vfs_vnops.c revision 1.98
1 /* $NetBSD: vfs_vnops.c,v 1.98 2005/10/15 21:18:54 elad 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 1.98 2005/10/15 21:18:54 elad 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(struct nameidata *ndp, int fmode, int cmode) 98 { 99 struct vnode *vp; 100 struct mount *mp; 101 struct proc *p = ndp->ni_cnd.cn_proc; 102 struct ucred *cred = p->p_ucred; 103 struct vattr va; 104 int error; 105 #ifdef VERIFIED_EXEC 106 struct veriexec_hash_entry *vhe = NULL; 107 char pathbuf[MAXPATHLEN]; 108 size_t pathlen; 109 int (*copyfun)(const void *, void *, size_t, size_t *) = 110 ndp->ni_segflg == UIO_SYSSPACE ? copystr : copyinstr; 111 #endif /* VERIFIED_EXEC */ 112 113 #ifdef VERIFIED_EXEC 114 error = (*copyfun)(ndp->ni_dirp, pathbuf, sizeof(pathbuf), &pathlen); 115 if (error) { 116 if (veriexec_verbose >= 1) 117 printf("veriexec: Can't copy path. (error=%d)\n", 118 error); 119 120 return (error); 121 } 122 #endif /* VERIFIED_EXEC */ 123 124 restart: 125 if (fmode & O_CREAT) { 126 ndp->ni_cnd.cn_nameiop = CREATE; 127 ndp->ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 128 if ((fmode & O_EXCL) == 0 && 129 ((fmode & O_NOFOLLOW) == 0)) 130 ndp->ni_cnd.cn_flags |= FOLLOW; 131 if ((error = namei(ndp)) != 0) 132 return (error); 133 if (ndp->ni_vp == NULL) { 134 #ifdef VERIFIED_EXEC 135 /* Lockdown mode: Prevent creation of new files. */ 136 if (veriexec_strict >= 3) { 137 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 138 139 printf("Veriexec: vn_open: Preventing " 140 "new file creation in %s.\n", 141 pathbuf); 142 143 vp = ndp->ni_dvp; 144 error = EPERM; 145 goto bad; 146 } 147 #endif /* VERIFIED_EXEC */ 148 149 VATTR_NULL(&va); 150 va.va_type = VREG; 151 va.va_mode = cmode; 152 if (fmode & O_EXCL) 153 va.va_vaflags |= VA_EXCLUSIVE; 154 if (vn_start_write(ndp->ni_dvp, &mp, V_NOWAIT) != 0) { 155 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 156 vput(ndp->ni_dvp); 157 if ((error = vn_start_write(NULL, &mp, 158 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 159 return (error); 160 goto restart; 161 } 162 VOP_LEASE(ndp->ni_dvp, p, cred, LEASE_WRITE); 163 error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp, 164 &ndp->ni_cnd, &va); 165 vn_finished_write(mp, 0); 166 if (error) 167 return (error); 168 fmode &= ~O_TRUNC; 169 vp = ndp->ni_vp; 170 } else { 171 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 172 if (ndp->ni_dvp == ndp->ni_vp) 173 vrele(ndp->ni_dvp); 174 else 175 vput(ndp->ni_dvp); 176 ndp->ni_dvp = NULL; 177 vp = ndp->ni_vp; 178 if (fmode & O_EXCL) { 179 error = EEXIST; 180 goto bad; 181 } 182 fmode &= ~O_CREAT; 183 } 184 } else { 185 ndp->ni_cnd.cn_nameiop = LOOKUP; 186 ndp->ni_cnd.cn_flags = LOCKLEAF; 187 if ((fmode & O_NOFOLLOW) == 0) 188 ndp->ni_cnd.cn_flags |= FOLLOW; 189 if ((error = namei(ndp)) != 0) 190 return (error); 191 vp = ndp->ni_vp; 192 } 193 if (vp->v_type == VSOCK) { 194 error = EOPNOTSUPP; 195 goto bad; 196 } 197 if (ndp->ni_vp->v_type == VLNK) { 198 error = EFTYPE; 199 goto bad; 200 } 201 202 #ifdef VERIFIED_EXEC 203 if ((error = VOP_GETATTR(vp, &va, cred, p)) != 0) 204 goto bad; 205 #endif 206 207 if ((fmode & O_CREAT) == 0) { 208 #ifdef VERIFIED_EXEC 209 if ((error = veriexec_verify(p, vp, &va, pathbuf, 210 VERIEXEC_FILE, &vhe)) != 0) 211 goto bad; 212 #endif 213 214 if (fmode & FREAD) { 215 if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0) 216 goto bad; 217 } 218 219 if (fmode & (FWRITE | O_TRUNC)) { 220 if (vp->v_type == VDIR) { 221 error = EISDIR; 222 goto bad; 223 } 224 if ((error = vn_writechk(vp)) != 0 || 225 (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0) 226 goto bad; 227 #ifdef VERIFIED_EXEC 228 if (vhe != NULL) { 229 veriexec_report("Write access request.", 230 pathbuf, &va, p, 231 REPORT_NOVERBOSE, 232 REPORT_ALARM, 233 REPORT_NOPANIC); 234 235 /* IPS mode: Deny writing to monitored files. */ 236 if (veriexec_strict >= 2) { 237 error = EPERM; 238 goto bad; 239 } else { 240 vhe->status = FINGERPRINT_NOTEVAL; 241 } 242 } 243 #endif 244 } 245 } 246 247 if (fmode & O_TRUNC) { 248 VOP_UNLOCK(vp, 0); /* XXX */ 249 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 250 vput(vp); 251 return (error); 252 } 253 VOP_LEASE(vp, p, cred, LEASE_WRITE); 254 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 255 VATTR_NULL(&va); 256 va.va_size = 0; 257 error = VOP_SETATTR(vp, &va, cred, p); 258 vn_finished_write(mp, 0); 259 if (error != 0) 260 goto bad; 261 } 262 if ((error = VOP_OPEN(vp, fmode, cred, p)) != 0) 263 goto bad; 264 if (vp->v_type == VREG && 265 uvn_attach(vp, fmode & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 266 error = EIO; 267 goto bad; 268 } 269 if (fmode & FWRITE) 270 vp->v_writecount++; 271 272 return (0); 273 bad: 274 vput(vp); 275 return (error); 276 } 277 278 /* 279 * Check for write permissions on the specified vnode. 280 * Prototype text segments cannot be written. 281 */ 282 int 283 vn_writechk(struct vnode *vp) 284 { 285 286 /* 287 * If the vnode is in use as a process's text, 288 * we can't allow writing. 289 */ 290 if (vp->v_flag & VTEXT) 291 return (ETXTBSY); 292 return (0); 293 } 294 295 /* 296 * Mark a vnode as having executable mappings. 297 */ 298 void 299 vn_markexec(struct vnode *vp) 300 { 301 if ((vp->v_flag & VEXECMAP) == 0) { 302 uvmexp.filepages -= vp->v_uobj.uo_npages; 303 uvmexp.execpages += vp->v_uobj.uo_npages; 304 } 305 vp->v_flag |= VEXECMAP; 306 } 307 308 /* 309 * Mark a vnode as being the text of a process. 310 * Fail if the vnode is currently writable. 311 */ 312 int 313 vn_marktext(struct vnode *vp) 314 { 315 316 if (vp->v_writecount != 0) { 317 KASSERT((vp->v_flag & VTEXT) == 0); 318 return (ETXTBSY); 319 } 320 vp->v_flag |= VTEXT; 321 vn_markexec(vp); 322 return (0); 323 } 324 325 /* 326 * Vnode close call 327 * 328 * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node. 329 */ 330 int 331 vn_close(struct vnode *vp, int flags, struct ucred *cred, struct proc *p) 332 { 333 int error; 334 335 if (flags & FWRITE) 336 vp->v_writecount--; 337 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 338 error = VOP_CLOSE(vp, flags, cred, p); 339 vput(vp); 340 return (error); 341 } 342 343 /* 344 * Package up an I/O request on a vnode into a uio and do it. 345 */ 346 int 347 vn_rdwr(enum uio_rw rw, struct vnode *vp, caddr_t base, int len, off_t offset, 348 enum uio_seg segflg, int ioflg, struct ucred *cred, 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(struct file *fp, char *bf, int segflg, u_int count, int *done, 396 struct proc *p, off_t **cookies, int *ncookies) 397 { 398 struct vnode *vp = (struct vnode *)fp->f_data; 399 struct iovec aiov; 400 struct uio auio; 401 int error, eofflag; 402 403 unionread: 404 if (vp->v_type != VDIR) 405 return (EINVAL); 406 aiov.iov_base = bf; 407 aiov.iov_len = count; 408 auio.uio_iov = &aiov; 409 auio.uio_iovcnt = 1; 410 auio.uio_rw = UIO_READ; 411 auio.uio_segflg = segflg; 412 auio.uio_procp = p; 413 auio.uio_resid = count; 414 vn_lock(vp, LK_SHARED | LK_RETRY); 415 auio.uio_offset = fp->f_offset; 416 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies, 417 ncookies); 418 fp->f_offset = auio.uio_offset; 419 VOP_UNLOCK(vp, 0); 420 if (error) 421 return (error); 422 423 #if defined(UNION) || defined(LKM) 424 if (count == auio.uio_resid && vn_union_readdir_hook) { 425 struct vnode *ovp = vp; 426 427 error = (*vn_union_readdir_hook)(&vp, fp, p); 428 if (error) 429 return (error); 430 if (vp != ovp) 431 goto unionread; 432 } 433 #endif /* UNION || LKM */ 434 435 if (count == auio.uio_resid && (vp->v_flag & VROOT) && 436 (vp->v_mount->mnt_flag & MNT_UNION)) { 437 struct vnode *tvp = vp; 438 vp = vp->v_mount->mnt_vnodecovered; 439 VREF(vp); 440 fp->f_data = vp; 441 fp->f_offset = 0; 442 vrele(tvp); 443 goto unionread; 444 } 445 *done = count - auio.uio_resid; 446 return error; 447 } 448 449 /* 450 * File table vnode read routine. 451 */ 452 static int 453 vn_read(struct file *fp, off_t *offset, struct uio *uio, struct ucred *cred, 454 int flags) 455 { 456 struct vnode *vp = (struct vnode *)fp->f_data; 457 int count, error, ioflag = 0; 458 459 VOP_LEASE(vp, uio->uio_procp, cred, LEASE_READ); 460 if (fp->f_flag & FNONBLOCK) 461 ioflag |= IO_NDELAY; 462 if ((fp->f_flag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC)) 463 ioflag |= IO_SYNC; 464 if (fp->f_flag & FALTIO) 465 ioflag |= IO_ALTSEMANTICS; 466 vn_lock(vp, LK_SHARED | LK_RETRY); 467 uio->uio_offset = *offset; 468 count = uio->uio_resid; 469 error = VOP_READ(vp, uio, ioflag, cred); 470 if (flags & FOF_UPDATE_OFFSET) 471 *offset += count - uio->uio_resid; 472 VOP_UNLOCK(vp, 0); 473 return (error); 474 } 475 476 /* 477 * File table vnode write routine. 478 */ 479 static int 480 vn_write(struct file *fp, off_t *offset, struct uio *uio, struct ucred *cred, 481 int flags) 482 { 483 struct vnode *vp = (struct vnode *)fp->f_data; 484 struct mount *mp; 485 int count, error, ioflag = IO_UNIT; 486 487 if (vp->v_type == VREG && (fp->f_flag & O_APPEND)) 488 ioflag |= IO_APPEND; 489 if (fp->f_flag & FNONBLOCK) 490 ioflag |= IO_NDELAY; 491 if (fp->f_flag & FFSYNC || 492 (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS))) 493 ioflag |= IO_SYNC; 494 else if (fp->f_flag & FDSYNC) 495 ioflag |= IO_DSYNC; 496 if (fp->f_flag & FALTIO) 497 ioflag |= IO_ALTSEMANTICS; 498 mp = NULL; 499 if (vp->v_type != VCHR && 500 (error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 501 return (error); 502 VOP_LEASE(vp, uio->uio_procp, cred, LEASE_WRITE); 503 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 504 uio->uio_offset = *offset; 505 count = uio->uio_resid; 506 error = VOP_WRITE(vp, uio, ioflag, cred); 507 if (flags & FOF_UPDATE_OFFSET) { 508 if (ioflag & IO_APPEND) 509 *offset = uio->uio_offset; 510 else 511 *offset += count - uio->uio_resid; 512 } 513 VOP_UNLOCK(vp, 0); 514 vn_finished_write(mp, 0); 515 return (error); 516 } 517 518 /* 519 * File table vnode stat routine. 520 */ 521 static int 522 vn_statfile(struct file *fp, struct stat *sb, struct proc *p) 523 { 524 struct vnode *vp = (struct vnode *)fp->f_data; 525 526 return vn_stat(vp, sb, p); 527 } 528 529 int 530 vn_stat(struct vnode *vp, struct stat *sb, struct proc *p) 531 { 532 struct vattr va; 533 int error; 534 mode_t mode; 535 536 error = VOP_GETATTR(vp, &va, p->p_ucred, p); 537 if (error) 538 return (error); 539 /* 540 * Copy from vattr table 541 */ 542 sb->st_dev = va.va_fsid; 543 sb->st_ino = va.va_fileid; 544 mode = va.va_mode; 545 switch (vp->v_type) { 546 case VREG: 547 mode |= S_IFREG; 548 break; 549 case VDIR: 550 mode |= S_IFDIR; 551 break; 552 case VBLK: 553 mode |= S_IFBLK; 554 break; 555 case VCHR: 556 mode |= S_IFCHR; 557 break; 558 case VLNK: 559 mode |= S_IFLNK; 560 break; 561 case VSOCK: 562 mode |= S_IFSOCK; 563 break; 564 case VFIFO: 565 mode |= S_IFIFO; 566 break; 567 default: 568 return (EBADF); 569 }; 570 sb->st_mode = mode; 571 sb->st_nlink = va.va_nlink; 572 sb->st_uid = va.va_uid; 573 sb->st_gid = va.va_gid; 574 sb->st_rdev = va.va_rdev; 575 sb->st_size = va.va_size; 576 sb->st_atimespec = va.va_atime; 577 sb->st_mtimespec = va.va_mtime; 578 sb->st_ctimespec = va.va_ctime; 579 sb->st_birthtimespec = va.va_birthtime; 580 sb->st_blksize = va.va_blocksize; 581 sb->st_flags = va.va_flags; 582 sb->st_gen = 0; 583 sb->st_blocks = va.va_bytes / S_BLKSIZE; 584 return (0); 585 } 586 587 /* 588 * File table vnode fcntl routine. 589 */ 590 static int 591 vn_fcntl(struct file *fp, u_int com, void *data, struct proc *p) 592 { 593 struct vnode *vp = ((struct vnode *)fp->f_data); 594 int error; 595 596 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 597 error = VOP_FCNTL(vp, com, data, fp->f_flag, p->p_ucred, p); 598 VOP_UNLOCK(vp, 0); 599 return (error); 600 } 601 602 /* 603 * File table vnode ioctl routine. 604 */ 605 static int 606 vn_ioctl(struct file *fp, u_long com, void *data, struct proc *p) 607 { 608 struct vnode *vp = ((struct vnode *)fp->f_data); 609 struct vattr vattr; 610 int error; 611 612 switch (vp->v_type) { 613 614 case VREG: 615 case VDIR: 616 if (com == FIONREAD) { 617 error = VOP_GETATTR(vp, &vattr, p->p_ucred, p); 618 if (error) 619 return (error); 620 *(int *)data = vattr.va_size - fp->f_offset; 621 return (0); 622 } 623 if ((com == FIONWRITE) || (com == FIONSPACE)) { 624 /* 625 * Files don't have send queues, so there never 626 * are any bytes in them, nor is there any 627 * open space in them. 628 */ 629 *(int *)data = 0; 630 return (0); 631 } 632 if (com == FIOGETBMAP) { 633 daddr_t *block; 634 635 if (*(daddr_t *)data < 0) 636 return (EINVAL); 637 block = (daddr_t *)data; 638 return (VOP_BMAP(vp, *block, NULL, block, NULL)); 639 } 640 if (com == OFIOGETBMAP) { 641 daddr_t ibn, obn; 642 643 if (*(int32_t *)data < 0) 644 return (EINVAL); 645 ibn = (daddr_t)*(int32_t *)data; 646 error = VOP_BMAP(vp, ibn, NULL, &obn, NULL); 647 *(int32_t *)data = (int32_t)obn; 648 return error; 649 } 650 if (com == FIONBIO || com == FIOASYNC) /* XXX */ 651 return (0); /* XXX */ 652 /* fall into ... */ 653 case VFIFO: 654 case VCHR: 655 case VBLK: 656 error = VOP_IOCTL(vp, com, data, fp->f_flag, p->p_ucred, p); 657 if (error == 0 && com == TIOCSCTTY) { 658 if (p->p_session->s_ttyvp) 659 vrele(p->p_session->s_ttyvp); 660 p->p_session->s_ttyvp = vp; 661 VREF(vp); 662 } 663 return (error); 664 665 default: 666 return (EPASSTHROUGH); 667 } 668 } 669 670 /* 671 * File table vnode poll routine. 672 */ 673 static int 674 vn_poll(struct file *fp, int events, struct proc *p) 675 { 676 677 return (VOP_POLL(((struct vnode *)fp->f_data), events, p)); 678 } 679 680 /* 681 * File table vnode kqfilter routine. 682 */ 683 int 684 vn_kqfilter(struct file *fp, struct knote *kn) 685 { 686 687 return (VOP_KQFILTER((struct vnode *)fp->f_data, kn)); 688 } 689 690 /* 691 * Check that the vnode is still valid, and if so 692 * acquire requested lock. 693 */ 694 int 695 vn_lock(struct vnode *vp, int flags) 696 { 697 int error; 698 699 #if 0 700 KASSERT(vp->v_usecount > 0 || (flags & LK_INTERLOCK) != 0 701 || (vp->v_flag & VONWORKLST) != 0); 702 #endif 703 704 do { 705 if ((flags & LK_INTERLOCK) == 0) 706 simple_lock(&vp->v_interlock); 707 if (vp->v_flag & VXLOCK) { 708 if (flags & LK_NOWAIT) { 709 simple_unlock(&vp->v_interlock); 710 return EBUSY; 711 } 712 vp->v_flag |= VXWANT; 713 ltsleep(vp, PINOD | PNORELOCK, 714 "vn_lock", 0, &vp->v_interlock); 715 error = ENOENT; 716 } else { 717 error = VOP_LOCK(vp, 718 (flags & ~LK_RETRY) | LK_INTERLOCK); 719 if (error == 0 || error == EDEADLK || error == EBUSY) 720 return (error); 721 } 722 flags &= ~LK_INTERLOCK; 723 } while (flags & LK_RETRY); 724 return (error); 725 } 726 727 /* 728 * File table vnode close routine. 729 */ 730 static int 731 vn_closefile(struct file *fp, struct proc *p) 732 { 733 734 return (vn_close(((struct vnode *)fp->f_data), fp->f_flag, 735 fp->f_cred, p)); 736 } 737 738 /* 739 * Enable LK_CANRECURSE on lock. Return prior status. 740 */ 741 u_int 742 vn_setrecurse(struct vnode *vp) 743 { 744 struct lock *lkp = &vp->v_lock; 745 u_int retval = lkp->lk_flags & LK_CANRECURSE; 746 747 lkp->lk_flags |= LK_CANRECURSE; 748 return retval; 749 } 750 751 /* 752 * Called when done with locksetrecurse. 753 */ 754 void 755 vn_restorerecurse(struct vnode *vp, u_int flags) 756 { 757 struct lock *lkp = &vp->v_lock; 758 759 lkp->lk_flags &= ~LK_CANRECURSE; 760 lkp->lk_flags |= flags; 761 } 762 763 int 764 vn_cow_establish(struct vnode *vp, 765 int (*func)(void *, struct buf *), void *cookie) 766 { 767 int s; 768 struct spec_cow_entry *e; 769 770 MALLOC(e, struct spec_cow_entry *, sizeof(struct spec_cow_entry), 771 M_DEVBUF, M_WAITOK); 772 e->ce_func = func; 773 e->ce_cookie = cookie; 774 775 SPEC_COW_LOCK(vp->v_specinfo, s); 776 vp->v_spec_cow_req++; 777 while (vp->v_spec_cow_count > 0) 778 ltsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 779 &vp->v_spec_cow_slock); 780 781 SLIST_INSERT_HEAD(&vp->v_spec_cow_head, e, ce_list); 782 783 vp->v_spec_cow_req--; 784 if (vp->v_spec_cow_req == 0) 785 wakeup(&vp->v_spec_cow_req); 786 SPEC_COW_UNLOCK(vp->v_specinfo, s); 787 788 return 0; 789 } 790 791 int 792 vn_cow_disestablish(struct vnode *vp, 793 int (*func)(void *, struct buf *), void *cookie) 794 { 795 int s; 796 struct spec_cow_entry *e; 797 798 SPEC_COW_LOCK(vp->v_specinfo, s); 799 vp->v_spec_cow_req++; 800 while (vp->v_spec_cow_count > 0) 801 ltsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 802 &vp->v_spec_cow_slock); 803 804 SLIST_FOREACH(e, &vp->v_spec_cow_head, ce_list) 805 if (e->ce_func == func && e->ce_cookie == cookie) { 806 SLIST_REMOVE(&vp->v_spec_cow_head, e, 807 spec_cow_entry, ce_list); 808 FREE(e, M_DEVBUF); 809 break; 810 } 811 812 vp->v_spec_cow_req--; 813 if (vp->v_spec_cow_req == 0) 814 wakeup(&vp->v_spec_cow_req); 815 SPEC_COW_UNLOCK(vp->v_specinfo, s); 816 817 return e ? 0 : EINVAL; 818 } 819 820 /* 821 * Simplified in-kernel wrapper calls for extended attribute access. 822 * Both calls pass in a NULL credential, authorizing a "kernel" access. 823 * Set IO_NODELOCKED in ioflg if the vnode is already locked. 824 */ 825 int 826 vn_extattr_get(struct vnode *vp, int ioflg, int attrnamespace, 827 const char *attrname, size_t *buflen, void *bf, struct proc *p) 828 { 829 struct uio auio; 830 struct iovec aiov; 831 int error; 832 833 aiov.iov_len = *buflen; 834 aiov.iov_base = bf; 835 836 auio.uio_iov = &aiov; 837 auio.uio_iovcnt = 1; 838 auio.uio_rw = UIO_READ; 839 auio.uio_segflg = UIO_SYSSPACE; 840 auio.uio_procp = p; 841 auio.uio_offset = 0; 842 auio.uio_resid = *buflen; 843 844 if ((ioflg & IO_NODELOCKED) == 0) 845 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 846 847 error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, NULL, 848 p); 849 850 if ((ioflg & IO_NODELOCKED) == 0) 851 VOP_UNLOCK(vp, 0); 852 853 if (error == 0) 854 *buflen = *buflen - auio.uio_resid; 855 856 return (error); 857 } 858 859 /* 860 * XXX Failure mode if partially written? 861 */ 862 int 863 vn_extattr_set(struct vnode *vp, int ioflg, int attrnamespace, 864 const char *attrname, size_t buflen, const void *bf, struct proc *p) 865 { 866 struct uio auio; 867 struct iovec aiov; 868 struct mount *mp; 869 int error; 870 871 aiov.iov_len = buflen; 872 aiov.iov_base = __UNCONST(bf); /* XXXUNCONST kills const */ 873 874 auio.uio_iov = &aiov; 875 auio.uio_iovcnt = 1; 876 auio.uio_rw = UIO_WRITE; 877 auio.uio_segflg = UIO_SYSSPACE; 878 auio.uio_procp = p; 879 auio.uio_offset = 0; 880 auio.uio_resid = buflen; 881 882 if ((ioflg & IO_NODELOCKED) == 0) { 883 if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0) 884 return (error); 885 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 886 } 887 888 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, p); 889 890 if ((ioflg & IO_NODELOCKED) == 0) { 891 vn_finished_write(mp, 0); 892 VOP_UNLOCK(vp, 0); 893 } 894 895 return (error); 896 } 897 898 int 899 vn_extattr_rm(struct vnode *vp, int ioflg, int attrnamespace, 900 const char *attrname, struct proc *p) 901 { 902 struct mount *mp; 903 int error; 904 905 if ((ioflg & IO_NODELOCKED) == 0) { 906 if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0) 907 return (error); 908 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 909 } 910 911 error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, NULL, p); 912 if (error == EOPNOTSUPP) 913 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL, 914 NULL, p); 915 916 if ((ioflg & IO_NODELOCKED) == 0) { 917 vn_finished_write(mp, 0); 918 VOP_UNLOCK(vp, 0); 919 } 920 921 return (error); 922 } 923 924 /* 925 * Preparing to start a filesystem write operation. If the operation is 926 * permitted, then we bump the count of operations in progress and 927 * proceed. If a suspend request is in progress, we wait until the 928 * suspension is over, and then proceed. 929 * V_PCATCH adds PCATCH to the tsleep flags. 930 * V_WAIT waits until suspension is over. Otherwise returns EWOULDBLOCK. 931 * V_SLEEPONLY wait, but do not bump the operations count. 932 * V_LOWER this is a lower level operation. No further vnodes should be 933 * locked. Otherwise it is a upper level operation. No vnodes 934 * should be locked. 935 */ 936 int 937 vn_start_write(struct vnode *vp, struct mount **mpp, int flags) 938 { 939 struct mount *mp; 940 int error, mask, prio; 941 942 /* 943 * If a vnode is provided, get and return the mount point that 944 * to which it will write. 945 */ 946 if (vp != NULL) { 947 *mpp = vp->v_mount; 948 } 949 if ((mp = *mpp) == NULL) 950 return (0); 951 mp = mp->mnt_leaf; 952 /* 953 * Check on status of suspension. 954 */ 955 prio = PUSER - 1; 956 if (flags & V_PCATCH) 957 prio |= PCATCH; 958 959 if ((flags & V_LOWER) == 0) 960 mask = IMNT_SUSPEND; 961 else 962 mask = IMNT_SUSPENDLOW; 963 964 while ((mp->mnt_iflag & mask) != 0) { 965 if ((flags & V_WAIT) == 0) 966 return (EWOULDBLOCK); 967 error = tsleep(&mp->mnt_flag, prio, "suspfs", 0); 968 if (error) 969 return (error); 970 } 971 if (flags & V_SLEEPONLY) 972 return (0); 973 simple_lock(&mp->mnt_slock); 974 if ((flags & V_LOWER) == 0) 975 mp->mnt_writeopcountupper++; 976 else 977 mp->mnt_writeopcountlower++; 978 simple_unlock(&mp->mnt_slock); 979 return (0); 980 } 981 982 /* 983 * Filesystem write operation has completed. If we are suspending and this 984 * operation is the last one, notify the suspender that the suspension is 985 * now in effect. 986 */ 987 void 988 vn_finished_write(struct mount *mp, int flags) 989 { 990 if (mp == NULL) 991 return; 992 mp = mp->mnt_leaf; 993 simple_lock(&mp->mnt_slock); 994 if ((flags & V_LOWER) == 0) { 995 mp->mnt_writeopcountupper--; 996 if (mp->mnt_writeopcountupper < 0) 997 printf("vn_finished_write: neg cnt upper=%d\n", 998 mp->mnt_writeopcountupper); 999 if ((mp->mnt_iflag & IMNT_SUSPEND) != 0 && 1000 mp->mnt_writeopcountupper <= 0) 1001 wakeup(&mp->mnt_writeopcountupper); 1002 } else { 1003 mp->mnt_writeopcountlower--; 1004 if (mp->mnt_writeopcountlower < 0) 1005 printf("vn_finished_write: neg cnt lower=%d\n", 1006 mp->mnt_writeopcountlower); 1007 if ((mp->mnt_iflag & IMNT_SUSPENDLOW) != 0 && 1008 mp->mnt_writeopcountupper <= 0) 1009 wakeup(&mp->mnt_writeopcountlower); 1010 } 1011 simple_unlock(&mp->mnt_slock); 1012 } 1013
Indexes created Sat Sep 20 22:09:52 GMT 2025