vfs_vnops.c revision 1.89
1 /* $NetBSD: vfs_vnops.c,v 1.89 2005/06/05 23:47:48 thorpej 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 1.89 2005/06/05 23:47:48 thorpej 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(struct nameidata *ndp, int fmode, int cmode) 96 { 97 struct vnode *vp; 98 struct mount *mp; 99 struct proc *p = ndp->ni_cnd.cn_proc; 100 struct ucred *cred = p->p_ucred; 101 struct vattr va; 102 int error; 103 #ifdef NEVER /* for the moment I am not convinced this is needed since NDINIT should do this lookup...XXXX blymn */ 104 char pathbuf[MAXPATHLEN]; 105 unsigned pathlen; 106 107 if (ndp->ni_segflg == UIO_SYSSPACE) 108 error = copystr(pathbuf, ndp->ni_dirp, MAXPATHLEN, &pathlen); 109 else 110 error = copyinstr(pathbuf, ndp->ni_dirp,MAXPATHLEN, &pathlen); 111 #endif 112 113 restart: 114 if (fmode & O_CREAT) { 115 ndp->ni_cnd.cn_nameiop = CREATE; 116 ndp->ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 117 if ((fmode & O_EXCL) == 0 && 118 ((fmode & O_NOFOLLOW) == 0)) 119 ndp->ni_cnd.cn_flags |= FOLLOW; 120 if ((error = namei(ndp)) != 0) 121 return (error); 122 if (ndp->ni_vp == NULL) { 123 VATTR_NULL(&va); 124 va.va_type = VREG; 125 va.va_mode = cmode; 126 if (fmode & O_EXCL) 127 va.va_vaflags |= VA_EXCLUSIVE; 128 if (vn_start_write(ndp->ni_dvp, &mp, V_NOWAIT) != 0) { 129 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 130 vput(ndp->ni_dvp); 131 if ((error = vn_start_write(NULL, &mp, 132 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 133 return (error); 134 goto restart; 135 } 136 VOP_LEASE(ndp->ni_dvp, p, cred, LEASE_WRITE); 137 error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp, 138 &ndp->ni_cnd, &va); 139 vn_finished_write(mp, 0); 140 if (error) 141 return (error); 142 fmode &= ~O_TRUNC; 143 vp = ndp->ni_vp; 144 } else { 145 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 146 if (ndp->ni_dvp == ndp->ni_vp) 147 vrele(ndp->ni_dvp); 148 else 149 vput(ndp->ni_dvp); 150 ndp->ni_dvp = NULL; 151 vp = ndp->ni_vp; 152 if (fmode & O_EXCL) { 153 error = EEXIST; 154 goto bad; 155 } 156 fmode &= ~O_CREAT; 157 } 158 } else { 159 ndp->ni_cnd.cn_nameiop = LOOKUP; 160 ndp->ni_cnd.cn_flags = LOCKLEAF; 161 if ((fmode & O_NOFOLLOW) == 0) 162 ndp->ni_cnd.cn_flags |= FOLLOW; 163 if ((error = namei(ndp)) != 0) 164 return (error); 165 vp = ndp->ni_vp; 166 } 167 if (vp->v_type == VSOCK) { 168 error = EOPNOTSUPP; 169 goto bad; 170 } 171 if (ndp->ni_vp->v_type == VLNK) { 172 error = EFTYPE; 173 goto bad; 174 } 175 176 #ifdef VERIFIED_EXEC 177 if ((error = VOP_GETATTR(vp, &va, cred, p)) != 0) 178 goto bad; 179 #endif 180 181 if ((fmode & O_CREAT) == 0) { 182 #ifdef VERIFIED_EXEC 183 /* XXX may need pathbuf instead */ 184 if ((vp->v_type == VREG) && 185 ((error = veriexec_verify(p, vp, &va, ndp->ni_dirp, 186 VERIEXEC_FILE)) != 0)) 187 goto bad; 188 #endif 189 if (fmode & FREAD) { 190 if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0) 191 goto bad; 192 193 } 194 195 if (fmode & (FWRITE | O_TRUNC)) { 196 if (vp->v_type == VDIR) { 197 error = EISDIR; 198 goto bad; 199 } 200 if ((error = vn_writechk(vp)) != 0 || 201 (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0) 202 goto bad; 203 #ifdef VERIFIED_EXEC 204 /* 205 * If file has a fingerprint then 206 * deny the write request, otherwise 207 * invalidate the status so we don't 208 * keep checking for the file having 209 * a fingerprint. 210 */ 211 if ((vp->fp_status == FINGERPRINT_VALID) || 212 (vp->fp_status == FINGERPRINT_INDIRECT)) { 213 printf( 214 "writing to fingerprinted file for dev %lu, file %lu\n", 215 va.va_fsid, va.va_fileid); 216 if (securelevel >= 2) { 217 error = EPERM; 218 goto bad; 219 } else { 220 vp->fp_status = FINGERPRINT_NOMATCH; 221 } 222 } 223 #endif 224 } 225 } 226 227 if (fmode & O_TRUNC) { 228 VOP_UNLOCK(vp, 0); /* XXX */ 229 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 230 vput(vp); 231 return (error); 232 } 233 VOP_LEASE(vp, p, cred, LEASE_WRITE); 234 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 235 VATTR_NULL(&va); 236 va.va_size = 0; 237 error = VOP_SETATTR(vp, &va, cred, p); 238 vn_finished_write(mp, 0); 239 if (error != 0) 240 goto bad; 241 } 242 if ((error = VOP_OPEN(vp, fmode, cred, p)) != 0) 243 goto bad; 244 if (vp->v_type == VREG && 245 uvn_attach(vp, fmode & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 246 error = EIO; 247 goto bad; 248 } 249 if (fmode & FWRITE) 250 vp->v_writecount++; 251 252 return (0); 253 bad: 254 vput(vp); 255 return (error); 256 } 257 258 /* 259 * Check for write permissions on the specified vnode. 260 * Prototype text segments cannot be written. 261 */ 262 int 263 vn_writechk(struct vnode *vp) 264 { 265 266 /* 267 * If the vnode is in use as a process's text, 268 * we can't allow writing. 269 */ 270 if (vp->v_flag & VTEXT) 271 return (ETXTBSY); 272 return (0); 273 } 274 275 /* 276 * Mark a vnode as having executable mappings. 277 */ 278 void 279 vn_markexec(struct vnode *vp) 280 { 281 if ((vp->v_flag & VEXECMAP) == 0) { 282 uvmexp.filepages -= vp->v_uobj.uo_npages; 283 uvmexp.execpages += vp->v_uobj.uo_npages; 284 } 285 vp->v_flag |= VEXECMAP; 286 } 287 288 /* 289 * Mark a vnode as being the text of a process. 290 * Fail if the vnode is currently writable. 291 */ 292 int 293 vn_marktext(struct vnode *vp) 294 { 295 296 if (vp->v_writecount != 0) { 297 KASSERT((vp->v_flag & VTEXT) == 0); 298 return (ETXTBSY); 299 } 300 vp->v_flag |= VTEXT; 301 vn_markexec(vp); 302 return (0); 303 } 304 305 /* 306 * Vnode close call 307 * 308 * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node. 309 */ 310 int 311 vn_close(struct vnode *vp, int flags, struct ucred *cred, struct proc *p) 312 { 313 int error; 314 315 if (flags & FWRITE) 316 vp->v_writecount--; 317 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 318 error = VOP_CLOSE(vp, flags, cred, p); 319 vput(vp); 320 return (error); 321 } 322 323 /* 324 * Package up an I/O request on a vnode into a uio and do it. 325 */ 326 int 327 vn_rdwr(enum uio_rw rw, struct vnode *vp, caddr_t base, int len, off_t offset, 328 enum uio_seg segflg, int ioflg, struct ucred *cred, size_t *aresid, 329 struct proc *p) 330 { 331 struct uio auio; 332 struct iovec aiov; 333 struct mount *mp; 334 int error; 335 336 if ((ioflg & IO_NODELOCKED) == 0) { 337 if (rw == UIO_READ) { 338 vn_lock(vp, LK_SHARED | LK_RETRY); 339 } else /* UIO_WRITE */ { 340 if (vp->v_type != VCHR && 341 (error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) 342 != 0) 343 return (error); 344 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 345 } 346 } 347 auio.uio_iov = &aiov; 348 auio.uio_iovcnt = 1; 349 aiov.iov_base = base; 350 aiov.iov_len = len; 351 auio.uio_resid = len; 352 auio.uio_offset = offset; 353 auio.uio_segflg = segflg; 354 auio.uio_rw = rw; 355 auio.uio_procp = p; 356 if (rw == UIO_READ) { 357 error = VOP_READ(vp, &auio, ioflg, cred); 358 } else { 359 error = VOP_WRITE(vp, &auio, ioflg, cred); 360 } 361 if (aresid) 362 *aresid = auio.uio_resid; 363 else 364 if (auio.uio_resid && error == 0) 365 error = EIO; 366 if ((ioflg & IO_NODELOCKED) == 0) { 367 if (rw == UIO_WRITE) 368 vn_finished_write(mp, 0); 369 VOP_UNLOCK(vp, 0); 370 } 371 return (error); 372 } 373 374 int 375 vn_readdir(struct file *fp, char *bf, int segflg, u_int count, int *done, 376 struct proc *p, off_t **cookies, int *ncookies) 377 { 378 struct vnode *vp = (struct vnode *)fp->f_data; 379 struct iovec aiov; 380 struct uio auio; 381 int error, eofflag; 382 383 unionread: 384 if (vp->v_type != VDIR) 385 return (EINVAL); 386 aiov.iov_base = bf; 387 aiov.iov_len = count; 388 auio.uio_iov = &aiov; 389 auio.uio_iovcnt = 1; 390 auio.uio_rw = UIO_READ; 391 auio.uio_segflg = segflg; 392 auio.uio_procp = p; 393 auio.uio_resid = count; 394 vn_lock(vp, LK_SHARED | LK_RETRY); 395 auio.uio_offset = fp->f_offset; 396 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies, 397 ncookies); 398 fp->f_offset = auio.uio_offset; 399 VOP_UNLOCK(vp, 0); 400 if (error) 401 return (error); 402 403 #if defined(UNION) || defined(LKM) 404 if (count == auio.uio_resid && vn_union_readdir_hook) { 405 struct vnode *ovp = vp; 406 407 error = (*vn_union_readdir_hook)(&vp, fp, p); 408 if (error) 409 return (error); 410 if (vp != ovp) 411 goto unionread; 412 } 413 #endif /* UNION || LKM */ 414 415 if (count == auio.uio_resid && (vp->v_flag & VROOT) && 416 (vp->v_mount->mnt_flag & MNT_UNION)) { 417 struct vnode *tvp = vp; 418 vp = vp->v_mount->mnt_vnodecovered; 419 VREF(vp); 420 fp->f_data = vp; 421 fp->f_offset = 0; 422 vrele(tvp); 423 goto unionread; 424 } 425 *done = count - auio.uio_resid; 426 return error; 427 } 428 429 /* 430 * File table vnode read routine. 431 */ 432 static int 433 vn_read(struct file *fp, off_t *offset, struct uio *uio, struct ucred *cred, 434 int flags) 435 { 436 struct vnode *vp = (struct vnode *)fp->f_data; 437 int count, error, ioflag = 0; 438 439 VOP_LEASE(vp, uio->uio_procp, cred, LEASE_READ); 440 if (fp->f_flag & FNONBLOCK) 441 ioflag |= IO_NDELAY; 442 if ((fp->f_flag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC)) 443 ioflag |= IO_SYNC; 444 if (fp->f_flag & FALTIO) 445 ioflag |= IO_ALTSEMANTICS; 446 vn_lock(vp, LK_SHARED | LK_RETRY); 447 uio->uio_offset = *offset; 448 count = uio->uio_resid; 449 error = VOP_READ(vp, uio, ioflag, cred); 450 if (flags & FOF_UPDATE_OFFSET) 451 *offset += count - uio->uio_resid; 452 VOP_UNLOCK(vp, 0); 453 return (error); 454 } 455 456 /* 457 * File table vnode write routine. 458 */ 459 static int 460 vn_write(struct file *fp, off_t *offset, struct uio *uio, struct ucred *cred, 461 int flags) 462 { 463 struct vnode *vp = (struct vnode *)fp->f_data; 464 struct mount *mp; 465 int count, error, ioflag = IO_UNIT; 466 467 if (vp->v_type == VREG && (fp->f_flag & O_APPEND)) 468 ioflag |= IO_APPEND; 469 if (fp->f_flag & FNONBLOCK) 470 ioflag |= IO_NDELAY; 471 if (fp->f_flag & FFSYNC || 472 (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS))) 473 ioflag |= IO_SYNC; 474 else if (fp->f_flag & FDSYNC) 475 ioflag |= IO_DSYNC; 476 if (fp->f_flag & FALTIO) 477 ioflag |= IO_ALTSEMANTICS; 478 mp = NULL; 479 if (vp->v_type != VCHR && 480 (error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 481 return (error); 482 VOP_LEASE(vp, uio->uio_procp, cred, LEASE_WRITE); 483 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 484 uio->uio_offset = *offset; 485 count = uio->uio_resid; 486 error = VOP_WRITE(vp, uio, ioflag, cred); 487 if (flags & FOF_UPDATE_OFFSET) { 488 if (ioflag & IO_APPEND) 489 *offset = uio->uio_offset; 490 else 491 *offset += count - uio->uio_resid; 492 } 493 VOP_UNLOCK(vp, 0); 494 vn_finished_write(mp, 0); 495 return (error); 496 } 497 498 /* 499 * File table vnode stat routine. 500 */ 501 static int 502 vn_statfile(struct file *fp, struct stat *sb, struct proc *p) 503 { 504 struct vnode *vp = (struct vnode *)fp->f_data; 505 506 return vn_stat(vp, sb, p); 507 } 508 509 int 510 vn_stat(struct vnode *vp, struct stat *sb, struct proc *p) 511 { 512 struct vattr va; 513 int error; 514 mode_t mode; 515 516 error = VOP_GETATTR(vp, &va, p->p_ucred, p); 517 if (error) 518 return (error); 519 /* 520 * Copy from vattr table 521 */ 522 sb->st_dev = va.va_fsid; 523 sb->st_ino = va.va_fileid; 524 mode = va.va_mode; 525 switch (vp->v_type) { 526 case VREG: 527 mode |= S_IFREG; 528 break; 529 case VDIR: 530 mode |= S_IFDIR; 531 break; 532 case VBLK: 533 mode |= S_IFBLK; 534 break; 535 case VCHR: 536 mode |= S_IFCHR; 537 break; 538 case VLNK: 539 mode |= S_IFLNK; 540 break; 541 case VSOCK: 542 mode |= S_IFSOCK; 543 break; 544 case VFIFO: 545 mode |= S_IFIFO; 546 break; 547 default: 548 return (EBADF); 549 }; 550 sb->st_mode = mode; 551 sb->st_nlink = va.va_nlink; 552 sb->st_uid = va.va_uid; 553 sb->st_gid = va.va_gid; 554 sb->st_rdev = va.va_rdev; 555 sb->st_size = va.va_size; 556 sb->st_atimespec = va.va_atime; 557 sb->st_mtimespec = va.va_mtime; 558 sb->st_ctimespec = va.va_ctime; 559 sb->st_birthtimespec = va.va_birthtime; 560 sb->st_blksize = va.va_blocksize; 561 sb->st_flags = va.va_flags; 562 sb->st_gen = 0; 563 sb->st_blocks = va.va_bytes / S_BLKSIZE; 564 return (0); 565 } 566 567 /* 568 * File table vnode fcntl routine. 569 */ 570 static int 571 vn_fcntl(struct file *fp, u_int com, void *data, struct proc *p) 572 { 573 struct vnode *vp = ((struct vnode *)fp->f_data); 574 int error; 575 576 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 577 error = VOP_FCNTL(vp, com, data, fp->f_flag, p->p_ucred, p); 578 VOP_UNLOCK(vp, 0); 579 return (error); 580 } 581 582 /* 583 * File table vnode ioctl routine. 584 */ 585 static int 586 vn_ioctl(struct file *fp, u_long com, void *data, struct proc *p) 587 { 588 struct vnode *vp = ((struct vnode *)fp->f_data); 589 struct vattr vattr; 590 int error; 591 592 switch (vp->v_type) { 593 594 case VREG: 595 case VDIR: 596 if (com == FIONREAD) { 597 error = VOP_GETATTR(vp, &vattr, p->p_ucred, p); 598 if (error) 599 return (error); 600 *(int *)data = vattr.va_size - fp->f_offset; 601 return (0); 602 } 603 if ((com == FIONWRITE) || (com == FIONSPACE)) { 604 /* 605 * Files don't have send queues, so there never 606 * are any bytes in them, nor is there any 607 * open space in them. 608 */ 609 *(int *)data = 0; 610 return (0); 611 } 612 if (com == FIOGETBMAP) { 613 daddr_t *block; 614 615 if (*(daddr_t *)data < 0) 616 return (EINVAL); 617 block = (daddr_t *)data; 618 return (VOP_BMAP(vp, *block, NULL, block, NULL)); 619 } 620 if (com == OFIOGETBMAP) { 621 daddr_t ibn, obn; 622 623 if (*(int32_t *)data < 0) 624 return (EINVAL); 625 ibn = (daddr_t)*(int32_t *)data; 626 error = VOP_BMAP(vp, ibn, NULL, &obn, NULL); 627 *(int32_t *)data = (int32_t)obn; 628 return error; 629 } 630 if (com == FIONBIO || com == FIOASYNC) /* XXX */ 631 return (0); /* XXX */ 632 /* fall into ... */ 633 case VFIFO: 634 case VCHR: 635 case VBLK: 636 error = VOP_IOCTL(vp, com, data, fp->f_flag, p->p_ucred, p); 637 if (error == 0 && com == TIOCSCTTY) { 638 if (p->p_session->s_ttyvp) 639 vrele(p->p_session->s_ttyvp); 640 p->p_session->s_ttyvp = vp; 641 VREF(vp); 642 } 643 return (error); 644 645 default: 646 return (EPASSTHROUGH); 647 } 648 } 649 650 /* 651 * File table vnode poll routine. 652 */ 653 static int 654 vn_poll(struct file *fp, int events, struct proc *p) 655 { 656 657 return (VOP_POLL(((struct vnode *)fp->f_data), events, p)); 658 } 659 660 /* 661 * File table vnode kqfilter routine. 662 */ 663 int 664 vn_kqfilter(struct file *fp, struct knote *kn) 665 { 666 667 return (VOP_KQFILTER((struct vnode *)fp->f_data, kn)); 668 } 669 670 /* 671 * Check that the vnode is still valid, and if so 672 * acquire requested lock. 673 */ 674 int 675 vn_lock(struct vnode *vp, int flags) 676 { 677 int error; 678 679 #if 0 680 KASSERT(vp->v_usecount > 0 || (flags & LK_INTERLOCK) != 0 681 || (vp->v_flag & VONWORKLST) != 0); 682 #endif 683 684 do { 685 if ((flags & LK_INTERLOCK) == 0) 686 simple_lock(&vp->v_interlock); 687 if (vp->v_flag & VXLOCK) { 688 if (flags & LK_NOWAIT) { 689 simple_unlock(&vp->v_interlock); 690 return EBUSY; 691 } 692 vp->v_flag |= VXWANT; 693 ltsleep(vp, PINOD | PNORELOCK, 694 "vn_lock", 0, &vp->v_interlock); 695 error = ENOENT; 696 } else { 697 error = VOP_LOCK(vp, 698 (flags & ~LK_RETRY) | LK_INTERLOCK); 699 if (error == 0 || error == EDEADLK || error == EBUSY) 700 return (error); 701 } 702 flags &= ~LK_INTERLOCK; 703 } while (flags & LK_RETRY); 704 return (error); 705 } 706 707 /* 708 * File table vnode close routine. 709 */ 710 static int 711 vn_closefile(struct file *fp, struct proc *p) 712 { 713 714 return (vn_close(((struct vnode *)fp->f_data), fp->f_flag, 715 fp->f_cred, p)); 716 } 717 718 /* 719 * Enable LK_CANRECURSE on lock. Return prior status. 720 */ 721 u_int 722 vn_setrecurse(struct vnode *vp) 723 { 724 struct lock *lkp = &vp->v_lock; 725 u_int retval = lkp->lk_flags & LK_CANRECURSE; 726 727 lkp->lk_flags |= LK_CANRECURSE; 728 return retval; 729 } 730 731 /* 732 * Called when done with locksetrecurse. 733 */ 734 void 735 vn_restorerecurse(struct vnode *vp, u_int flags) 736 { 737 struct lock *lkp = &vp->v_lock; 738 739 lkp->lk_flags &= ~LK_CANRECURSE; 740 lkp->lk_flags |= flags; 741 } 742 743 int 744 vn_cow_establish(struct vnode *vp, 745 int (*func)(void *, struct buf *), void *cookie) 746 { 747 int s; 748 struct spec_cow_entry *e; 749 750 MALLOC(e, struct spec_cow_entry *, sizeof(struct spec_cow_entry), 751 M_DEVBUF, M_WAITOK); 752 e->ce_func = func; 753 e->ce_cookie = cookie; 754 755 SPEC_COW_LOCK(vp->v_specinfo, s); 756 vp->v_spec_cow_req++; 757 while (vp->v_spec_cow_count > 0) 758 ltsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 759 &vp->v_spec_cow_slock); 760 761 SLIST_INSERT_HEAD(&vp->v_spec_cow_head, e, ce_list); 762 763 vp->v_spec_cow_req--; 764 if (vp->v_spec_cow_req == 0) 765 wakeup(&vp->v_spec_cow_req); 766 SPEC_COW_UNLOCK(vp->v_specinfo, s); 767 768 return 0; 769 } 770 771 int 772 vn_cow_disestablish(struct vnode *vp, 773 int (*func)(void *, struct buf *), void *cookie) 774 { 775 int s; 776 struct spec_cow_entry *e; 777 778 SPEC_COW_LOCK(vp->v_specinfo, s); 779 vp->v_spec_cow_req++; 780 while (vp->v_spec_cow_count > 0) 781 ltsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 782 &vp->v_spec_cow_slock); 783 784 SLIST_FOREACH(e, &vp->v_spec_cow_head, ce_list) 785 if (e->ce_func == func && e->ce_cookie == cookie) { 786 SLIST_REMOVE(&vp->v_spec_cow_head, e, 787 spec_cow_entry, ce_list); 788 FREE(e, M_DEVBUF); 789 break; 790 } 791 792 vp->v_spec_cow_req--; 793 if (vp->v_spec_cow_req == 0) 794 wakeup(&vp->v_spec_cow_req); 795 SPEC_COW_UNLOCK(vp->v_specinfo, s); 796 797 return e ? 0 : EINVAL; 798 } 799 800 /* 801 * Simplified in-kernel wrapper calls for extended attribute access. 802 * Both calls pass in a NULL credential, authorizing a "kernel" access. 803 * Set IO_NODELOCKED in ioflg if the vnode is already locked. 804 */ 805 int 806 vn_extattr_get(struct vnode *vp, int ioflg, int attrnamespace, 807 const char *attrname, size_t *buflen, void *bf, struct proc *p) 808 { 809 struct uio auio; 810 struct iovec aiov; 811 int error; 812 813 aiov.iov_len = *buflen; 814 aiov.iov_base = bf; 815 816 auio.uio_iov = &aiov; 817 auio.uio_iovcnt = 1; 818 auio.uio_rw = UIO_READ; 819 auio.uio_segflg = UIO_SYSSPACE; 820 auio.uio_procp = p; 821 auio.uio_offset = 0; 822 auio.uio_resid = *buflen; 823 824 if ((ioflg & IO_NODELOCKED) == 0) 825 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 826 827 error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, NULL, 828 p); 829 830 if ((ioflg & IO_NODELOCKED) == 0) 831 VOP_UNLOCK(vp, 0); 832 833 if (error == 0) 834 *buflen = *buflen - auio.uio_resid; 835 836 return (error); 837 } 838 839 /* 840 * XXX Failure mode if partially written? 841 */ 842 int 843 vn_extattr_set(struct vnode *vp, int ioflg, int attrnamespace, 844 const char *attrname, size_t buflen, const void *bf, struct proc *p) 845 { 846 struct uio auio; 847 struct iovec aiov; 848 struct mount *mp; 849 int error; 850 851 aiov.iov_len = buflen; 852 aiov.iov_base = __UNCONST(bf); /* XXXUNCONST kills const */ 853 854 auio.uio_iov = &aiov; 855 auio.uio_iovcnt = 1; 856 auio.uio_rw = UIO_WRITE; 857 auio.uio_segflg = UIO_SYSSPACE; 858 auio.uio_procp = p; 859 auio.uio_offset = 0; 860 auio.uio_resid = buflen; 861 862 if ((ioflg & IO_NODELOCKED) == 0) { 863 if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0) 864 return (error); 865 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 866 } 867 868 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, p); 869 870 if ((ioflg & IO_NODELOCKED) == 0) { 871 vn_finished_write(mp, 0); 872 VOP_UNLOCK(vp, 0); 873 } 874 875 return (error); 876 } 877 878 int 879 vn_extattr_rm(struct vnode *vp, int ioflg, int attrnamespace, 880 const char *attrname, struct proc *p) 881 { 882 struct mount *mp; 883 int error; 884 885 if ((ioflg & IO_NODELOCKED) == 0) { 886 if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0) 887 return (error); 888 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 889 } 890 891 error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, NULL, p); 892 if (error == EOPNOTSUPP) 893 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL, 894 NULL, p); 895 896 if ((ioflg & IO_NODELOCKED) == 0) { 897 vn_finished_write(mp, 0); 898 VOP_UNLOCK(vp, 0); 899 } 900 901 return (error); 902 } 903
Indexes created Sat Sep 20 22:09:52 GMT 2025