vfs_vnops.c revision 1.86.2.2
1 /* $NetBSD: vfs_vnops.c,v 1.86.2.2 2005/06/13 22:09:09 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 1.86.2.2 2005/06/13 22:09:09 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 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 veriexec_report("Write access request.", 216 ndp->ni_dirp, &va, p, 217 REPORT_NOVERBOSE, 218 REPORT_ALARM, 219 REPORT_NOPANIC); 220 221 if (veriexec_strict > 0) { 222 error = EPERM; 223 goto bad; 224 } else { 225 vp->fp_status = FINGERPRINT_NOTEVAL; 226 } 227 } 228 #endif 229 } 230 } 231 232 if (fmode & O_TRUNC) { 233 VOP_UNLOCK(vp, 0); /* XXX */ 234 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 235 vput(vp); 236 return (error); 237 } 238 VOP_LEASE(vp, p, cred, LEASE_WRITE); 239 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 240 VATTR_NULL(&va); 241 va.va_size = 0; 242 error = VOP_SETATTR(vp, &va, cred, p); 243 vn_finished_write(mp, 0); 244 if (error != 0) 245 goto bad; 246 } 247 if ((error = VOP_OPEN(vp, fmode, cred, p)) != 0) 248 goto bad; 249 if (vp->v_type == VREG && 250 uvn_attach(vp, fmode & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 251 error = EIO; 252 goto bad; 253 } 254 if (fmode & FWRITE) 255 vp->v_writecount++; 256 257 return (0); 258 bad: 259 vput(vp); 260 return (error); 261 } 262 263 /* 264 * Check for write permissions on the specified vnode. 265 * Prototype text segments cannot be written. 266 */ 267 int 268 vn_writechk(vp) 269 struct vnode *vp; 270 { 271 272 /* 273 * If the vnode is in use as a process's text, 274 * we can't allow writing. 275 */ 276 if (vp->v_flag & VTEXT) 277 return (ETXTBSY); 278 return (0); 279 } 280 281 /* 282 * Mark a vnode as having executable mappings. 283 */ 284 void 285 vn_markexec(vp) 286 struct vnode *vp; 287 { 288 if ((vp->v_flag & VEXECMAP) == 0) { 289 uvmexp.filepages -= vp->v_uobj.uo_npages; 290 uvmexp.execpages += vp->v_uobj.uo_npages; 291 } 292 vp->v_flag |= VEXECMAP; 293 } 294 295 /* 296 * Mark a vnode as being the text of a process. 297 * Fail if the vnode is currently writable. 298 */ 299 int 300 vn_marktext(vp) 301 struct vnode *vp; 302 { 303 304 if (vp->v_writecount != 0) { 305 KASSERT((vp->v_flag & VTEXT) == 0); 306 return (ETXTBSY); 307 } 308 vp->v_flag |= VTEXT; 309 vn_markexec(vp); 310 return (0); 311 } 312 313 /* 314 * Vnode close call 315 * 316 * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node. 317 */ 318 int 319 vn_close(vp, flags, cred, p) 320 struct vnode *vp; 321 int flags; 322 struct ucred *cred; 323 struct proc *p; 324 { 325 int error; 326 327 if (flags & FWRITE) 328 vp->v_writecount--; 329 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 330 error = VOP_CLOSE(vp, flags, cred, p); 331 vput(vp); 332 return (error); 333 } 334 335 /* 336 * Package up an I/O request on a vnode into a uio and do it. 337 */ 338 int 339 vn_rdwr(rw, vp, base, len, offset, segflg, ioflg, cred, aresid, p) 340 enum uio_rw rw; 341 struct vnode *vp; 342 caddr_t base; 343 int len; 344 off_t offset; 345 enum uio_seg segflg; 346 int ioflg; 347 struct ucred *cred; 348 size_t *aresid; 349 struct proc *p; 350 { 351 struct uio auio; 352 struct iovec aiov; 353 struct mount *mp; 354 int error; 355 356 if ((ioflg & IO_NODELOCKED) == 0) { 357 if (rw == UIO_READ) { 358 vn_lock(vp, LK_SHARED | LK_RETRY); 359 } else /* UIO_WRITE */ { 360 if (vp->v_type != VCHR && 361 (error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) 362 != 0) 363 return (error); 364 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 365 } 366 } 367 auio.uio_iov = &aiov; 368 auio.uio_iovcnt = 1; 369 aiov.iov_base = base; 370 aiov.iov_len = len; 371 auio.uio_resid = len; 372 auio.uio_offset = offset; 373 auio.uio_segflg = segflg; 374 auio.uio_rw = rw; 375 auio.uio_procp = p; 376 if (rw == UIO_READ) { 377 error = VOP_READ(vp, &auio, ioflg, cred); 378 } else { 379 error = VOP_WRITE(vp, &auio, ioflg, cred); 380 } 381 if (aresid) 382 *aresid = auio.uio_resid; 383 else 384 if (auio.uio_resid && error == 0) 385 error = EIO; 386 if ((ioflg & IO_NODELOCKED) == 0) { 387 if (rw == UIO_WRITE) 388 vn_finished_write(mp, 0); 389 VOP_UNLOCK(vp, 0); 390 } 391 return (error); 392 } 393 394 int 395 vn_readdir(fp, buf, segflg, count, done, p, cookies, ncookies) 396 struct file *fp; 397 char *buf; 398 int segflg, *done, *ncookies; 399 u_int count; 400 struct proc *p; 401 off_t **cookies; 402 { 403 struct vnode *vp = (struct vnode *)fp->f_data; 404 struct iovec aiov; 405 struct uio auio; 406 int error, eofflag; 407 408 unionread: 409 if (vp->v_type != VDIR) 410 return (EINVAL); 411 aiov.iov_base = buf; 412 aiov.iov_len = count; 413 auio.uio_iov = &aiov; 414 auio.uio_iovcnt = 1; 415 auio.uio_rw = UIO_READ; 416 auio.uio_segflg = segflg; 417 auio.uio_procp = p; 418 auio.uio_resid = count; 419 vn_lock(vp, LK_SHARED | LK_RETRY); 420 auio.uio_offset = fp->f_offset; 421 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies, 422 ncookies); 423 fp->f_offset = auio.uio_offset; 424 VOP_UNLOCK(vp, 0); 425 if (error) 426 return (error); 427 428 #if defined(UNION) || defined(LKM) 429 if (count == auio.uio_resid && vn_union_readdir_hook) { 430 struct vnode *ovp = vp; 431 432 error = (*vn_union_readdir_hook)(&vp, fp, p); 433 if (error) 434 return (error); 435 if (vp != ovp) 436 goto unionread; 437 } 438 #endif /* UNION || LKM */ 439 440 if (count == auio.uio_resid && (vp->v_flag & VROOT) && 441 (vp->v_mount->mnt_flag & MNT_UNION)) { 442 struct vnode *tvp = vp; 443 vp = vp->v_mount->mnt_vnodecovered; 444 VREF(vp); 445 fp->f_data = vp; 446 fp->f_offset = 0; 447 vrele(tvp); 448 goto unionread; 449 } 450 *done = count - auio.uio_resid; 451 return error; 452 } 453 454 /* 455 * File table vnode read routine. 456 */ 457 static int 458 vn_read(fp, offset, uio, cred, flags) 459 struct file *fp; 460 off_t *offset; 461 struct uio *uio; 462 struct ucred *cred; 463 int flags; 464 { 465 struct vnode *vp = (struct vnode *)fp->f_data; 466 int count, error, ioflag = 0; 467 468 VOP_LEASE(vp, uio->uio_procp, cred, LEASE_READ); 469 if (fp->f_flag & FNONBLOCK) 470 ioflag |= IO_NDELAY; 471 if ((fp->f_flag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC)) 472 ioflag |= IO_SYNC; 473 if (fp->f_flag & FALTIO) 474 ioflag |= IO_ALTSEMANTICS; 475 vn_lock(vp, LK_SHARED | LK_RETRY); 476 uio->uio_offset = *offset; 477 count = uio->uio_resid; 478 error = VOP_READ(vp, uio, ioflag, cred); 479 if (flags & FOF_UPDATE_OFFSET) 480 *offset += count - uio->uio_resid; 481 VOP_UNLOCK(vp, 0); 482 return (error); 483 } 484 485 /* 486 * File table vnode write routine. 487 */ 488 static int 489 vn_write(fp, offset, uio, cred, flags) 490 struct file *fp; 491 off_t *offset; 492 struct uio *uio; 493 struct ucred *cred; 494 int flags; 495 { 496 struct vnode *vp = (struct vnode *)fp->f_data; 497 struct mount *mp; 498 int count, error, ioflag = IO_UNIT; 499 500 if (vp->v_type == VREG && (fp->f_flag & O_APPEND)) 501 ioflag |= IO_APPEND; 502 if (fp->f_flag & FNONBLOCK) 503 ioflag |= IO_NDELAY; 504 if (fp->f_flag & FFSYNC || 505 (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS))) 506 ioflag |= IO_SYNC; 507 else if (fp->f_flag & FDSYNC) 508 ioflag |= IO_DSYNC; 509 if (fp->f_flag & FALTIO) 510 ioflag |= IO_ALTSEMANTICS; 511 mp = NULL; 512 if (vp->v_type != VCHR && 513 (error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 514 return (error); 515 VOP_LEASE(vp, uio->uio_procp, cred, LEASE_WRITE); 516 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 517 uio->uio_offset = *offset; 518 count = uio->uio_resid; 519 error = VOP_WRITE(vp, uio, ioflag, cred); 520 if (flags & FOF_UPDATE_OFFSET) { 521 if (ioflag & IO_APPEND) 522 *offset = uio->uio_offset; 523 else 524 *offset += count - uio->uio_resid; 525 } 526 VOP_UNLOCK(vp, 0); 527 vn_finished_write(mp, 0); 528 return (error); 529 } 530 531 /* 532 * File table vnode stat routine. 533 */ 534 static int 535 vn_statfile(fp, sb, p) 536 struct file *fp; 537 struct stat *sb; 538 struct proc *p; 539 { 540 struct vnode *vp = (struct vnode *)fp->f_data; 541 542 return vn_stat(vp, sb, p); 543 } 544 545 int 546 vn_stat(vp, sb, p) 547 struct vnode *vp; 548 struct stat *sb; 549 struct proc *p; 550 { 551 struct vattr va; 552 int error; 553 mode_t mode; 554 555 error = VOP_GETATTR(vp, &va, p->p_ucred, p); 556 if (error) 557 return (error); 558 /* 559 * Copy from vattr table 560 */ 561 sb->st_dev = va.va_fsid; 562 sb->st_ino = va.va_fileid; 563 mode = va.va_mode; 564 switch (vp->v_type) { 565 case VREG: 566 mode |= S_IFREG; 567 break; 568 case VDIR: 569 mode |= S_IFDIR; 570 break; 571 case VBLK: 572 mode |= S_IFBLK; 573 break; 574 case VCHR: 575 mode |= S_IFCHR; 576 break; 577 case VLNK: 578 mode |= S_IFLNK; 579 break; 580 case VSOCK: 581 mode |= S_IFSOCK; 582 break; 583 case VFIFO: 584 mode |= S_IFIFO; 585 break; 586 default: 587 return (EBADF); 588 }; 589 sb->st_mode = mode; 590 sb->st_nlink = va.va_nlink; 591 sb->st_uid = va.va_uid; 592 sb->st_gid = va.va_gid; 593 sb->st_rdev = va.va_rdev; 594 sb->st_size = va.va_size; 595 sb->st_atimespec = va.va_atime; 596 sb->st_mtimespec = va.va_mtime; 597 sb->st_ctimespec = va.va_ctime; 598 sb->st_birthtimespec = va.va_birthtime; 599 sb->st_blksize = va.va_blocksize; 600 sb->st_flags = va.va_flags; 601 sb->st_gen = 0; 602 sb->st_blocks = va.va_bytes / S_BLKSIZE; 603 return (0); 604 } 605 606 /* 607 * File table vnode fcntl routine. 608 */ 609 static int 610 vn_fcntl(fp, com, data, p) 611 struct file *fp; 612 u_int com; 613 void *data; 614 struct proc *p; 615 { 616 struct vnode *vp = ((struct vnode *)fp->f_data); 617 int error; 618 619 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 620 error = VOP_FCNTL(vp, com, data, fp->f_flag, p->p_ucred, p); 621 VOP_UNLOCK(vp, 0); 622 return (error); 623 } 624 625 /* 626 * File table vnode ioctl routine. 627 */ 628 static int 629 vn_ioctl(fp, com, data, p) 630 struct file *fp; 631 u_long com; 632 void *data; 633 struct proc *p; 634 { 635 struct vnode *vp = ((struct vnode *)fp->f_data); 636 struct vattr vattr; 637 int error; 638 639 switch (vp->v_type) { 640 641 case VREG: 642 case VDIR: 643 if (com == FIONREAD) { 644 error = VOP_GETATTR(vp, &vattr, p->p_ucred, p); 645 if (error) 646 return (error); 647 *(int *)data = vattr.va_size - fp->f_offset; 648 return (0); 649 } 650 if ((com == FIONWRITE) || (com == FIONSPACE)) { 651 /* 652 * Files don't have send queues, so there never 653 * are any bytes in them, nor is there any 654 * open space in them. 655 */ 656 *(int *)data = 0; 657 return (0); 658 } 659 if (com == FIOGETBMAP) { 660 daddr_t *block; 661 662 if (*(daddr_t *)data < 0) 663 return (EINVAL); 664 block = (daddr_t *)data; 665 return (VOP_BMAP(vp, *block, NULL, block, NULL)); 666 } 667 if (com == OFIOGETBMAP) { 668 daddr_t ibn, obn; 669 670 if (*(int32_t *)data < 0) 671 return (EINVAL); 672 ibn = (daddr_t)*(int32_t *)data; 673 error = VOP_BMAP(vp, ibn, NULL, &obn, NULL); 674 *(int32_t *)data = (int32_t)obn; 675 return error; 676 } 677 if (com == FIONBIO || com == FIOASYNC) /* XXX */ 678 return (0); /* XXX */ 679 /* fall into ... */ 680 case VFIFO: 681 case VCHR: 682 case VBLK: 683 error = VOP_IOCTL(vp, com, data, fp->f_flag, p->p_ucred, p); 684 if (error == 0 && com == TIOCSCTTY) { 685 if (p->p_session->s_ttyvp) 686 vrele(p->p_session->s_ttyvp); 687 p->p_session->s_ttyvp = vp; 688 VREF(vp); 689 } 690 return (error); 691 692 default: 693 return (EPASSTHROUGH); 694 } 695 } 696 697 /* 698 * File table vnode poll routine. 699 */ 700 static int 701 vn_poll(fp, events, p) 702 struct file *fp; 703 int events; 704 struct proc *p; 705 { 706 707 return (VOP_POLL(((struct vnode *)fp->f_data), events, p)); 708 } 709 710 /* 711 * File table vnode kqfilter routine. 712 */ 713 int 714 vn_kqfilter(fp, kn) 715 struct file *fp; 716 struct knote *kn; 717 { 718 719 return (VOP_KQFILTER((struct vnode *)fp->f_data, kn)); 720 } 721 722 /* 723 * Check that the vnode is still valid, and if so 724 * acquire requested lock. 725 */ 726 int 727 vn_lock(vp, flags) 728 struct vnode *vp; 729 int flags; 730 { 731 int error; 732 733 #if 0 734 KASSERT(vp->v_usecount > 0 || (flags & LK_INTERLOCK) != 0 735 || (vp->v_flag & VONWORKLST) != 0); 736 #endif 737 738 do { 739 if ((flags & LK_INTERLOCK) == 0) 740 simple_lock(&vp->v_interlock); 741 if (vp->v_flag & VXLOCK) { 742 if (flags & LK_NOWAIT) { 743 simple_unlock(&vp->v_interlock); 744 return EBUSY; 745 } 746 vp->v_flag |= VXWANT; 747 ltsleep(vp, PINOD | PNORELOCK, 748 "vn_lock", 0, &vp->v_interlock); 749 error = ENOENT; 750 } else { 751 error = VOP_LOCK(vp, 752 (flags & ~LK_RETRY) | LK_INTERLOCK); 753 if (error == 0 || error == EDEADLK || error == EBUSY) 754 return (error); 755 } 756 flags &= ~LK_INTERLOCK; 757 } while (flags & LK_RETRY); 758 return (error); 759 } 760 761 /* 762 * File table vnode close routine. 763 */ 764 static int 765 vn_closefile(fp, p) 766 struct file *fp; 767 struct proc *p; 768 { 769 770 return (vn_close(((struct vnode *)fp->f_data), fp->f_flag, 771 fp->f_cred, p)); 772 } 773 774 /* 775 * Enable LK_CANRECURSE on lock. Return prior status. 776 */ 777 u_int 778 vn_setrecurse(vp) 779 struct vnode *vp; 780 { 781 struct lock *lkp = &vp->v_lock; 782 u_int retval = lkp->lk_flags & LK_CANRECURSE; 783 784 lkp->lk_flags |= LK_CANRECURSE; 785 return retval; 786 } 787 788 /* 789 * Called when done with locksetrecurse. 790 */ 791 void 792 vn_restorerecurse(vp, flags) 793 struct vnode *vp; 794 u_int flags; 795 { 796 struct lock *lkp = &vp->v_lock; 797 798 lkp->lk_flags &= ~LK_CANRECURSE; 799 lkp->lk_flags |= flags; 800 } 801 802 int 803 vn_cow_establish(struct vnode *vp, 804 int (*func)(void *, struct buf *), void *cookie) 805 { 806 int s; 807 struct spec_cow_entry *e; 808 809 MALLOC(e, struct spec_cow_entry *, sizeof(struct spec_cow_entry), 810 M_DEVBUF, M_WAITOK); 811 e->ce_func = func; 812 e->ce_cookie = cookie; 813 814 SPEC_COW_LOCK(vp->v_specinfo, s); 815 vp->v_spec_cow_req++; 816 while (vp->v_spec_cow_count > 0) 817 ltsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 818 &vp->v_spec_cow_slock); 819 820 SLIST_INSERT_HEAD(&vp->v_spec_cow_head, e, ce_list); 821 822 vp->v_spec_cow_req--; 823 if (vp->v_spec_cow_req == 0) 824 wakeup(&vp->v_spec_cow_req); 825 SPEC_COW_UNLOCK(vp->v_specinfo, s); 826 827 return 0; 828 } 829 830 int 831 vn_cow_disestablish(struct vnode *vp, 832 int (*func)(void *, struct buf *), void *cookie) 833 { 834 int s; 835 struct spec_cow_entry *e; 836 837 SPEC_COW_LOCK(vp->v_specinfo, s); 838 vp->v_spec_cow_req++; 839 while (vp->v_spec_cow_count > 0) 840 ltsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 841 &vp->v_spec_cow_slock); 842 843 SLIST_FOREACH(e, &vp->v_spec_cow_head, ce_list) 844 if (e->ce_func == func && e->ce_cookie == cookie) { 845 SLIST_REMOVE(&vp->v_spec_cow_head, e, 846 spec_cow_entry, ce_list); 847 FREE(e, M_DEVBUF); 848 break; 849 } 850 851 vp->v_spec_cow_req--; 852 if (vp->v_spec_cow_req == 0) 853 wakeup(&vp->v_spec_cow_req); 854 SPEC_COW_UNLOCK(vp->v_specinfo, s); 855 856 return e ? 0 : EINVAL; 857 } 858 859 /* 860 * Simplified in-kernel wrapper calls for extended attribute access. 861 * Both calls pass in a NULL credential, authorizing a "kernel" access. 862 * Set IO_NODELOCKED in ioflg if the vnode is already locked. 863 */ 864 int 865 vn_extattr_get(struct vnode *vp, int ioflg, int attrnamespace, 866 const char *attrname, size_t *buflen, void *buf, struct proc *p) 867 { 868 struct uio auio; 869 struct iovec aiov; 870 int error; 871 872 aiov.iov_len = *buflen; 873 aiov.iov_base = buf; 874 875 auio.uio_iov = &aiov; 876 auio.uio_iovcnt = 1; 877 auio.uio_rw = UIO_READ; 878 auio.uio_segflg = UIO_SYSSPACE; 879 auio.uio_procp = p; 880 auio.uio_offset = 0; 881 auio.uio_resid = *buflen; 882 883 if ((ioflg & IO_NODELOCKED) == 0) 884 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 885 886 error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, NULL, 887 p); 888 889 if ((ioflg & IO_NODELOCKED) == 0) 890 VOP_UNLOCK(vp, 0); 891 892 if (error == 0) 893 *buflen = *buflen - auio.uio_resid; 894 895 return (error); 896 } 897 898 /* 899 * XXX Failure mode if partially written? 900 */ 901 int 902 vn_extattr_set(struct vnode *vp, int ioflg, int attrnamespace, 903 const char *attrname, size_t buflen, const void *buf, struct proc *p) 904 { 905 struct uio auio; 906 struct iovec aiov; 907 struct mount *mp; 908 int error; 909 910 aiov.iov_len = buflen; 911 aiov.iov_base = (caddr_t) buf; /* XXX kills const */ 912 913 auio.uio_iov = &aiov; 914 auio.uio_iovcnt = 1; 915 auio.uio_rw = UIO_WRITE; 916 auio.uio_segflg = UIO_SYSSPACE; 917 auio.uio_procp = p; 918 auio.uio_offset = 0; 919 auio.uio_resid = buflen; 920 921 if ((ioflg & IO_NODELOCKED) == 0) { 922 if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0) 923 return (error); 924 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 925 } 926 927 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, p); 928 929 if ((ioflg & IO_NODELOCKED) == 0) { 930 vn_finished_write(mp, 0); 931 VOP_UNLOCK(vp, 0); 932 } 933 934 return (error); 935 } 936 937 int 938 vn_extattr_rm(struct vnode *vp, int ioflg, int attrnamespace, 939 const char *attrname, struct proc *p) 940 { 941 struct mount *mp; 942 int error; 943 944 if ((ioflg & IO_NODELOCKED) == 0) { 945 if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0) 946 return (error); 947 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 948 } 949 950 error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, NULL, p); 951 if (error == EOPNOTSUPP) 952 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL, 953 NULL, p); 954 955 if ((ioflg & IO_NODELOCKED) == 0) { 956 vn_finished_write(mp, 0); 957 VOP_UNLOCK(vp, 0); 958 } 959 960 return (error); 961 } 962
Indexes created Sat Sep 20 22:09:52 GMT 2025