vfs_vnops.c revision 1.87
1 /* $NetBSD: vfs_vnops.c,v 1.87 2005/04/20 13:44:46 blymn 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 1.87 2005/04/20 13:44:46 blymn 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 NEVER /* for the moment I am not convinced this is needed since NDINIT should do this lookup...XXXX blymn */ 106 char pathbuf[MAXPATHLEN]; 107 unsigned pathlen; 108 109 if (ndp->ni_segflg == UIO_SYSSPACE) 110 error = copystr(pathbuf, ndp->ni_dirp, MAXPATHLEN, &pathlen); 111 else 112 error = copyinstr(pathbuf, ndp->ni_dirp,MAXPATHLEN, &pathlen); 113 #endif 114 115 restart: 116 if (fmode & O_CREAT) { 117 ndp->ni_cnd.cn_nameiop = CREATE; 118 ndp->ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 119 if ((fmode & O_EXCL) == 0 && 120 ((fmode & O_NOFOLLOW) == 0)) 121 ndp->ni_cnd.cn_flags |= FOLLOW; 122 if ((error = namei(ndp)) != 0) 123 return (error); 124 if (ndp->ni_vp == NULL) { 125 VATTR_NULL(&va); 126 va.va_type = VREG; 127 va.va_mode = cmode; 128 if (fmode & O_EXCL) 129 va.va_vaflags |= VA_EXCLUSIVE; 130 if (vn_start_write(ndp->ni_dvp, &mp, V_NOWAIT) != 0) { 131 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 132 vput(ndp->ni_dvp); 133 if ((error = vn_start_write(NULL, &mp, 134 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 135 return (error); 136 goto restart; 137 } 138 VOP_LEASE(ndp->ni_dvp, p, cred, LEASE_WRITE); 139 error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp, 140 &ndp->ni_cnd, &va); 141 vn_finished_write(mp, 0); 142 if (error) 143 return (error); 144 fmode &= ~O_TRUNC; 145 vp = ndp->ni_vp; 146 } else { 147 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 148 if (ndp->ni_dvp == ndp->ni_vp) 149 vrele(ndp->ni_dvp); 150 else 151 vput(ndp->ni_dvp); 152 ndp->ni_dvp = NULL; 153 vp = ndp->ni_vp; 154 if (fmode & O_EXCL) { 155 error = EEXIST; 156 goto bad; 157 } 158 fmode &= ~O_CREAT; 159 } 160 } else { 161 ndp->ni_cnd.cn_nameiop = LOOKUP; 162 ndp->ni_cnd.cn_flags = LOCKLEAF; 163 if ((fmode & O_NOFOLLOW) == 0) 164 ndp->ni_cnd.cn_flags |= FOLLOW; 165 if ((error = namei(ndp)) != 0) 166 return (error); 167 vp = ndp->ni_vp; 168 } 169 if (vp->v_type == VSOCK) { 170 error = EOPNOTSUPP; 171 goto bad; 172 } 173 if (ndp->ni_vp->v_type == VLNK) { 174 error = EFTYPE; 175 goto bad; 176 } 177 178 #ifdef VERIFIED_EXEC 179 if ((error = VOP_GETATTR(vp, &va, cred, p)) != 0) 180 goto bad; 181 #endif 182 183 if ((fmode & O_CREAT) == 0) { 184 #ifdef VERIFIED_EXEC 185 /* XXX may need pathbuf instead */ 186 if ((vp->v_type == VREG) && 187 ((error = veriexec_verify(p, vp, &va, ndp->ni_dirp, 188 VERIEXEC_FILE)) != 0)) 189 goto bad; 190 #endif 191 if (fmode & FREAD) { 192 if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0) 193 goto bad; 194 195 } 196 197 if (fmode & (FWRITE | O_TRUNC)) { 198 if (vp->v_type == VDIR) { 199 error = EISDIR; 200 goto bad; 201 } 202 if ((error = vn_writechk(vp)) != 0 || 203 (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0) 204 goto bad; 205 #ifdef VERIFIED_EXEC 206 /* 207 * If file has a fingerprint then 208 * deny the write request, otherwise 209 * invalidate the status so we don't 210 * keep checking for the file having 211 * a fingerprint. 212 */ 213 if ((vp->fp_status == FINGERPRINT_VALID) || 214 (vp->fp_status == FINGERPRINT_INDIRECT)) { 215 printf( 216 "writing to fingerprinted file for dev %lu, file %lu\n", 217 va.va_fsid, va.va_fileid); 218 if (securelevel >= 2) { 219 error = EPERM; 220 goto bad; 221 } else { 222 vp->fp_status = FINGERPRINT_NOMATCH; 223 } 224 } 225 #endif 226 } 227 } 228 229 if (fmode & O_TRUNC) { 230 VOP_UNLOCK(vp, 0); /* XXX */ 231 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 232 vput(vp); 233 return (error); 234 } 235 VOP_LEASE(vp, p, cred, LEASE_WRITE); 236 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 237 VATTR_NULL(&va); 238 va.va_size = 0; 239 error = VOP_SETATTR(vp, &va, cred, p); 240 vn_finished_write(mp, 0); 241 if (error != 0) 242 goto bad; 243 } 244 if ((error = VOP_OPEN(vp, fmode, cred, p)) != 0) 245 goto bad; 246 if (vp->v_type == VREG && 247 uvn_attach(vp, fmode & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 248 error = EIO; 249 goto bad; 250 } 251 if (fmode & FWRITE) 252 vp->v_writecount++; 253 254 return (0); 255 bad: 256 vput(vp); 257 return (error); 258 } 259 260 /* 261 * Check for write permissions on the specified vnode. 262 * Prototype text segments cannot be written. 263 */ 264 int 265 vn_writechk(vp) 266 struct vnode *vp; 267 { 268 269 /* 270 * If the vnode is in use as a process's text, 271 * we can't allow writing. 272 */ 273 if (vp->v_flag & VTEXT) 274 return (ETXTBSY); 275 return (0); 276 } 277 278 /* 279 * Mark a vnode as having executable mappings. 280 */ 281 void 282 vn_markexec(vp) 283 struct vnode *vp; 284 { 285 if ((vp->v_flag & VEXECMAP) == 0) { 286 uvmexp.filepages -= vp->v_uobj.uo_npages; 287 uvmexp.execpages += vp->v_uobj.uo_npages; 288 } 289 vp->v_flag |= VEXECMAP; 290 } 291 292 /* 293 * Mark a vnode as being the text of a process. 294 * Fail if the vnode is currently writable. 295 */ 296 int 297 vn_marktext(vp) 298 struct vnode *vp; 299 { 300 301 if (vp->v_writecount != 0) { 302 KASSERT((vp->v_flag & VTEXT) == 0); 303 return (ETXTBSY); 304 } 305 vp->v_flag |= VTEXT; 306 vn_markexec(vp); 307 return (0); 308 } 309 310 /* 311 * Vnode close call 312 * 313 * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node. 314 */ 315 int 316 vn_close(vp, flags, cred, p) 317 struct vnode *vp; 318 int flags; 319 struct ucred *cred; 320 struct proc *p; 321 { 322 int error; 323 324 if (flags & FWRITE) 325 vp->v_writecount--; 326 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 327 error = VOP_CLOSE(vp, flags, cred, p); 328 vput(vp); 329 return (error); 330 } 331 332 /* 333 * Package up an I/O request on a vnode into a uio and do it. 334 */ 335 int 336 vn_rdwr(rw, vp, base, len, offset, segflg, ioflg, cred, aresid, p) 337 enum uio_rw rw; 338 struct vnode *vp; 339 caddr_t base; 340 int len; 341 off_t offset; 342 enum uio_seg segflg; 343 int ioflg; 344 struct ucred *cred; 345 size_t *aresid; 346 struct proc *p; 347 { 348 struct uio auio; 349 struct iovec aiov; 350 struct mount *mp; 351 int error; 352 353 if ((ioflg & IO_NODELOCKED) == 0) { 354 if (rw == UIO_READ) { 355 vn_lock(vp, LK_SHARED | LK_RETRY); 356 } else /* UIO_WRITE */ { 357 if (vp->v_type != VCHR && 358 (error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) 359 != 0) 360 return (error); 361 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 362 } 363 } 364 auio.uio_iov = &aiov; 365 auio.uio_iovcnt = 1; 366 aiov.iov_base = base; 367 aiov.iov_len = len; 368 auio.uio_resid = len; 369 auio.uio_offset = offset; 370 auio.uio_segflg = segflg; 371 auio.uio_rw = rw; 372 auio.uio_procp = p; 373 if (rw == UIO_READ) { 374 error = VOP_READ(vp, &auio, ioflg, cred); 375 } else { 376 error = VOP_WRITE(vp, &auio, ioflg, cred); 377 } 378 if (aresid) 379 *aresid = auio.uio_resid; 380 else 381 if (auio.uio_resid && error == 0) 382 error = EIO; 383 if ((ioflg & IO_NODELOCKED) == 0) { 384 if (rw == UIO_WRITE) 385 vn_finished_write(mp, 0); 386 VOP_UNLOCK(vp, 0); 387 } 388 return (error); 389 } 390 391 int 392 vn_readdir(fp, buf, segflg, count, done, p, cookies, ncookies) 393 struct file *fp; 394 char *buf; 395 int segflg, *done, *ncookies; 396 u_int count; 397 struct proc *p; 398 off_t **cookies; 399 { 400 struct vnode *vp = (struct vnode *)fp->f_data; 401 struct iovec aiov; 402 struct uio auio; 403 int error, eofflag; 404 405 unionread: 406 if (vp->v_type != VDIR) 407 return (EINVAL); 408 aiov.iov_base = buf; 409 aiov.iov_len = count; 410 auio.uio_iov = &aiov; 411 auio.uio_iovcnt = 1; 412 auio.uio_rw = UIO_READ; 413 auio.uio_segflg = segflg; 414 auio.uio_procp = p; 415 auio.uio_resid = count; 416 vn_lock(vp, LK_SHARED | LK_RETRY); 417 auio.uio_offset = fp->f_offset; 418 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies, 419 ncookies); 420 fp->f_offset = auio.uio_offset; 421 VOP_UNLOCK(vp, 0); 422 if (error) 423 return (error); 424 425 #if defined(UNION) || defined(LKM) 426 if (count == auio.uio_resid && vn_union_readdir_hook) { 427 struct vnode *ovp = vp; 428 429 error = (*vn_union_readdir_hook)(&vp, fp, p); 430 if (error) 431 return (error); 432 if (vp != ovp) 433 goto unionread; 434 } 435 #endif /* UNION || LKM */ 436 437 if (count == auio.uio_resid && (vp->v_flag & VROOT) && 438 (vp->v_mount->mnt_flag & MNT_UNION)) { 439 struct vnode *tvp = vp; 440 vp = vp->v_mount->mnt_vnodecovered; 441 VREF(vp); 442 fp->f_data = vp; 443 fp->f_offset = 0; 444 vrele(tvp); 445 goto unionread; 446 } 447 *done = count - auio.uio_resid; 448 return error; 449 } 450 451 /* 452 * File table vnode read routine. 453 */ 454 static int 455 vn_read(fp, offset, uio, cred, flags) 456 struct file *fp; 457 off_t *offset; 458 struct uio *uio; 459 struct ucred *cred; 460 int flags; 461 { 462 struct vnode *vp = (struct vnode *)fp->f_data; 463 int count, error, ioflag = 0; 464 465 VOP_LEASE(vp, uio->uio_procp, cred, LEASE_READ); 466 if (fp->f_flag & FNONBLOCK) 467 ioflag |= IO_NDELAY; 468 if ((fp->f_flag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC)) 469 ioflag |= IO_SYNC; 470 if (fp->f_flag & FALTIO) 471 ioflag |= IO_ALTSEMANTICS; 472 vn_lock(vp, LK_SHARED | LK_RETRY); 473 uio->uio_offset = *offset; 474 count = uio->uio_resid; 475 error = VOP_READ(vp, uio, ioflag, cred); 476 if (flags & FOF_UPDATE_OFFSET) 477 *offset += count - uio->uio_resid; 478 VOP_UNLOCK(vp, 0); 479 return (error); 480 } 481 482 /* 483 * File table vnode write routine. 484 */ 485 static int 486 vn_write(fp, offset, uio, cred, flags) 487 struct file *fp; 488 off_t *offset; 489 struct uio *uio; 490 struct ucred *cred; 491 int flags; 492 { 493 struct vnode *vp = (struct vnode *)fp->f_data; 494 struct mount *mp; 495 int count, error, ioflag = IO_UNIT; 496 497 if (vp->v_type == VREG && (fp->f_flag & O_APPEND)) 498 ioflag |= IO_APPEND; 499 if (fp->f_flag & FNONBLOCK) 500 ioflag |= IO_NDELAY; 501 if (fp->f_flag & FFSYNC || 502 (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS))) 503 ioflag |= IO_SYNC; 504 else if (fp->f_flag & FDSYNC) 505 ioflag |= IO_DSYNC; 506 if (fp->f_flag & FALTIO) 507 ioflag |= IO_ALTSEMANTICS; 508 mp = NULL; 509 if (vp->v_type != VCHR && 510 (error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 511 return (error); 512 VOP_LEASE(vp, uio->uio_procp, cred, LEASE_WRITE); 513 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 514 uio->uio_offset = *offset; 515 count = uio->uio_resid; 516 error = VOP_WRITE(vp, uio, ioflag, cred); 517 if (flags & FOF_UPDATE_OFFSET) { 518 if (ioflag & IO_APPEND) 519 *offset = uio->uio_offset; 520 else 521 *offset += count - uio->uio_resid; 522 } 523 VOP_UNLOCK(vp, 0); 524 vn_finished_write(mp, 0); 525 return (error); 526 } 527 528 /* 529 * File table vnode stat routine. 530 */ 531 static int 532 vn_statfile(fp, sb, p) 533 struct file *fp; 534 struct stat *sb; 535 struct proc *p; 536 { 537 struct vnode *vp = (struct vnode *)fp->f_data; 538 539 return vn_stat(vp, sb, p); 540 } 541 542 int 543 vn_stat(vp, sb, p) 544 struct vnode *vp; 545 struct stat *sb; 546 struct proc *p; 547 { 548 struct vattr va; 549 int error; 550 mode_t mode; 551 552 error = VOP_GETATTR(vp, &va, p->p_ucred, p); 553 if (error) 554 return (error); 555 /* 556 * Copy from vattr table 557 */ 558 sb->st_dev = va.va_fsid; 559 sb->st_ino = va.va_fileid; 560 mode = va.va_mode; 561 switch (vp->v_type) { 562 case VREG: 563 mode |= S_IFREG; 564 break; 565 case VDIR: 566 mode |= S_IFDIR; 567 break; 568 case VBLK: 569 mode |= S_IFBLK; 570 break; 571 case VCHR: 572 mode |= S_IFCHR; 573 break; 574 case VLNK: 575 mode |= S_IFLNK; 576 break; 577 case VSOCK: 578 mode |= S_IFSOCK; 579 break; 580 case VFIFO: 581 mode |= S_IFIFO; 582 break; 583 default: 584 return (EBADF); 585 }; 586 sb->st_mode = mode; 587 sb->st_nlink = va.va_nlink; 588 sb->st_uid = va.va_uid; 589 sb->st_gid = va.va_gid; 590 sb->st_rdev = va.va_rdev; 591 sb->st_size = va.va_size; 592 sb->st_atimespec = va.va_atime; 593 sb->st_mtimespec = va.va_mtime; 594 sb->st_ctimespec = va.va_ctime; 595 sb->st_birthtimespec = va.va_birthtime; 596 sb->st_blksize = va.va_blocksize; 597 sb->st_flags = va.va_flags; 598 sb->st_gen = 0; 599 sb->st_blocks = va.va_bytes / S_BLKSIZE; 600 return (0); 601 } 602 603 /* 604 * File table vnode fcntl routine. 605 */ 606 static int 607 vn_fcntl(fp, com, data, p) 608 struct file *fp; 609 u_int com; 610 void *data; 611 struct proc *p; 612 { 613 struct vnode *vp = ((struct vnode *)fp->f_data); 614 int error; 615 616 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 617 error = VOP_FCNTL(vp, com, data, fp->f_flag, p->p_ucred, p); 618 VOP_UNLOCK(vp, 0); 619 return (error); 620 } 621 622 /* 623 * File table vnode ioctl routine. 624 */ 625 static int 626 vn_ioctl(fp, com, data, p) 627 struct file *fp; 628 u_long com; 629 void *data; 630 struct proc *p; 631 { 632 struct vnode *vp = ((struct vnode *)fp->f_data); 633 struct vattr vattr; 634 int error; 635 636 switch (vp->v_type) { 637 638 case VREG: 639 case VDIR: 640 if (com == FIONREAD) { 641 error = VOP_GETATTR(vp, &vattr, p->p_ucred, p); 642 if (error) 643 return (error); 644 *(int *)data = vattr.va_size - fp->f_offset; 645 return (0); 646 } 647 if ((com == FIONWRITE) || (com == FIONSPACE)) { 648 /* 649 * Files don't have send queues, so there never 650 * are any bytes in them, nor is there any 651 * open space in them. 652 */ 653 *(int *)data = 0; 654 return (0); 655 } 656 if (com == FIOGETBMAP) { 657 daddr_t *block; 658 659 if (*(daddr_t *)data < 0) 660 return (EINVAL); 661 block = (daddr_t *)data; 662 return (VOP_BMAP(vp, *block, NULL, block, NULL)); 663 } 664 if (com == OFIOGETBMAP) { 665 daddr_t ibn, obn; 666 667 if (*(int32_t *)data < 0) 668 return (EINVAL); 669 ibn = (daddr_t)*(int32_t *)data; 670 error = VOP_BMAP(vp, ibn, NULL, &obn, NULL); 671 *(int32_t *)data = (int32_t)obn; 672 return error; 673 } 674 if (com == FIONBIO || com == FIOASYNC) /* XXX */ 675 return (0); /* XXX */ 676 /* fall into ... */ 677 case VFIFO: 678 case VCHR: 679 case VBLK: 680 error = VOP_IOCTL(vp, com, data, fp->f_flag, p->p_ucred, p); 681 if (error == 0 && com == TIOCSCTTY) { 682 if (p->p_session->s_ttyvp) 683 vrele(p->p_session->s_ttyvp); 684 p->p_session->s_ttyvp = vp; 685 VREF(vp); 686 } 687 return (error); 688 689 default: 690 return (EPASSTHROUGH); 691 } 692 } 693 694 /* 695 * File table vnode poll routine. 696 */ 697 static int 698 vn_poll(fp, events, p) 699 struct file *fp; 700 int events; 701 struct proc *p; 702 { 703 704 return (VOP_POLL(((struct vnode *)fp->f_data), events, p)); 705 } 706 707 /* 708 * File table vnode kqfilter routine. 709 */ 710 int 711 vn_kqfilter(fp, kn) 712 struct file *fp; 713 struct knote *kn; 714 { 715 716 return (VOP_KQFILTER((struct vnode *)fp->f_data, kn)); 717 } 718 719 /* 720 * Check that the vnode is still valid, and if so 721 * acquire requested lock. 722 */ 723 int 724 vn_lock(vp, flags) 725 struct vnode *vp; 726 int flags; 727 { 728 int error; 729 730 #if 0 731 KASSERT(vp->v_usecount > 0 || (flags & LK_INTERLOCK) != 0 732 || (vp->v_flag & VONWORKLST) != 0); 733 #endif 734 735 do { 736 if ((flags & LK_INTERLOCK) == 0) 737 simple_lock(&vp->v_interlock); 738 if (vp->v_flag & VXLOCK) { 739 if (flags & LK_NOWAIT) { 740 simple_unlock(&vp->v_interlock); 741 return EBUSY; 742 } 743 vp->v_flag |= VXWANT; 744 ltsleep(vp, PINOD | PNORELOCK, 745 "vn_lock", 0, &vp->v_interlock); 746 error = ENOENT; 747 } else { 748 error = VOP_LOCK(vp, 749 (flags & ~LK_RETRY) | LK_INTERLOCK); 750 if (error == 0 || error == EDEADLK || error == EBUSY) 751 return (error); 752 } 753 flags &= ~LK_INTERLOCK; 754 } while (flags & LK_RETRY); 755 return (error); 756 } 757 758 /* 759 * File table vnode close routine. 760 */ 761 static int 762 vn_closefile(fp, p) 763 struct file *fp; 764 struct proc *p; 765 { 766 767 return (vn_close(((struct vnode *)fp->f_data), fp->f_flag, 768 fp->f_cred, p)); 769 } 770 771 /* 772 * Enable LK_CANRECURSE on lock. Return prior status. 773 */ 774 u_int 775 vn_setrecurse(vp) 776 struct vnode *vp; 777 { 778 struct lock *lkp = &vp->v_lock; 779 u_int retval = lkp->lk_flags & LK_CANRECURSE; 780 781 lkp->lk_flags |= LK_CANRECURSE; 782 return retval; 783 } 784 785 /* 786 * Called when done with locksetrecurse. 787 */ 788 void 789 vn_restorerecurse(vp, flags) 790 struct vnode *vp; 791 u_int flags; 792 { 793 struct lock *lkp = &vp->v_lock; 794 795 lkp->lk_flags &= ~LK_CANRECURSE; 796 lkp->lk_flags |= flags; 797 } 798 799 int 800 vn_cow_establish(struct vnode *vp, 801 int (*func)(void *, struct buf *), void *cookie) 802 { 803 int s; 804 struct spec_cow_entry *e; 805 806 MALLOC(e, struct spec_cow_entry *, sizeof(struct spec_cow_entry), 807 M_DEVBUF, M_WAITOK); 808 e->ce_func = func; 809 e->ce_cookie = cookie; 810 811 SPEC_COW_LOCK(vp->v_specinfo, s); 812 vp->v_spec_cow_req++; 813 while (vp->v_spec_cow_count > 0) 814 ltsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 815 &vp->v_spec_cow_slock); 816 817 SLIST_INSERT_HEAD(&vp->v_spec_cow_head, e, ce_list); 818 819 vp->v_spec_cow_req--; 820 if (vp->v_spec_cow_req == 0) 821 wakeup(&vp->v_spec_cow_req); 822 SPEC_COW_UNLOCK(vp->v_specinfo, s); 823 824 return 0; 825 } 826 827 int 828 vn_cow_disestablish(struct vnode *vp, 829 int (*func)(void *, struct buf *), void *cookie) 830 { 831 int s; 832 struct spec_cow_entry *e; 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_FOREACH(e, &vp->v_spec_cow_head, ce_list) 841 if (e->ce_func == func && e->ce_cookie == cookie) { 842 SLIST_REMOVE(&vp->v_spec_cow_head, e, 843 spec_cow_entry, ce_list); 844 FREE(e, M_DEVBUF); 845 break; 846 } 847 848 vp->v_spec_cow_req--; 849 if (vp->v_spec_cow_req == 0) 850 wakeup(&vp->v_spec_cow_req); 851 SPEC_COW_UNLOCK(vp->v_specinfo, s); 852 853 return e ? 0 : EINVAL; 854 } 855 856 /* 857 * Simplified in-kernel wrapper calls for extended attribute access. 858 * Both calls pass in a NULL credential, authorizing a "kernel" access. 859 * Set IO_NODELOCKED in ioflg if the vnode is already locked. 860 */ 861 int 862 vn_extattr_get(struct vnode *vp, int ioflg, int attrnamespace, 863 const char *attrname, size_t *buflen, void *buf, struct proc *p) 864 { 865 struct uio auio; 866 struct iovec aiov; 867 int error; 868 869 aiov.iov_len = *buflen; 870 aiov.iov_base = buf; 871 872 auio.uio_iov = &aiov; 873 auio.uio_iovcnt = 1; 874 auio.uio_rw = UIO_READ; 875 auio.uio_segflg = UIO_SYSSPACE; 876 auio.uio_procp = p; 877 auio.uio_offset = 0; 878 auio.uio_resid = *buflen; 879 880 if ((ioflg & IO_NODELOCKED) == 0) 881 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 882 883 error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, NULL, 884 p); 885 886 if ((ioflg & IO_NODELOCKED) == 0) 887 VOP_UNLOCK(vp, 0); 888 889 if (error == 0) 890 *buflen = *buflen - auio.uio_resid; 891 892 return (error); 893 } 894 895 /* 896 * XXX Failure mode if partially written? 897 */ 898 int 899 vn_extattr_set(struct vnode *vp, int ioflg, int attrnamespace, 900 const char *attrname, size_t buflen, const void *buf, struct proc *p) 901 { 902 struct uio auio; 903 struct iovec aiov; 904 struct mount *mp; 905 int error; 906 907 aiov.iov_len = buflen; 908 aiov.iov_base = (caddr_t) buf; /* XXX kills const */ 909 910 auio.uio_iov = &aiov; 911 auio.uio_iovcnt = 1; 912 auio.uio_rw = UIO_WRITE; 913 auio.uio_segflg = UIO_SYSSPACE; 914 auio.uio_procp = p; 915 auio.uio_offset = 0; 916 auio.uio_resid = buflen; 917 918 if ((ioflg & IO_NODELOCKED) == 0) { 919 if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0) 920 return (error); 921 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 922 } 923 924 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, p); 925 926 if ((ioflg & IO_NODELOCKED) == 0) { 927 vn_finished_write(mp, 0); 928 VOP_UNLOCK(vp, 0); 929 } 930 931 return (error); 932 } 933 934 int 935 vn_extattr_rm(struct vnode *vp, int ioflg, int attrnamespace, 936 const char *attrname, struct proc *p) 937 { 938 struct mount *mp; 939 int error; 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_DELETEEXTATTR(vp, attrnamespace, attrname, NULL, p); 948 if (error == EOPNOTSUPP) 949 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL, 950 NULL, p); 951 952 if ((ioflg & IO_NODELOCKED) == 0) { 953 vn_finished_write(mp, 0); 954 VOP_UNLOCK(vp, 0); 955 } 956 957 return (error); 958 } 959
Indexes created Sat Sep 20 22:09:52 GMT 2025