vfs_vnops.c revision 1.96
1 /* $NetBSD: vfs_vnops.c,v 1.96 2005/10/14 12:47:04 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 1.96 2005/10/14 12:47:04 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 #endif /* VERIFIED_EXEC */ 110 111 #ifdef VERIFIED_EXEC 112 if (ndp->ni_segflg == UIO_SYSSPACE) 113 error = copystr(__UNCONST(ndp->ni_dirp), pathbuf, 114 sizeof(pathbuf), &pathlen); 115 else 116 error = copyinstr(__UNCONST(ndp->ni_dirp), pathbuf, 117 sizeof(pathbuf), &pathlen); 118 if (error) { 119 if (veriexec_verbose >= 1) 120 printf("veriexec: Can't copy path. (error=%d)\n", 121 error); 122 123 return (error); 124 } 125 #endif /* VERIFIED_EXEC */ 126 127 restart: 128 if (fmode & O_CREAT) { 129 ndp->ni_cnd.cn_nameiop = CREATE; 130 ndp->ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 131 if ((fmode & O_EXCL) == 0 && 132 ((fmode & O_NOFOLLOW) == 0)) 133 ndp->ni_cnd.cn_flags |= FOLLOW; 134 if ((error = namei(ndp)) != 0) 135 return (error); 136 if (ndp->ni_vp == NULL) { 137 #ifdef VERIFIED_EXEC 138 /* Lockdown mode: Prevent creation of new files. */ 139 if (veriexec_strict >= 3) { 140 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 141 142 printf("Veriexec: vn_open: Preventing " 143 "new file creation in %s.\n", 144 pathbuf); 145 146 vp = ndp->ni_dvp; 147 error = EPERM; 148 goto bad; 149 } 150 #endif /* VERIFIED_EXEC */ 151 152 VATTR_NULL(&va); 153 va.va_type = VREG; 154 va.va_mode = cmode; 155 if (fmode & O_EXCL) 156 va.va_vaflags |= VA_EXCLUSIVE; 157 if (vn_start_write(ndp->ni_dvp, &mp, V_NOWAIT) != 0) { 158 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 159 vput(ndp->ni_dvp); 160 if ((error = vn_start_write(NULL, &mp, 161 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 162 return (error); 163 goto restart; 164 } 165 VOP_LEASE(ndp->ni_dvp, p, cred, LEASE_WRITE); 166 error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp, 167 &ndp->ni_cnd, &va); 168 vn_finished_write(mp, 0); 169 if (error) 170 return (error); 171 fmode &= ~O_TRUNC; 172 vp = ndp->ni_vp; 173 } else { 174 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 175 if (ndp->ni_dvp == ndp->ni_vp) 176 vrele(ndp->ni_dvp); 177 else 178 vput(ndp->ni_dvp); 179 ndp->ni_dvp = NULL; 180 vp = ndp->ni_vp; 181 if (fmode & O_EXCL) { 182 error = EEXIST; 183 goto bad; 184 } 185 fmode &= ~O_CREAT; 186 } 187 } else { 188 ndp->ni_cnd.cn_nameiop = LOOKUP; 189 ndp->ni_cnd.cn_flags = LOCKLEAF; 190 if ((fmode & O_NOFOLLOW) == 0) 191 ndp->ni_cnd.cn_flags |= FOLLOW; 192 if ((error = namei(ndp)) != 0) 193 return (error); 194 vp = ndp->ni_vp; 195 } 196 if (vp->v_type == VSOCK) { 197 error = EOPNOTSUPP; 198 goto bad; 199 } 200 if (ndp->ni_vp->v_type == VLNK) { 201 error = EFTYPE; 202 goto bad; 203 } 204 205 #ifdef VERIFIED_EXEC 206 if ((error = VOP_GETATTR(vp, &va, cred, p)) != 0) 207 goto bad; 208 #endif 209 210 if ((fmode & O_CREAT) == 0) { 211 #ifdef VERIFIED_EXEC 212 if ((error = veriexec_verify(p, vp, &va, pathbuf, 213 VERIEXEC_FILE, &vhe)) != 0) 214 goto bad; 215 #endif 216 217 if (fmode & FREAD) { 218 if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0) 219 goto bad; 220 } 221 222 if (fmode & (FWRITE | O_TRUNC)) { 223 if (vp->v_type == VDIR) { 224 error = EISDIR; 225 goto bad; 226 } 227 if ((error = vn_writechk(vp)) != 0 || 228 (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0) 229 goto bad; 230 #ifdef VERIFIED_EXEC 231 if (vhe != NULL) { 232 veriexec_report("Write access request.", 233 pathbuf, &va, p, 234 REPORT_NOVERBOSE, 235 REPORT_ALARM, 236 REPORT_NOPANIC); 237 238 /* IPS mode: Deny writing to monitored files. */ 239 if (veriexec_strict >= 2) { 240 error = EPERM; 241 goto bad; 242 } else { 243 vhe->status = FINGERPRINT_NOTEVAL; 244 } 245 } 246 #endif 247 } 248 } 249 250 if (fmode & O_TRUNC) { 251 VOP_UNLOCK(vp, 0); /* XXX */ 252 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 253 vput(vp); 254 return (error); 255 } 256 VOP_LEASE(vp, p, cred, LEASE_WRITE); 257 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 258 VATTR_NULL(&va); 259 va.va_size = 0; 260 error = VOP_SETATTR(vp, &va, cred, p); 261 vn_finished_write(mp, 0); 262 if (error != 0) 263 goto bad; 264 } 265 if ((error = VOP_OPEN(vp, fmode, cred, p)) != 0) 266 goto bad; 267 if (vp->v_type == VREG && 268 uvn_attach(vp, fmode & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 269 error = EIO; 270 goto bad; 271 } 272 if (fmode & FWRITE) 273 vp->v_writecount++; 274 275 return (0); 276 bad: 277 vput(vp); 278 return (error); 279 } 280 281 /* 282 * Check for write permissions on the specified vnode. 283 * Prototype text segments cannot be written. 284 */ 285 int 286 vn_writechk(struct vnode *vp) 287 { 288 289 /* 290 * If the vnode is in use as a process's text, 291 * we can't allow writing. 292 */ 293 if (vp->v_flag & VTEXT) 294 return (ETXTBSY); 295 return (0); 296 } 297 298 /* 299 * Mark a vnode as having executable mappings. 300 */ 301 void 302 vn_markexec(struct vnode *vp) 303 { 304 if ((vp->v_flag & VEXECMAP) == 0) { 305 uvmexp.filepages -= vp->v_uobj.uo_npages; 306 uvmexp.execpages += vp->v_uobj.uo_npages; 307 } 308 vp->v_flag |= VEXECMAP; 309 } 310 311 /* 312 * Mark a vnode as being the text of a process. 313 * Fail if the vnode is currently writable. 314 */ 315 int 316 vn_marktext(struct vnode *vp) 317 { 318 319 if (vp->v_writecount != 0) { 320 KASSERT((vp->v_flag & VTEXT) == 0); 321 return (ETXTBSY); 322 } 323 vp->v_flag |= VTEXT; 324 vn_markexec(vp); 325 return (0); 326 } 327 328 /* 329 * Vnode close call 330 * 331 * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node. 332 */ 333 int 334 vn_close(struct vnode *vp, int flags, struct ucred *cred, struct proc *p) 335 { 336 int error; 337 338 if (flags & FWRITE) 339 vp->v_writecount--; 340 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 341 error = VOP_CLOSE(vp, flags, cred, p); 342 vput(vp); 343 return (error); 344 } 345 346 /* 347 * Package up an I/O request on a vnode into a uio and do it. 348 */ 349 int 350 vn_rdwr(enum uio_rw rw, struct vnode *vp, caddr_t base, int len, off_t offset, 351 enum uio_seg segflg, int ioflg, struct ucred *cred, size_t *aresid, 352 struct proc *p) 353 { 354 struct uio auio; 355 struct iovec aiov; 356 struct mount *mp; 357 int error; 358 359 if ((ioflg & IO_NODELOCKED) == 0) { 360 if (rw == UIO_READ) { 361 vn_lock(vp, LK_SHARED | LK_RETRY); 362 } else /* UIO_WRITE */ { 363 if (vp->v_type != VCHR && 364 (error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) 365 != 0) 366 return (error); 367 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 368 } 369 } 370 auio.uio_iov = &aiov; 371 auio.uio_iovcnt = 1; 372 aiov.iov_base = base; 373 aiov.iov_len = len; 374 auio.uio_resid = len; 375 auio.uio_offset = offset; 376 auio.uio_segflg = segflg; 377 auio.uio_rw = rw; 378 auio.uio_procp = p; 379 if (rw == UIO_READ) { 380 error = VOP_READ(vp, &auio, ioflg, cred); 381 } else { 382 error = VOP_WRITE(vp, &auio, ioflg, cred); 383 } 384 if (aresid) 385 *aresid = auio.uio_resid; 386 else 387 if (auio.uio_resid && error == 0) 388 error = EIO; 389 if ((ioflg & IO_NODELOCKED) == 0) { 390 if (rw == UIO_WRITE) 391 vn_finished_write(mp, 0); 392 VOP_UNLOCK(vp, 0); 393 } 394 return (error); 395 } 396 397 int 398 vn_readdir(struct file *fp, char *bf, int segflg, u_int count, int *done, 399 struct proc *p, off_t **cookies, int *ncookies) 400 { 401 struct vnode *vp = (struct vnode *)fp->f_data; 402 struct iovec aiov; 403 struct uio auio; 404 int error, eofflag; 405 406 unionread: 407 if (vp->v_type != VDIR) 408 return (EINVAL); 409 aiov.iov_base = bf; 410 aiov.iov_len = count; 411 auio.uio_iov = &aiov; 412 auio.uio_iovcnt = 1; 413 auio.uio_rw = UIO_READ; 414 auio.uio_segflg = segflg; 415 auio.uio_procp = p; 416 auio.uio_resid = count; 417 vn_lock(vp, LK_SHARED | LK_RETRY); 418 auio.uio_offset = fp->f_offset; 419 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies, 420 ncookies); 421 fp->f_offset = auio.uio_offset; 422 VOP_UNLOCK(vp, 0); 423 if (error) 424 return (error); 425 426 #if defined(UNION) || defined(LKM) 427 if (count == auio.uio_resid && vn_union_readdir_hook) { 428 struct vnode *ovp = vp; 429 430 error = (*vn_union_readdir_hook)(&vp, fp, p); 431 if (error) 432 return (error); 433 if (vp != ovp) 434 goto unionread; 435 } 436 #endif /* UNION || LKM */ 437 438 if (count == auio.uio_resid && (vp->v_flag & VROOT) && 439 (vp->v_mount->mnt_flag & MNT_UNION)) { 440 struct vnode *tvp = vp; 441 vp = vp->v_mount->mnt_vnodecovered; 442 VREF(vp); 443 fp->f_data = vp; 444 fp->f_offset = 0; 445 vrele(tvp); 446 goto unionread; 447 } 448 *done = count - auio.uio_resid; 449 return error; 450 } 451 452 /* 453 * File table vnode read routine. 454 */ 455 static int 456 vn_read(struct file *fp, off_t *offset, struct uio *uio, struct ucred *cred, 457 int flags) 458 { 459 struct vnode *vp = (struct vnode *)fp->f_data; 460 int count, error, ioflag = 0; 461 462 VOP_LEASE(vp, uio->uio_procp, cred, LEASE_READ); 463 if (fp->f_flag & FNONBLOCK) 464 ioflag |= IO_NDELAY; 465 if ((fp->f_flag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC)) 466 ioflag |= IO_SYNC; 467 if (fp->f_flag & FALTIO) 468 ioflag |= IO_ALTSEMANTICS; 469 vn_lock(vp, LK_SHARED | LK_RETRY); 470 uio->uio_offset = *offset; 471 count = uio->uio_resid; 472 error = VOP_READ(vp, uio, ioflag, cred); 473 if (flags & FOF_UPDATE_OFFSET) 474 *offset += count - uio->uio_resid; 475 VOP_UNLOCK(vp, 0); 476 return (error); 477 } 478 479 /* 480 * File table vnode write routine. 481 */ 482 static int 483 vn_write(struct file *fp, off_t *offset, struct uio *uio, struct ucred *cred, 484 int flags) 485 { 486 struct vnode *vp = (struct vnode *)fp->f_data; 487 struct mount *mp; 488 int count, error, ioflag = IO_UNIT; 489 490 if (vp->v_type == VREG && (fp->f_flag & O_APPEND)) 491 ioflag |= IO_APPEND; 492 if (fp->f_flag & FNONBLOCK) 493 ioflag |= IO_NDELAY; 494 if (fp->f_flag & FFSYNC || 495 (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS))) 496 ioflag |= IO_SYNC; 497 else if (fp->f_flag & FDSYNC) 498 ioflag |= IO_DSYNC; 499 if (fp->f_flag & FALTIO) 500 ioflag |= IO_ALTSEMANTICS; 501 mp = NULL; 502 if (vp->v_type != VCHR && 503 (error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 504 return (error); 505 VOP_LEASE(vp, uio->uio_procp, cred, LEASE_WRITE); 506 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 507 uio->uio_offset = *offset; 508 count = uio->uio_resid; 509 error = VOP_WRITE(vp, uio, ioflag, cred); 510 if (flags & FOF_UPDATE_OFFSET) { 511 if (ioflag & IO_APPEND) 512 *offset = uio->uio_offset; 513 else 514 *offset += count - uio->uio_resid; 515 } 516 VOP_UNLOCK(vp, 0); 517 vn_finished_write(mp, 0); 518 return (error); 519 } 520 521 /* 522 * File table vnode stat routine. 523 */ 524 static int 525 vn_statfile(struct file *fp, struct stat *sb, struct proc *p) 526 { 527 struct vnode *vp = (struct vnode *)fp->f_data; 528 529 return vn_stat(vp, sb, p); 530 } 531 532 int 533 vn_stat(struct vnode *vp, struct stat *sb, struct proc *p) 534 { 535 struct vattr va; 536 int error; 537 mode_t mode; 538 539 error = VOP_GETATTR(vp, &va, p->p_ucred, p); 540 if (error) 541 return (error); 542 /* 543 * Copy from vattr table 544 */ 545 sb->st_dev = va.va_fsid; 546 sb->st_ino = va.va_fileid; 547 mode = va.va_mode; 548 switch (vp->v_type) { 549 case VREG: 550 mode |= S_IFREG; 551 break; 552 case VDIR: 553 mode |= S_IFDIR; 554 break; 555 case VBLK: 556 mode |= S_IFBLK; 557 break; 558 case VCHR: 559 mode |= S_IFCHR; 560 break; 561 case VLNK: 562 mode |= S_IFLNK; 563 break; 564 case VSOCK: 565 mode |= S_IFSOCK; 566 break; 567 case VFIFO: 568 mode |= S_IFIFO; 569 break; 570 default: 571 return (EBADF); 572 }; 573 sb->st_mode = mode; 574 sb->st_nlink = va.va_nlink; 575 sb->st_uid = va.va_uid; 576 sb->st_gid = va.va_gid; 577 sb->st_rdev = va.va_rdev; 578 sb->st_size = va.va_size; 579 sb->st_atimespec = va.va_atime; 580 sb->st_mtimespec = va.va_mtime; 581 sb->st_ctimespec = va.va_ctime; 582 sb->st_birthtimespec = va.va_birthtime; 583 sb->st_blksize = va.va_blocksize; 584 sb->st_flags = va.va_flags; 585 sb->st_gen = 0; 586 sb->st_blocks = va.va_bytes / S_BLKSIZE; 587 return (0); 588 } 589 590 /* 591 * File table vnode fcntl routine. 592 */ 593 static int 594 vn_fcntl(struct file *fp, u_int com, void *data, struct proc *p) 595 { 596 struct vnode *vp = ((struct vnode *)fp->f_data); 597 int error; 598 599 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 600 error = VOP_FCNTL(vp, com, data, fp->f_flag, p->p_ucred, p); 601 VOP_UNLOCK(vp, 0); 602 return (error); 603 } 604 605 /* 606 * File table vnode ioctl routine. 607 */ 608 static int 609 vn_ioctl(struct file *fp, u_long com, void *data, struct proc *p) 610 { 611 struct vnode *vp = ((struct vnode *)fp->f_data); 612 struct vattr vattr; 613 int error; 614 615 switch (vp->v_type) { 616 617 case VREG: 618 case VDIR: 619 if (com == FIONREAD) { 620 error = VOP_GETATTR(vp, &vattr, p->p_ucred, p); 621 if (error) 622 return (error); 623 *(int *)data = vattr.va_size - fp->f_offset; 624 return (0); 625 } 626 if ((com == FIONWRITE) || (com == FIONSPACE)) { 627 /* 628 * Files don't have send queues, so there never 629 * are any bytes in them, nor is there any 630 * open space in them. 631 */ 632 *(int *)data = 0; 633 return (0); 634 } 635 if (com == FIOGETBMAP) { 636 daddr_t *block; 637 638 if (*(daddr_t *)data < 0) 639 return (EINVAL); 640 block = (daddr_t *)data; 641 return (VOP_BMAP(vp, *block, NULL, block, NULL)); 642 } 643 if (com == OFIOGETBMAP) { 644 daddr_t ibn, obn; 645 646 if (*(int32_t *)data < 0) 647 return (EINVAL); 648 ibn = (daddr_t)*(int32_t *)data; 649 error = VOP_BMAP(vp, ibn, NULL, &obn, NULL); 650 *(int32_t *)data = (int32_t)obn; 651 return error; 652 } 653 if (com == FIONBIO || com == FIOASYNC) /* XXX */ 654 return (0); /* XXX */ 655 /* fall into ... */ 656 case VFIFO: 657 case VCHR: 658 case VBLK: 659 error = VOP_IOCTL(vp, com, data, fp->f_flag, p->p_ucred, p); 660 if (error == 0 && com == TIOCSCTTY) { 661 if (p->p_session->s_ttyvp) 662 vrele(p->p_session->s_ttyvp); 663 p->p_session->s_ttyvp = vp; 664 VREF(vp); 665 } 666 return (error); 667 668 default: 669 return (EPASSTHROUGH); 670 } 671 } 672 673 /* 674 * File table vnode poll routine. 675 */ 676 static int 677 vn_poll(struct file *fp, int events, struct proc *p) 678 { 679 680 return (VOP_POLL(((struct vnode *)fp->f_data), events, p)); 681 } 682 683 /* 684 * File table vnode kqfilter routine. 685 */ 686 int 687 vn_kqfilter(struct file *fp, struct knote *kn) 688 { 689 690 return (VOP_KQFILTER((struct vnode *)fp->f_data, kn)); 691 } 692 693 /* 694 * Check that the vnode is still valid, and if so 695 * acquire requested lock. 696 */ 697 int 698 vn_lock(struct vnode *vp, int flags) 699 { 700 int error; 701 702 #if 0 703 KASSERT(vp->v_usecount > 0 || (flags & LK_INTERLOCK) != 0 704 || (vp->v_flag & VONWORKLST) != 0); 705 #endif 706 707 do { 708 if ((flags & LK_INTERLOCK) == 0) 709 simple_lock(&vp->v_interlock); 710 if (vp->v_flag & VXLOCK) { 711 if (flags & LK_NOWAIT) { 712 simple_unlock(&vp->v_interlock); 713 return EBUSY; 714 } 715 vp->v_flag |= VXWANT; 716 ltsleep(vp, PINOD | PNORELOCK, 717 "vn_lock", 0, &vp->v_interlock); 718 error = ENOENT; 719 } else { 720 error = VOP_LOCK(vp, 721 (flags & ~LK_RETRY) | LK_INTERLOCK); 722 if (error == 0 || error == EDEADLK || error == EBUSY) 723 return (error); 724 } 725 flags &= ~LK_INTERLOCK; 726 } while (flags & LK_RETRY); 727 return (error); 728 } 729 730 /* 731 * File table vnode close routine. 732 */ 733 static int 734 vn_closefile(struct file *fp, struct proc *p) 735 { 736 737 return (vn_close(((struct vnode *)fp->f_data), fp->f_flag, 738 fp->f_cred, p)); 739 } 740 741 /* 742 * Enable LK_CANRECURSE on lock. Return prior status. 743 */ 744 u_int 745 vn_setrecurse(struct vnode *vp) 746 { 747 struct lock *lkp = &vp->v_lock; 748 u_int retval = lkp->lk_flags & LK_CANRECURSE; 749 750 lkp->lk_flags |= LK_CANRECURSE; 751 return retval; 752 } 753 754 /* 755 * Called when done with locksetrecurse. 756 */ 757 void 758 vn_restorerecurse(struct vnode *vp, u_int flags) 759 { 760 struct lock *lkp = &vp->v_lock; 761 762 lkp->lk_flags &= ~LK_CANRECURSE; 763 lkp->lk_flags |= flags; 764 } 765 766 int 767 vn_cow_establish(struct vnode *vp, 768 int (*func)(void *, struct buf *), void *cookie) 769 { 770 int s; 771 struct spec_cow_entry *e; 772 773 MALLOC(e, struct spec_cow_entry *, sizeof(struct spec_cow_entry), 774 M_DEVBUF, M_WAITOK); 775 e->ce_func = func; 776 e->ce_cookie = cookie; 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_INSERT_HEAD(&vp->v_spec_cow_head, e, ce_list); 785 786 vp->v_spec_cow_req--; 787 if (vp->v_spec_cow_req == 0) 788 wakeup(&vp->v_spec_cow_req); 789 SPEC_COW_UNLOCK(vp->v_specinfo, s); 790 791 return 0; 792 } 793 794 int 795 vn_cow_disestablish(struct vnode *vp, 796 int (*func)(void *, struct buf *), void *cookie) 797 { 798 int s; 799 struct spec_cow_entry *e; 800 801 SPEC_COW_LOCK(vp->v_specinfo, s); 802 vp->v_spec_cow_req++; 803 while (vp->v_spec_cow_count > 0) 804 ltsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 805 &vp->v_spec_cow_slock); 806 807 SLIST_FOREACH(e, &vp->v_spec_cow_head, ce_list) 808 if (e->ce_func == func && e->ce_cookie == cookie) { 809 SLIST_REMOVE(&vp->v_spec_cow_head, e, 810 spec_cow_entry, ce_list); 811 FREE(e, M_DEVBUF); 812 break; 813 } 814 815 vp->v_spec_cow_req--; 816 if (vp->v_spec_cow_req == 0) 817 wakeup(&vp->v_spec_cow_req); 818 SPEC_COW_UNLOCK(vp->v_specinfo, s); 819 820 return e ? 0 : EINVAL; 821 } 822 823 /* 824 * Simplified in-kernel wrapper calls for extended attribute access. 825 * Both calls pass in a NULL credential, authorizing a "kernel" access. 826 * Set IO_NODELOCKED in ioflg if the vnode is already locked. 827 */ 828 int 829 vn_extattr_get(struct vnode *vp, int ioflg, int attrnamespace, 830 const char *attrname, size_t *buflen, void *bf, struct proc *p) 831 { 832 struct uio auio; 833 struct iovec aiov; 834 int error; 835 836 aiov.iov_len = *buflen; 837 aiov.iov_base = bf; 838 839 auio.uio_iov = &aiov; 840 auio.uio_iovcnt = 1; 841 auio.uio_rw = UIO_READ; 842 auio.uio_segflg = UIO_SYSSPACE; 843 auio.uio_procp = p; 844 auio.uio_offset = 0; 845 auio.uio_resid = *buflen; 846 847 if ((ioflg & IO_NODELOCKED) == 0) 848 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 849 850 error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, NULL, 851 p); 852 853 if ((ioflg & IO_NODELOCKED) == 0) 854 VOP_UNLOCK(vp, 0); 855 856 if (error == 0) 857 *buflen = *buflen - auio.uio_resid; 858 859 return (error); 860 } 861 862 /* 863 * XXX Failure mode if partially written? 864 */ 865 int 866 vn_extattr_set(struct vnode *vp, int ioflg, int attrnamespace, 867 const char *attrname, size_t buflen, const void *bf, struct proc *p) 868 { 869 struct uio auio; 870 struct iovec aiov; 871 struct mount *mp; 872 int error; 873 874 aiov.iov_len = buflen; 875 aiov.iov_base = __UNCONST(bf); /* XXXUNCONST kills const */ 876 877 auio.uio_iov = &aiov; 878 auio.uio_iovcnt = 1; 879 auio.uio_rw = UIO_WRITE; 880 auio.uio_segflg = UIO_SYSSPACE; 881 auio.uio_procp = p; 882 auio.uio_offset = 0; 883 auio.uio_resid = buflen; 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_SETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, p); 892 893 if ((ioflg & IO_NODELOCKED) == 0) { 894 vn_finished_write(mp, 0); 895 VOP_UNLOCK(vp, 0); 896 } 897 898 return (error); 899 } 900 901 int 902 vn_extattr_rm(struct vnode *vp, int ioflg, int attrnamespace, 903 const char *attrname, struct proc *p) 904 { 905 struct mount *mp; 906 int error; 907 908 if ((ioflg & IO_NODELOCKED) == 0) { 909 if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0) 910 return (error); 911 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 912 } 913 914 error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, NULL, p); 915 if (error == EOPNOTSUPP) 916 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL, 917 NULL, p); 918 919 if ((ioflg & IO_NODELOCKED) == 0) { 920 vn_finished_write(mp, 0); 921 VOP_UNLOCK(vp, 0); 922 } 923 924 return (error); 925 } 926 927 /* 928 * Preparing to start a filesystem write operation. If the operation is 929 * permitted, then we bump the count of operations in progress and 930 * proceed. If a suspend request is in progress, we wait until the 931 * suspension is over, and then proceed. 932 * V_PCATCH adds PCATCH to the tsleep flags. 933 * V_WAIT waits until suspension is over. Otherwise returns EWOULDBLOCK. 934 * V_SLEEPONLY wait, but do not bump the operations count. 935 * V_LOWER this is a lower level operation. No further vnodes should be 936 * locked. Otherwise it is a upper level operation. No vnodes 937 * should be locked. 938 */ 939 int 940 vn_start_write(struct vnode *vp, struct mount **mpp, int flags) 941 { 942 struct mount *mp; 943 int error, mask, prio; 944 945 /* 946 * If a vnode is provided, get and return the mount point that 947 * to which it will write. 948 */ 949 if (vp != NULL) { 950 *mpp = vp->v_mount; 951 } 952 if ((mp = *mpp) == NULL) 953 return (0); 954 mp = mp->mnt_leaf; 955 /* 956 * Check on status of suspension. 957 */ 958 prio = PUSER - 1; 959 if (flags & V_PCATCH) 960 prio |= PCATCH; 961 962 if ((flags & V_LOWER) == 0) 963 mask = IMNT_SUSPEND; 964 else 965 mask = IMNT_SUSPENDLOW; 966 967 while ((mp->mnt_iflag & mask) != 0) { 968 if ((flags & V_WAIT) == 0) 969 return (EWOULDBLOCK); 970 error = tsleep(&mp->mnt_flag, prio, "suspfs", 0); 971 if (error) 972 return (error); 973 } 974 if (flags & V_SLEEPONLY) 975 return (0); 976 simple_lock(&mp->mnt_slock); 977 if ((flags & V_LOWER) == 0) 978 mp->mnt_writeopcountupper++; 979 else 980 mp->mnt_writeopcountlower++; 981 simple_unlock(&mp->mnt_slock); 982 return (0); 983 } 984 985 /* 986 * Filesystem write operation has completed. If we are suspending and this 987 * operation is the last one, notify the suspender that the suspension is 988 * now in effect. 989 */ 990 void 991 vn_finished_write(struct mount *mp, int flags) 992 { 993 if (mp == NULL) 994 return; 995 mp = mp->mnt_leaf; 996 simple_lock(&mp->mnt_slock); 997 if ((flags & V_LOWER) == 0) { 998 mp->mnt_writeopcountupper--; 999 if (mp->mnt_writeopcountupper < 0) 1000 printf("vn_finished_write: neg cnt upper=%d\n", 1001 mp->mnt_writeopcountupper); 1002 if ((mp->mnt_iflag & IMNT_SUSPEND) != 0 && 1003 mp->mnt_writeopcountupper <= 0) 1004 wakeup(&mp->mnt_writeopcountupper); 1005 } else { 1006 mp->mnt_writeopcountlower--; 1007 if (mp->mnt_writeopcountlower < 0) 1008 printf("vn_finished_write: neg cnt lower=%d\n", 1009 mp->mnt_writeopcountlower); 1010 if ((mp->mnt_iflag & IMNT_SUSPENDLOW) != 0 && 1011 mp->mnt_writeopcountupper <= 0) 1012 wakeup(&mp->mnt_writeopcountlower); 1013 } 1014 simple_unlock(&mp->mnt_slock); 1015 } 1016
Indexes created Sat Sep 20 22:09:52 GMT 2025