vfs_vnops.c revision 1.86.2.8
1 /* $NetBSD: vfs_vnops.c,v 1.86.2.8 2005/09/08 21:06:31 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 1.86.2.8 2005/09/08 21:06:31 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 #endif /* VERIFIED_EXEC */ 110 111 restart: 112 if (fmode & O_CREAT) { 113 ndp->ni_cnd.cn_nameiop = CREATE; 114 ndp->ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 115 if ((fmode & O_EXCL) == 0 && 116 ((fmode & O_NOFOLLOW) == 0)) 117 ndp->ni_cnd.cn_flags |= FOLLOW; 118 if ((error = namei(ndp)) != 0) 119 return (error); 120 if (ndp->ni_vp == NULL) { 121 #ifdef VERIFIED_EXEC 122 /* Lockdown mode: Prevent creation of new files. */ 123 if (veriexec_strict >= 3) { 124 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 125 126 printf("Veriexec: vn_open: Preventing " 127 "new file creation in %s.\n", 128 ndp->ni_dirp); 129 130 vp = ndp->ni_dvp; 131 error = EPERM; 132 goto bad; 133 } 134 #endif /* VERIFIED_EXEC */ 135 136 VATTR_NULL(&va); 137 va.va_type = VREG; 138 va.va_mode = cmode; 139 if (fmode & O_EXCL) 140 va.va_vaflags |= VA_EXCLUSIVE; 141 if (vn_start_write(ndp->ni_dvp, &mp, V_NOWAIT) != 0) { 142 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 143 vput(ndp->ni_dvp); 144 if ((error = vn_start_write(NULL, &mp, 145 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 146 return (error); 147 goto restart; 148 } 149 VOP_LEASE(ndp->ni_dvp, p, cred, LEASE_WRITE); 150 error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp, 151 &ndp->ni_cnd, &va); 152 vn_finished_write(mp, 0); 153 if (error) 154 return (error); 155 fmode &= ~O_TRUNC; 156 vp = ndp->ni_vp; 157 } else { 158 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 159 if (ndp->ni_dvp == ndp->ni_vp) 160 vrele(ndp->ni_dvp); 161 else 162 vput(ndp->ni_dvp); 163 ndp->ni_dvp = NULL; 164 vp = ndp->ni_vp; 165 if (fmode & O_EXCL) { 166 error = EEXIST; 167 goto bad; 168 } 169 fmode &= ~O_CREAT; 170 } 171 } else { 172 ndp->ni_cnd.cn_nameiop = LOOKUP; 173 ndp->ni_cnd.cn_flags = LOCKLEAF; 174 if ((fmode & O_NOFOLLOW) == 0) 175 ndp->ni_cnd.cn_flags |= FOLLOW; 176 if ((error = namei(ndp)) != 0) 177 return (error); 178 vp = ndp->ni_vp; 179 } 180 if (vp->v_type == VSOCK) { 181 error = EOPNOTSUPP; 182 goto bad; 183 } 184 if (ndp->ni_vp->v_type == VLNK) { 185 error = EFTYPE; 186 goto bad; 187 } 188 189 #ifdef VERIFIED_EXEC 190 if ((error = VOP_GETATTR(vp, &va, cred, p)) != 0) 191 goto bad; 192 #endif 193 194 if ((fmode & O_CREAT) == 0) { 195 #ifdef VERIFIED_EXEC 196 /* XXX may need pathbuf instead */ 197 if ((error = veriexec_verify(p, vp, &va, ndp->ni_dirp, 198 VERIEXEC_FILE, &vhe)) != 0) 199 goto bad; 200 #endif 201 202 if (fmode & FREAD) { 203 if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0) 204 goto bad; 205 } 206 207 if (fmode & (FWRITE | O_TRUNC)) { 208 if (vp->v_type == VDIR) { 209 error = EISDIR; 210 goto bad; 211 } 212 if ((error = vn_writechk(vp)) != 0 || 213 (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0) 214 goto bad; 215 #ifdef VERIFIED_EXEC 216 if (vhe != NULL) { 217 veriexec_report("Write access request.", 218 ndp->ni_dirp, &va, p, 219 REPORT_NOVERBOSE, 220 REPORT_ALARM, 221 REPORT_NOPANIC); 222 223 /* IPS mode: Deny writing to monitored files. */ 224 if (veriexec_strict >= 2) { 225 error = EPERM; 226 goto bad; 227 } else { 228 vhe->status = FINGERPRINT_NOTEVAL; 229 } 230 } 231 #endif 232 } 233 } 234 235 if (fmode & O_TRUNC) { 236 VOP_UNLOCK(vp, 0); /* XXX */ 237 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 238 vput(vp); 239 return (error); 240 } 241 VOP_LEASE(vp, p, cred, LEASE_WRITE); 242 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 243 VATTR_NULL(&va); 244 va.va_size = 0; 245 error = VOP_SETATTR(vp, &va, cred, p); 246 vn_finished_write(mp, 0); 247 if (error != 0) 248 goto bad; 249 } 250 if ((error = VOP_OPEN(vp, fmode, cred, p)) != 0) 251 goto bad; 252 if (vp->v_type == VREG && 253 uvn_attach(vp, fmode & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 254 error = EIO; 255 goto bad; 256 } 257 if (fmode & FWRITE) 258 vp->v_writecount++; 259 260 return (0); 261 bad: 262 vput(vp); 263 return (error); 264 } 265 266 /* 267 * Check for write permissions on the specified vnode. 268 * Prototype text segments cannot be written. 269 */ 270 int 271 vn_writechk(vp) 272 struct vnode *vp; 273 { 274 275 /* 276 * If the vnode is in use as a process's text, 277 * we can't allow writing. 278 */ 279 if (vp->v_flag & VTEXT) 280 return (ETXTBSY); 281 return (0); 282 } 283 284 /* 285 * Mark a vnode as having executable mappings. 286 */ 287 void 288 vn_markexec(vp) 289 struct vnode *vp; 290 { 291 if ((vp->v_flag & VEXECMAP) == 0) { 292 uvmexp.filepages -= vp->v_uobj.uo_npages; 293 uvmexp.execpages += vp->v_uobj.uo_npages; 294 } 295 vp->v_flag |= VEXECMAP; 296 } 297 298 /* 299 * Mark a vnode as being the text of a process. 300 * Fail if the vnode is currently writable. 301 */ 302 int 303 vn_marktext(vp) 304 struct vnode *vp; 305 { 306 307 if (vp->v_writecount != 0) { 308 KASSERT((vp->v_flag & VTEXT) == 0); 309 return (ETXTBSY); 310 } 311 vp->v_flag |= VTEXT; 312 vn_markexec(vp); 313 return (0); 314 } 315 316 /* 317 * Vnode close call 318 * 319 * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node. 320 */ 321 int 322 vn_close(vp, flags, cred, p) 323 struct vnode *vp; 324 int flags; 325 struct ucred *cred; 326 struct proc *p; 327 { 328 int error; 329 330 if (flags & FWRITE) 331 vp->v_writecount--; 332 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 333 error = VOP_CLOSE(vp, flags, cred, p); 334 vput(vp); 335 return (error); 336 } 337 338 /* 339 * Package up an I/O request on a vnode into a uio and do it. 340 */ 341 int 342 vn_rdwr(rw, vp, base, len, offset, segflg, ioflg, cred, aresid, p) 343 enum uio_rw rw; 344 struct vnode *vp; 345 caddr_t base; 346 int len; 347 off_t offset; 348 enum uio_seg segflg; 349 int ioflg; 350 struct ucred *cred; 351 size_t *aresid; 352 struct proc *p; 353 { 354 struct uio auio; 355 struct iovec aiov; 356 struct mount *mp; 357 int error; 358 359 if ((ioflg & IO_NODELOCKED) == 0) { 360 if (rw == UIO_READ) { 361 vn_lock(vp, LK_SHARED | LK_RETRY); 362 } else /* UIO_WRITE */ { 363 if (vp->v_type != VCHR && 364 (error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) 365 != 0) 366 return (error); 367 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 368 } 369 } 370 auio.uio_iov = &aiov; 371 auio.uio_iovcnt = 1; 372 aiov.iov_base = base; 373 aiov.iov_len = len; 374 auio.uio_resid = len; 375 auio.uio_offset = offset; 376 auio.uio_segflg = segflg; 377 auio.uio_rw = rw; 378 auio.uio_procp = p; 379 if (rw == UIO_READ) { 380 error = VOP_READ(vp, &auio, ioflg, cred); 381 } else { 382 error = VOP_WRITE(vp, &auio, ioflg, cred); 383 } 384 if (aresid) 385 *aresid = auio.uio_resid; 386 else 387 if (auio.uio_resid && error == 0) 388 error = EIO; 389 if ((ioflg & IO_NODELOCKED) == 0) { 390 if (rw == UIO_WRITE) 391 vn_finished_write(mp, 0); 392 VOP_UNLOCK(vp, 0); 393 } 394 return (error); 395 } 396 397 int 398 vn_readdir(fp, buf, segflg, count, done, p, cookies, ncookies) 399 struct file *fp; 400 char *buf; 401 int segflg, *done, *ncookies; 402 u_int count; 403 struct proc *p; 404 off_t **cookies; 405 { 406 struct vnode *vp = (struct vnode *)fp->f_data; 407 struct iovec aiov; 408 struct uio auio; 409 int error, eofflag; 410 411 unionread: 412 if (vp->v_type != VDIR) 413 return (EINVAL); 414 aiov.iov_base = buf; 415 aiov.iov_len = count; 416 auio.uio_iov = &aiov; 417 auio.uio_iovcnt = 1; 418 auio.uio_rw = UIO_READ; 419 auio.uio_segflg = segflg; 420 auio.uio_procp = p; 421 auio.uio_resid = count; 422 vn_lock(vp, LK_SHARED | LK_RETRY); 423 auio.uio_offset = fp->f_offset; 424 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies, 425 ncookies); 426 fp->f_offset = auio.uio_offset; 427 VOP_UNLOCK(vp, 0); 428 if (error) 429 return (error); 430 431 #if defined(UNION) || defined(LKM) 432 if (count == auio.uio_resid && vn_union_readdir_hook) { 433 struct vnode *ovp = vp; 434 435 error = (*vn_union_readdir_hook)(&vp, fp, p); 436 if (error) 437 return (error); 438 if (vp != ovp) 439 goto unionread; 440 } 441 #endif /* UNION || LKM */ 442 443 if (count == auio.uio_resid && (vp->v_flag & VROOT) && 444 (vp->v_mount->mnt_flag & MNT_UNION)) { 445 struct vnode *tvp = vp; 446 vp = vp->v_mount->mnt_vnodecovered; 447 VREF(vp); 448 fp->f_data = vp; 449 fp->f_offset = 0; 450 vrele(tvp); 451 goto unionread; 452 } 453 *done = count - auio.uio_resid; 454 return error; 455 } 456 457 /* 458 * File table vnode read routine. 459 */ 460 static int 461 vn_read(fp, offset, uio, cred, flags) 462 struct file *fp; 463 off_t *offset; 464 struct uio *uio; 465 struct ucred *cred; 466 int flags; 467 { 468 struct vnode *vp = (struct vnode *)fp->f_data; 469 int count, error, ioflag = 0; 470 471 VOP_LEASE(vp, uio->uio_procp, cred, LEASE_READ); 472 if (fp->f_flag & FNONBLOCK) 473 ioflag |= IO_NDELAY; 474 if ((fp->f_flag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC)) 475 ioflag |= IO_SYNC; 476 if (fp->f_flag & FALTIO) 477 ioflag |= IO_ALTSEMANTICS; 478 vn_lock(vp, LK_SHARED | LK_RETRY); 479 uio->uio_offset = *offset; 480 count = uio->uio_resid; 481 error = VOP_READ(vp, uio, ioflag, cred); 482 if (flags & FOF_UPDATE_OFFSET) 483 *offset += count - uio->uio_resid; 484 VOP_UNLOCK(vp, 0); 485 return (error); 486 } 487 488 /* 489 * File table vnode write routine. 490 */ 491 static int 492 vn_write(fp, offset, uio, cred, flags) 493 struct file *fp; 494 off_t *offset; 495 struct uio *uio; 496 struct ucred *cred; 497 int flags; 498 { 499 struct vnode *vp = (struct vnode *)fp->f_data; 500 struct mount *mp; 501 int count, error, ioflag = IO_UNIT; 502 503 if (vp->v_type == VREG && (fp->f_flag & O_APPEND)) 504 ioflag |= IO_APPEND; 505 if (fp->f_flag & FNONBLOCK) 506 ioflag |= IO_NDELAY; 507 if (fp->f_flag & FFSYNC || 508 (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS))) 509 ioflag |= IO_SYNC; 510 else if (fp->f_flag & FDSYNC) 511 ioflag |= IO_DSYNC; 512 if (fp->f_flag & FALTIO) 513 ioflag |= IO_ALTSEMANTICS; 514 mp = NULL; 515 if (vp->v_type != VCHR && 516 (error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 517 return (error); 518 VOP_LEASE(vp, uio->uio_procp, cred, LEASE_WRITE); 519 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 520 uio->uio_offset = *offset; 521 count = uio->uio_resid; 522 error = VOP_WRITE(vp, uio, ioflag, cred); 523 if (flags & FOF_UPDATE_OFFSET) { 524 if (ioflag & IO_APPEND) 525 *offset = uio->uio_offset; 526 else 527 *offset += count - uio->uio_resid; 528 } 529 VOP_UNLOCK(vp, 0); 530 vn_finished_write(mp, 0); 531 return (error); 532 } 533 534 /* 535 * File table vnode stat routine. 536 */ 537 static int 538 vn_statfile(fp, sb, p) 539 struct file *fp; 540 struct stat *sb; 541 struct proc *p; 542 { 543 struct vnode *vp = (struct vnode *)fp->f_data; 544 545 return vn_stat(vp, sb, p); 546 } 547 548 int 549 vn_stat(vp, sb, p) 550 struct vnode *vp; 551 struct stat *sb; 552 struct proc *p; 553 { 554 struct vattr va; 555 int error; 556 mode_t mode; 557 558 error = VOP_GETATTR(vp, &va, p->p_ucred, p); 559 if (error) 560 return (error); 561 /* 562 * Copy from vattr table 563 */ 564 sb->st_dev = va.va_fsid; 565 sb->st_ino = va.va_fileid; 566 mode = va.va_mode; 567 switch (vp->v_type) { 568 case VREG: 569 mode |= S_IFREG; 570 break; 571 case VDIR: 572 mode |= S_IFDIR; 573 break; 574 case VBLK: 575 mode |= S_IFBLK; 576 break; 577 case VCHR: 578 mode |= S_IFCHR; 579 break; 580 case VLNK: 581 mode |= S_IFLNK; 582 break; 583 case VSOCK: 584 mode |= S_IFSOCK; 585 break; 586 case VFIFO: 587 mode |= S_IFIFO; 588 break; 589 default: 590 return (EBADF); 591 }; 592 sb->st_mode = mode; 593 sb->st_nlink = va.va_nlink; 594 sb->st_uid = va.va_uid; 595 sb->st_gid = va.va_gid; 596 sb->st_rdev = va.va_rdev; 597 sb->st_size = va.va_size; 598 sb->st_atimespec = va.va_atime; 599 sb->st_mtimespec = va.va_mtime; 600 sb->st_ctimespec = va.va_ctime; 601 sb->st_birthtimespec = va.va_birthtime; 602 sb->st_blksize = va.va_blocksize; 603 sb->st_flags = va.va_flags; 604 sb->st_gen = 0; 605 sb->st_blocks = va.va_bytes / S_BLKSIZE; 606 return (0); 607 } 608 609 /* 610 * File table vnode fcntl routine. 611 */ 612 static int 613 vn_fcntl(fp, com, data, p) 614 struct file *fp; 615 u_int com; 616 void *data; 617 struct proc *p; 618 { 619 struct vnode *vp = ((struct vnode *)fp->f_data); 620 int error; 621 622 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 623 error = VOP_FCNTL(vp, com, data, fp->f_flag, p->p_ucred, p); 624 VOP_UNLOCK(vp, 0); 625 return (error); 626 } 627 628 /* 629 * File table vnode ioctl routine. 630 */ 631 static int 632 vn_ioctl(fp, com, data, p) 633 struct file *fp; 634 u_long com; 635 void *data; 636 struct proc *p; 637 { 638 struct vnode *vp = ((struct vnode *)fp->f_data); 639 struct vattr vattr; 640 int error; 641 642 switch (vp->v_type) { 643 644 case VREG: 645 case VDIR: 646 if (com == FIONREAD) { 647 error = VOP_GETATTR(vp, &vattr, p->p_ucred, p); 648 if (error) 649 return (error); 650 *(int *)data = vattr.va_size - fp->f_offset; 651 return (0); 652 } 653 if ((com == FIONWRITE) || (com == FIONSPACE)) { 654 /* 655 * Files don't have send queues, so there never 656 * are any bytes in them, nor is there any 657 * open space in them. 658 */ 659 *(int *)data = 0; 660 return (0); 661 } 662 if (com == FIOGETBMAP) { 663 daddr_t *block; 664 665 if (*(daddr_t *)data < 0) 666 return (EINVAL); 667 block = (daddr_t *)data; 668 return (VOP_BMAP(vp, *block, NULL, block, NULL)); 669 } 670 if (com == OFIOGETBMAP) { 671 daddr_t ibn, obn; 672 673 if (*(int32_t *)data < 0) 674 return (EINVAL); 675 ibn = (daddr_t)*(int32_t *)data; 676 error = VOP_BMAP(vp, ibn, NULL, &obn, NULL); 677 *(int32_t *)data = (int32_t)obn; 678 return error; 679 } 680 if (com == FIONBIO || com == FIOASYNC) /* XXX */ 681 return (0); /* XXX */ 682 /* fall into ... */ 683 case VFIFO: 684 case VCHR: 685 case VBLK: 686 error = VOP_IOCTL(vp, com, data, fp->f_flag, p->p_ucred, p); 687 if (error == 0 && com == TIOCSCTTY) { 688 if (p->p_session->s_ttyvp) 689 vrele(p->p_session->s_ttyvp); 690 p->p_session->s_ttyvp = vp; 691 VREF(vp); 692 } 693 return (error); 694 695 default: 696 return (EPASSTHROUGH); 697 } 698 } 699 700 /* 701 * File table vnode poll routine. 702 */ 703 static int 704 vn_poll(fp, events, p) 705 struct file *fp; 706 int events; 707 struct proc *p; 708 { 709 710 return (VOP_POLL(((struct vnode *)fp->f_data), events, p)); 711 } 712 713 /* 714 * File table vnode kqfilter routine. 715 */ 716 int 717 vn_kqfilter(fp, kn) 718 struct file *fp; 719 struct knote *kn; 720 { 721 722 return (VOP_KQFILTER((struct vnode *)fp->f_data, kn)); 723 } 724 725 /* 726 * Check that the vnode is still valid, and if so 727 * acquire requested lock. 728 */ 729 int 730 vn_lock(vp, flags) 731 struct vnode *vp; 732 int flags; 733 { 734 int error; 735 736 #if 0 737 KASSERT(vp->v_usecount > 0 || (flags & LK_INTERLOCK) != 0 738 || (vp->v_flag & VONWORKLST) != 0); 739 #endif 740 741 do { 742 if ((flags & LK_INTERLOCK) == 0) 743 simple_lock(&vp->v_interlock); 744 if (vp->v_flag & VXLOCK) { 745 if (flags & LK_NOWAIT) { 746 simple_unlock(&vp->v_interlock); 747 return EBUSY; 748 } 749 vp->v_flag |= VXWANT; 750 ltsleep(vp, PINOD | PNORELOCK, 751 "vn_lock", 0, &vp->v_interlock); 752 error = ENOENT; 753 } else { 754 error = VOP_LOCK(vp, 755 (flags & ~LK_RETRY) | LK_INTERLOCK); 756 if (error == 0 || error == EDEADLK || error == EBUSY) 757 return (error); 758 } 759 flags &= ~LK_INTERLOCK; 760 } while (flags & LK_RETRY); 761 return (error); 762 } 763 764 /* 765 * File table vnode close routine. 766 */ 767 static int 768 vn_closefile(fp, p) 769 struct file *fp; 770 struct proc *p; 771 { 772 773 return (vn_close(((struct vnode *)fp->f_data), fp->f_flag, 774 fp->f_cred, p)); 775 } 776 777 /* 778 * Enable LK_CANRECURSE on lock. Return prior status. 779 */ 780 u_int 781 vn_setrecurse(vp) 782 struct vnode *vp; 783 { 784 struct lock *lkp = &vp->v_lock; 785 u_int retval = lkp->lk_flags & LK_CANRECURSE; 786 787 lkp->lk_flags |= LK_CANRECURSE; 788 return retval; 789 } 790 791 /* 792 * Called when done with locksetrecurse. 793 */ 794 void 795 vn_restorerecurse(vp, flags) 796 struct vnode *vp; 797 u_int flags; 798 { 799 struct lock *lkp = &vp->v_lock; 800 801 lkp->lk_flags &= ~LK_CANRECURSE; 802 lkp->lk_flags |= flags; 803 } 804 805 int 806 vn_cow_establish(struct vnode *vp, 807 int (*func)(void *, struct buf *), void *cookie) 808 { 809 int s; 810 struct spec_cow_entry *e; 811 812 MALLOC(e, struct spec_cow_entry *, sizeof(struct spec_cow_entry), 813 M_DEVBUF, M_WAITOK); 814 e->ce_func = func; 815 e->ce_cookie = cookie; 816 817 SPEC_COW_LOCK(vp->v_specinfo, s); 818 vp->v_spec_cow_req++; 819 while (vp->v_spec_cow_count > 0) 820 ltsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 821 &vp->v_spec_cow_slock); 822 823 SLIST_INSERT_HEAD(&vp->v_spec_cow_head, e, ce_list); 824 825 vp->v_spec_cow_req--; 826 if (vp->v_spec_cow_req == 0) 827 wakeup(&vp->v_spec_cow_req); 828 SPEC_COW_UNLOCK(vp->v_specinfo, s); 829 830 return 0; 831 } 832 833 int 834 vn_cow_disestablish(struct vnode *vp, 835 int (*func)(void *, struct buf *), void *cookie) 836 { 837 int s; 838 struct spec_cow_entry *e; 839 840 SPEC_COW_LOCK(vp->v_specinfo, s); 841 vp->v_spec_cow_req++; 842 while (vp->v_spec_cow_count > 0) 843 ltsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 844 &vp->v_spec_cow_slock); 845 846 SLIST_FOREACH(e, &vp->v_spec_cow_head, ce_list) 847 if (e->ce_func == func && e->ce_cookie == cookie) { 848 SLIST_REMOVE(&vp->v_spec_cow_head, e, 849 spec_cow_entry, ce_list); 850 FREE(e, M_DEVBUF); 851 break; 852 } 853 854 vp->v_spec_cow_req--; 855 if (vp->v_spec_cow_req == 0) 856 wakeup(&vp->v_spec_cow_req); 857 SPEC_COW_UNLOCK(vp->v_specinfo, s); 858 859 return e ? 0 : EINVAL; 860 } 861 862 /* 863 * Simplified in-kernel wrapper calls for extended attribute access. 864 * Both calls pass in a NULL credential, authorizing a "kernel" access. 865 * Set IO_NODELOCKED in ioflg if the vnode is already locked. 866 */ 867 int 868 vn_extattr_get(struct vnode *vp, int ioflg, int attrnamespace, 869 const char *attrname, size_t *buflen, void *buf, struct proc *p) 870 { 871 struct uio auio; 872 struct iovec aiov; 873 int error; 874 875 aiov.iov_len = *buflen; 876 aiov.iov_base = buf; 877 878 auio.uio_iov = &aiov; 879 auio.uio_iovcnt = 1; 880 auio.uio_rw = UIO_READ; 881 auio.uio_segflg = UIO_SYSSPACE; 882 auio.uio_procp = p; 883 auio.uio_offset = 0; 884 auio.uio_resid = *buflen; 885 886 if ((ioflg & IO_NODELOCKED) == 0) 887 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 888 889 error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, NULL, 890 p); 891 892 if ((ioflg & IO_NODELOCKED) == 0) 893 VOP_UNLOCK(vp, 0); 894 895 if (error == 0) 896 *buflen = *buflen - auio.uio_resid; 897 898 return (error); 899 } 900 901 /* 902 * XXX Failure mode if partially written? 903 */ 904 int 905 vn_extattr_set(struct vnode *vp, int ioflg, int attrnamespace, 906 const char *attrname, size_t buflen, const void *buf, struct proc *p) 907 { 908 struct uio auio; 909 struct iovec aiov; 910 struct mount *mp; 911 int error; 912 913 aiov.iov_len = buflen; 914 aiov.iov_base = (caddr_t) buf; /* XXX kills const */ 915 916 auio.uio_iov = &aiov; 917 auio.uio_iovcnt = 1; 918 auio.uio_rw = UIO_WRITE; 919 auio.uio_segflg = UIO_SYSSPACE; 920 auio.uio_procp = p; 921 auio.uio_offset = 0; 922 auio.uio_resid = buflen; 923 924 if ((ioflg & IO_NODELOCKED) == 0) { 925 if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0) 926 return (error); 927 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 928 } 929 930 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, p); 931 932 if ((ioflg & IO_NODELOCKED) == 0) { 933 vn_finished_write(mp, 0); 934 VOP_UNLOCK(vp, 0); 935 } 936 937 return (error); 938 } 939 940 int 941 vn_extattr_rm(struct vnode *vp, int ioflg, int attrnamespace, 942 const char *attrname, struct proc *p) 943 { 944 struct mount *mp; 945 int error; 946 947 if ((ioflg & IO_NODELOCKED) == 0) { 948 if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0) 949 return (error); 950 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 951 } 952 953 error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, NULL, p); 954 if (error == EOPNOTSUPP) 955 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL, 956 NULL, p); 957 958 if ((ioflg & IO_NODELOCKED) == 0) { 959 vn_finished_write(mp, 0); 960 VOP_UNLOCK(vp, 0); 961 } 962 963 return (error); 964 } 965
Indexes created Sun Sep 21 14:09:52 GMT 2025