vp->v_spec_cow_req); 855 SPEC_COW_UNLOCK(vp->v_specinfo, s); 856 857 return e ? 0 : EINVAL; 858 } 859 860 /* 861 * Simplified in-kernel wrapper calls for extended attribute access. 862 * Both calls pass in a NULL credential, authorizing a "kernel" access. 863 * Set IO_NODELOCKED in ioflg if the vnode is already locked. 864 */ 865 int 866 vn_extattr_get(struct vnode *vp, int ioflg, int attrnamespace, 867 const char *attrname, size_t *buflen, void *buf, struct proc *p) 868 { 869 struct uio auio; 870 struct iovec aiov; 871 int error; 872 873 aiov.iov_len = *buflen; 874 aiov.iov_base = buf; 875 876 auio.uio_iov = &aiov; 877 auio.uio_iovcnt = 1; 878 auio.uio_rw = UIO_READ; 879 auio.uio_segflg = UIO_SYSSPACE; 880 auio.uio_procp = p; 881 auio.uio_offset = 0; 882 auio.uio_resid = *buflen; 883 884 if ((ioflg & IO_NODELOCKED) == 0) 885 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 886 887 error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, NULL, 888 p); 889 890 if ((ioflg & IO_NODELOCKED) == 0) 891 VOP_UNLOCK(vp, 0); 892 893 if (error == 0) 894 *buflen = *buflen - auio.uio_resid; 895 896 return (error); 897 } 898 899 /* 900 * XXX Failure mode if partially written? 901 */ 902 int 903 vn_extattr_set(struct vnode *vp, int ioflg, int attrnamespace, 904 const char *attrname, size_t buflen, const void *buf, struct proc *p) 905 { 906 struct uio auio; 907 struct iovec aiov; 908 struct mount *mp; 909 int error; 910 911 aiov.iov_len = buflen; 912 aiov.iov_base = (caddr_t) buf; /* XXX kills const */ 913 914 auio.uio_iov = &aiov; 915 auio.uio_iovcnt = 1; 916 auio.uio_rw = UIO_WRITE; 917 auio.uio_segflg = UIO_SYSSPACE; 918 auio.uio_procp = p; 919 auio.uio_offset = 0; 920 auio.uio_resid = buflen; 921 922 if ((ioflg & IO_NODELOCKED) == 0) { 923 if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0) 924 return (error); 925 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 926 } 927 928 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, p); 929 930 if ((ioflg & IO_NODELOCKED) == 0) { 931 vn_finished_write(mp, 0); 932 VOP_UNLOCK(vp, 0); 933 } 934 935 return (error); 936 } 937 938 int 939 vn_extattr_rm(struct vnode *vp, int ioflg, int attrnamespace, 940 const char *attrname, struct proc *p) 941 { 942 struct mount *mp; 943 int error; 944 945 if ((ioflg & IO_NODELOCKED) == 0) { 946 if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0) 947 return (error); 948 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 949 } 950 951 error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, NULL, p); 952 if (error == EOPNOTSUPP) 953 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL, 954 NULL, p); 955 956 if ((ioflg & IO_NODELOCKED) == 0) { 957 vn_finished_write(mp, 0); 958 VOP_UNLOCK(vp, 0); 959 } 960 961 return (error); 962 } 963
Indexes created Sat Sep 20 22:09:52 GMT 2025