vfs_vnops.c revision 1.99.2.2
1 /* $NetBSD: vfs_vnops.c,v 1.99.2.2 2005/11/15 05:24:48 yamt 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 1.99.2.2 2005/11/15 05:24:48 yamt 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 #include <uvm/uvm_readahead.h> 65 66 #ifdef UNION 67 #include <fs/union/union.h> 68 #endif 69 70 #if defined(LKM) || defined(UNION) 71 int (*vn_union_readdir_hook) (struct vnode **, struct file *, struct proc *); 72 #endif 73 74 #ifdef VERIFIED_EXEC 75 #include <sys/verified_exec.h> 76 #endif 77 78 static int vn_read(struct file *fp, off_t *offset, struct uio *uio, 79 struct ucred *cred, int flags); 80 static int vn_write(struct file *fp, off_t *offset, struct uio *uio, 81 struct ucred *cred, int flags); 82 static int vn_closefile(struct file *fp, struct proc *p); 83 static int vn_poll(struct file *fp, int events, struct proc *p); 84 static int vn_fcntl(struct file *fp, u_int com, void *data, struct proc *p); 85 static int vn_statfile(struct file *fp, struct stat *sb, struct proc *p); 86 static int vn_ioctl(struct file *fp, u_long com, void *data, struct proc *p); 87 88 const struct fileops vnops = { 89 vn_read, vn_write, vn_ioctl, vn_fcntl, vn_poll, 90 vn_statfile, vn_closefile, vn_kqfilter 91 }; 92 93 /* 94 * Common code for vnode open operations. 95 * Check permissions, and call the VOP_OPEN or VOP_CREATE routine. 96 */ 97 int 98 vn_open(struct nameidata *ndp, int fmode, int cmode) 99 { 100 struct vnode *vp; 101 struct mount *mp; 102 struct proc *p = ndp->ni_cnd.cn_proc; 103 struct ucred *cred = p->p_ucred; 104 struct vattr va; 105 int error; 106 #ifdef VERIFIED_EXEC 107 struct veriexec_hash_entry *vhe = NULL; 108 char pathbuf[MAXPATHLEN]; 109 size_t pathlen; 110 int (*copyfun)(const void *, void *, size_t, size_t *) = 111 ndp->ni_segflg == UIO_SYSSPACE ? copystr : copyinstr; 112 #endif /* VERIFIED_EXEC */ 113 114 #ifdef VERIFIED_EXEC 115 error = (*copyfun)(ndp->ni_dirp, pathbuf, sizeof(pathbuf), &pathlen); 116 if (error) { 117 if (veriexec_verbose >= 1) 118 printf("veriexec: Can't copy path. (error=%d)\n", 119 error); 120 121 return (error); 122 } 123 #endif /* VERIFIED_EXEC */ 124 125 restart: 126 if (fmode & O_CREAT) { 127 ndp->ni_cnd.cn_nameiop = CREATE; 128 ndp->ni_cnd.cn_flags = LOCKPARENT | LOCKLEAF; 129 if ((fmode & O_EXCL) == 0 && 130 ((fmode & O_NOFOLLOW) == 0)) 131 ndp->ni_cnd.cn_flags |= FOLLOW; 132 if ((error = namei(ndp)) != 0) 133 return (error); 134 if (ndp->ni_vp == NULL) { 135 #ifdef VERIFIED_EXEC 136 /* Lockdown mode: Prevent creation of new files. */ 137 if (veriexec_strict >= 3) { 138 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 139 140 printf("Veriexec: vn_open: Preventing " 141 "new file creation in %s.\n", 142 pathbuf); 143 144 vp = ndp->ni_dvp; 145 error = EPERM; 146 goto bad; 147 } 148 #endif /* VERIFIED_EXEC */ 149 150 VATTR_NULL(&va); 151 va.va_type = VREG; 152 va.va_mode = cmode; 153 if (fmode & O_EXCL) 154 va.va_vaflags |= VA_EXCLUSIVE; 155 if (vn_start_write(ndp->ni_dvp, &mp, V_NOWAIT) != 0) { 156 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 157 vput(ndp->ni_dvp); 158 if ((error = vn_start_write(NULL, &mp, 159 V_WAIT | V_SLEEPONLY | V_PCATCH)) != 0) 160 return (error); 161 goto restart; 162 } 163 VOP_LEASE(ndp->ni_dvp, p, cred, LEASE_WRITE); 164 error = VOP_CREATE(ndp->ni_dvp, &ndp->ni_vp, 165 &ndp->ni_cnd, &va); 166 vn_finished_write(mp, 0); 167 if (error) 168 return (error); 169 fmode &= ~O_TRUNC; 170 vp = ndp->ni_vp; 171 } else { 172 VOP_ABORTOP(ndp->ni_dvp, &ndp->ni_cnd); 173 if (ndp->ni_dvp == ndp->ni_vp) 174 vrele(ndp->ni_dvp); 175 else 176 vput(ndp->ni_dvp); 177 ndp->ni_dvp = NULL; 178 vp = ndp->ni_vp; 179 if (fmode & O_EXCL) { 180 error = EEXIST; 181 goto bad; 182 } 183 fmode &= ~O_CREAT; 184 } 185 } else { 186 ndp->ni_cnd.cn_nameiop = LOOKUP; 187 ndp->ni_cnd.cn_flags = LOCKLEAF; 188 if ((fmode & O_NOFOLLOW) == 0) 189 ndp->ni_cnd.cn_flags |= FOLLOW; 190 if ((error = namei(ndp)) != 0) 191 return (error); 192 vp = ndp->ni_vp; 193 } 194 if (vp->v_type == VSOCK) { 195 error = EOPNOTSUPP; 196 goto bad; 197 } 198 if (ndp->ni_vp->v_type == VLNK) { 199 error = EFTYPE; 200 goto bad; 201 } 202 203 #ifdef VERIFIED_EXEC 204 if ((error = VOP_GETATTR(vp, &va, cred, p)) != 0) 205 goto bad; 206 #endif 207 208 if ((fmode & O_CREAT) == 0) { 209 #ifdef VERIFIED_EXEC 210 if ((error = veriexec_verify(p, vp, &va, pathbuf, 211 VERIEXEC_FILE, &vhe)) != 0) 212 goto bad; 213 #endif 214 215 if (fmode & FREAD) { 216 if ((error = VOP_ACCESS(vp, VREAD, cred, p)) != 0) 217 goto bad; 218 } 219 220 if (fmode & (FWRITE | O_TRUNC)) { 221 if (vp->v_type == VDIR) { 222 error = EISDIR; 223 goto bad; 224 } 225 if ((error = vn_writechk(vp)) != 0 || 226 (error = VOP_ACCESS(vp, VWRITE, cred, p)) != 0) 227 goto bad; 228 #ifdef VERIFIED_EXEC 229 if (vhe != NULL) { 230 veriexec_report("Write access request.", 231 pathbuf, &va, p, 232 REPORT_NOVERBOSE, 233 REPORT_ALARM, 234 REPORT_NOPANIC); 235 236 /* IPS mode: Deny writing to monitored files. */ 237 if (veriexec_strict >= 2) { 238 error = EPERM; 239 goto bad; 240 } else { 241 vhe->status = FINGERPRINT_NOTEVAL; 242 } 243 } 244 #endif 245 } 246 } 247 248 if (fmode & O_TRUNC) { 249 VOP_UNLOCK(vp, 0); /* XXX */ 250 if ((error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) { 251 vrele(vp); 252 return (error); 253 } 254 VOP_LEASE(vp, p, cred, LEASE_WRITE); 255 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); /* XXX */ 256 VATTR_NULL(&va); 257 va.va_size = 0; 258 error = VOP_SETATTR(vp, &va, cred, p); 259 vn_finished_write(mp, 0); 260 if (error != 0) 261 goto bad; 262 } 263 if ((error = VOP_OPEN(vp, fmode, cred, p)) != 0) 264 goto bad; 265 if (vp->v_type == VREG && 266 uvn_attach(vp, fmode & FWRITE ? VM_PROT_WRITE : 0) == NULL) { 267 error = EIO; 268 goto bad; 269 } 270 if (fmode & FWRITE) 271 vp->v_writecount++; 272 273 return (0); 274 bad: 275 vput(vp); 276 return (error); 277 } 278 279 /* 280 * Check for write permissions on the specified vnode. 281 * Prototype text segments cannot be written. 282 */ 283 int 284 vn_writechk(struct vnode *vp) 285 { 286 287 /* 288 * If the vnode is in use as a process's text, 289 * we can't allow writing. 290 */ 291 if (vp->v_flag & VTEXT) 292 return (ETXTBSY); 293 return (0); 294 } 295 296 /* 297 * Mark a vnode as having executable mappings. 298 */ 299 void 300 vn_markexec(struct vnode *vp) 301 { 302 if ((vp->v_flag & VEXECMAP) == 0) { 303 uvmexp.filepages -= vp->v_uobj.uo_npages; 304 uvmexp.execpages += vp->v_uobj.uo_npages; 305 } 306 vp->v_flag |= VEXECMAP; 307 } 308 309 /* 310 * Mark a vnode as being the text of a process. 311 * Fail if the vnode is currently writable. 312 */ 313 int 314 vn_marktext(struct vnode *vp) 315 { 316 317 if (vp->v_writecount != 0) { 318 KASSERT((vp->v_flag & VTEXT) == 0); 319 return (ETXTBSY); 320 } 321 vp->v_flag |= VTEXT; 322 vn_markexec(vp); 323 return (0); 324 } 325 326 /* 327 * Vnode close call 328 * 329 * Note: takes an unlocked vnode, while VOP_CLOSE takes a locked node. 330 */ 331 int 332 vn_close(struct vnode *vp, int flags, struct ucred *cred, struct proc *p) 333 { 334 int error; 335 336 if (flags & FWRITE) 337 vp->v_writecount--; 338 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 339 error = VOP_CLOSE(vp, flags, cred, p); 340 vput(vp); 341 return (error); 342 } 343 344 /* 345 * Package up an I/O request on a vnode into a uio and do it. 346 */ 347 int 348 vn_rdwr(enum uio_rw rw, struct vnode *vp, caddr_t base, int len, off_t offset, 349 enum uio_seg segflg, int ioflg, struct ucred *cred, size_t *aresid, 350 struct proc *p) 351 { 352 struct uio auio; 353 struct iovec aiov; 354 struct mount *mp; 355 int error; 356 357 if ((ioflg & IO_NODELOCKED) == 0) { 358 if (rw == UIO_READ) { 359 vn_lock(vp, LK_SHARED | LK_RETRY); 360 } else /* UIO_WRITE */ { 361 if (vp->v_type != VCHR && 362 (error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) 363 != 0) 364 return (error); 365 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 366 } 367 } 368 auio.uio_iov = &aiov; 369 auio.uio_iovcnt = 1; 370 aiov.iov_base = base; 371 aiov.iov_len = len; 372 auio.uio_resid = len; 373 auio.uio_offset = offset; 374 auio.uio_segflg = segflg; 375 auio.uio_rw = rw; 376 auio.uio_procp = p; 377 if (rw == UIO_READ) { 378 error = VOP_READ(vp, &auio, NULL, ioflg, cred); 379 } else { 380 error = VOP_WRITE(vp, &auio, ioflg, cred); 381 } 382 if (aresid) 383 *aresid = auio.uio_resid; 384 else 385 if (auio.uio_resid && error == 0) 386 error = EIO; 387 if ((ioflg & IO_NODELOCKED) == 0) { 388 if (rw == UIO_WRITE) 389 vn_finished_write(mp, 0); 390 VOP_UNLOCK(vp, 0); 391 } 392 return (error); 393 } 394 395 int 396 vn_readdir(struct file *fp, char *bf, int segflg, u_int count, int *done, 397 struct proc *p, off_t **cookies, int *ncookies) 398 { 399 struct vnode *vp = (struct vnode *)fp->f_data; 400 struct iovec aiov; 401 struct uio auio; 402 int error, eofflag; 403 404 unionread: 405 if (vp->v_type != VDIR) 406 return (EINVAL); 407 aiov.iov_base = bf; 408 aiov.iov_len = count; 409 auio.uio_iov = &aiov; 410 auio.uio_iovcnt = 1; 411 auio.uio_rw = UIO_READ; 412 auio.uio_segflg = segflg; 413 auio.uio_procp = p; 414 auio.uio_resid = count; 415 vn_lock(vp, LK_SHARED | LK_RETRY); 416 auio.uio_offset = fp->f_offset; 417 error = VOP_READDIR(vp, &auio, fp->f_cred, &eofflag, cookies, 418 ncookies); 419 fp->f_offset = auio.uio_offset; 420 VOP_UNLOCK(vp, 0); 421 if (error) 422 return (error); 423 424 #if defined(UNION) || defined(LKM) 425 if (count == auio.uio_resid && vn_union_readdir_hook) { 426 struct vnode *ovp = vp; 427 428 error = (*vn_union_readdir_hook)(&vp, fp, p); 429 if (error) 430 return (error); 431 if (vp != ovp) 432 goto unionread; 433 } 434 #endif /* UNION || LKM */ 435 436 if (count == auio.uio_resid && (vp->v_flag & VROOT) && 437 (vp->v_mount->mnt_flag & MNT_UNION)) { 438 struct vnode *tvp = vp; 439 vp = vp->v_mount->mnt_vnodecovered; 440 VREF(vp); 441 fp->f_data = vp; 442 fp->f_offset = 0; 443 vrele(tvp); 444 goto unionread; 445 } 446 *done = count - auio.uio_resid; 447 return error; 448 } 449 450 /* 451 * File table vnode read routine. 452 */ 453 static int 454 vn_read(struct file *fp, off_t *offset, struct uio *uio, struct ucred *cred, 455 int flags) 456 { 457 struct vnode *vp = (struct vnode *)fp->f_data; 458 int count, error, ioflag = 0; 459 struct uvm_ractx *ra; 460 461 VOP_LEASE(vp, uio->uio_procp, cred, LEASE_READ); 462 if (fp->f_flag & FNONBLOCK) 463 ioflag |= IO_NDELAY; 464 if ((fp->f_flag & (FFSYNC | FRSYNC)) == (FFSYNC | FRSYNC)) 465 ioflag |= IO_SYNC; 466 if (fp->f_flag & FALTIO) 467 ioflag |= IO_ALTSEMANTICS; 468 469 simple_lock(&fp->f_slock); 470 ra = fp->f_ractx; 471 fp->f_ractx = NULL; 472 simple_unlock(&fp->f_slock); 473 if (ra == NULL) { 474 ra = uvm_ra_allocctx(fp->f_advice); 475 } 476 477 vn_lock(vp, LK_SHARED | LK_RETRY); 478 uio->uio_offset = *offset; 479 count = uio->uio_resid; 480 error = VOP_READ(vp, uio, ra, ioflag, cred); 481 if (flags & FOF_UPDATE_OFFSET) 482 *offset += count - uio->uio_resid; 483 VOP_UNLOCK(vp, 0); 484 485 simple_lock(&fp->f_slock); 486 if (fp->f_ractx == NULL) { 487 fp->f_ractx = ra; 488 ra = NULL; 489 } 490 simple_unlock(&fp->f_slock); 491 if (ra != NULL) { 492 uvm_ra_freectx(ra); 493 } 494 495 return (error); 496 } 497 498 /* 499 * File table vnode write routine. 500 */ 501 static int 502 vn_write(struct file *fp, off_t *offset, struct uio *uio, struct ucred *cred, 503 int flags) 504 { 505 struct vnode *vp = (struct vnode *)fp->f_data; 506 struct mount *mp; 507 int count, error, ioflag = IO_UNIT; 508 509 if (vp->v_type == VREG && (fp->f_flag & O_APPEND)) 510 ioflag |= IO_APPEND; 511 if (fp->f_flag & FNONBLOCK) 512 ioflag |= IO_NDELAY; 513 if (fp->f_flag & FFSYNC || 514 (vp->v_mount && (vp->v_mount->mnt_flag & MNT_SYNCHRONOUS))) 515 ioflag |= IO_SYNC; 516 else if (fp->f_flag & FDSYNC) 517 ioflag |= IO_DSYNC; 518 if (fp->f_flag & FALTIO) 519 ioflag |= IO_ALTSEMANTICS; 520 mp = NULL; 521 if (vp->v_type != VCHR && 522 (error = vn_start_write(vp, &mp, V_WAIT | V_PCATCH)) != 0) 523 return (error); 524 VOP_LEASE(vp, uio->uio_procp, cred, LEASE_WRITE); 525 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 526 uio->uio_offset = *offset; 527 count = uio->uio_resid; 528 error = VOP_WRITE(vp, uio, ioflag, cred); 529 if (flags & FOF_UPDATE_OFFSET) { 530 if (ioflag & IO_APPEND) 531 *offset = uio->uio_offset; 532 else 533 *offset += count - uio->uio_resid; 534 } 535 VOP_UNLOCK(vp, 0); 536 vn_finished_write(mp, 0); 537 return (error); 538 } 539 540 /* 541 * File table vnode stat routine. 542 */ 543 static int 544 vn_statfile(struct file *fp, struct stat *sb, struct proc *p) 545 { 546 struct vnode *vp = (struct vnode *)fp->f_data; 547 548 return vn_stat(vp, sb, p); 549 } 550 551 int 552 vn_stat(struct vnode *vp, struct stat *sb, struct proc *p) 553 { 554 struct vattr va; 555 int error; 556 mode_t mode; 557 558 error = VOP_GETATTR(vp, &va, p->p_ucred, p); 559 if (error) 560 return (error); 561 /* 562 * Copy from vattr table 563 */ 564 sb->st_dev = va.va_fsid; 565 sb->st_ino = va.va_fileid; 566 mode = va.va_mode; 567 switch (vp->v_type) { 568 case VREG: 569 mode |= S_IFREG; 570 break; 571 case VDIR: 572 mode |= S_IFDIR; 573 break; 574 case VBLK: 575 mode |= S_IFBLK; 576 break; 577 case VCHR: 578 mode |= S_IFCHR; 579 break; 580 case VLNK: 581 mode |= S_IFLNK; 582 break; 583 case VSOCK: 584 mode |= S_IFSOCK; 585 break; 586 case VFIFO: 587 mode |= S_IFIFO; 588 break; 589 default: 590 return (EBADF); 591 }; 592 sb->st_mode = mode; 593 sb->st_nlink = va.va_nlink; 594 sb->st_uid = va.va_uid; 595 sb->st_gid = va.va_gid; 596 sb->st_rdev = va.va_rdev; 597 sb->st_size = va.va_size; 598 sb->st_atimespec = va.va_atime; 599 sb->st_mtimespec = va.va_mtime; 600 sb->st_ctimespec = va.va_ctime; 601 sb->st_birthtimespec = va.va_birthtime; 602 sb->st_blksize = va.va_blocksize; 603 sb->st_flags = va.va_flags; 604 sb->st_gen = 0; 605 sb->st_blocks = va.va_bytes / S_BLKSIZE; 606 return (0); 607 } 608 609 /* 610 * File table vnode fcntl routine. 611 */ 612 static int 613 vn_fcntl(struct file *fp, u_int com, void *data, struct proc *p) 614 { 615 struct vnode *vp = ((struct vnode *)fp->f_data); 616 int error; 617 618 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 619 error = VOP_FCNTL(vp, com, data, fp->f_flag, p->p_ucred, p); 620 VOP_UNLOCK(vp, 0); 621 return (error); 622 } 623 624 /* 625 * File table vnode ioctl routine. 626 */ 627 static int 628 vn_ioctl(struct file *fp, u_long com, void *data, struct proc *p) 629 { 630 struct vnode *vp = ((struct vnode *)fp->f_data); 631 struct vattr vattr; 632 int error; 633 634 switch (vp->v_type) { 635 636 case VREG: 637 case VDIR: 638 if (com == FIONREAD) { 639 error = VOP_GETATTR(vp, &vattr, p->p_ucred, p); 640 if (error) 641 return (error); 642 *(int *)data = vattr.va_size - fp->f_offset; 643 return (0); 644 } 645 if ((com == FIONWRITE) || (com == FIONSPACE)) { 646 /* 647 * Files don't have send queues, so there never 648 * are any bytes in them, nor is there any 649 * open space in them. 650 */ 651 *(int *)data = 0; 652 return (0); 653 } 654 if (com == FIOGETBMAP) { 655 daddr_t *block; 656 657 if (*(daddr_t *)data < 0) 658 return (EINVAL); 659 block = (daddr_t *)data; 660 return (VOP_BMAP(vp, *block, NULL, block, NULL)); 661 } 662 if (com == OFIOGETBMAP) { 663 daddr_t ibn, obn; 664 665 if (*(int32_t *)data < 0) 666 return (EINVAL); 667 ibn = (daddr_t)*(int32_t *)data; 668 error = VOP_BMAP(vp, ibn, NULL, &obn, NULL); 669 *(int32_t *)data = (int32_t)obn; 670 return error; 671 } 672 if (com == FIONBIO || com == FIOASYNC) /* XXX */ 673 return (0); /* XXX */ 674 /* fall into ... */ 675 case VFIFO: 676 case VCHR: 677 case VBLK: 678 error = VOP_IOCTL(vp, com, data, fp->f_flag, p->p_ucred, p); 679 if (error == 0 && com == TIOCSCTTY) { 680 if (p->p_session->s_ttyvp) 681 vrele(p->p_session->s_ttyvp); 682 p->p_session->s_ttyvp = vp; 683 VREF(vp); 684 } 685 return (error); 686 687 default: 688 return (EPASSTHROUGH); 689 } 690 } 691 692 /* 693 * File table vnode poll routine. 694 */ 695 static int 696 vn_poll(struct file *fp, int events, struct proc *p) 697 { 698 699 return (VOP_POLL(((struct vnode *)fp->f_data), events, p)); 700 } 701 702 /* 703 * File table vnode kqfilter routine. 704 */ 705 int 706 vn_kqfilter(struct file *fp, struct knote *kn) 707 { 708 709 return (VOP_KQFILTER((struct vnode *)fp->f_data, kn)); 710 } 711 712 /* 713 * Check that the vnode is still valid, and if so 714 * acquire requested lock. 715 */ 716 int 717 vn_lock(struct vnode *vp, int flags) 718 { 719 int error; 720 721 #if 0 722 KASSERT(vp->v_usecount > 0 || (flags & LK_INTERLOCK) != 0 723 || (vp->v_flag & VONWORKLST) != 0); 724 #endif 725 726 do { 727 if ((flags & LK_INTERLOCK) == 0) 728 simple_lock(&vp->v_interlock); 729 if (vp->v_flag & VXLOCK) { 730 if (flags & LK_NOWAIT) { 731 simple_unlock(&vp->v_interlock); 732 return EBUSY; 733 } 734 vp->v_flag |= VXWANT; 735 ltsleep(vp, PINOD | PNORELOCK, 736 "vn_lock", 0, &vp->v_interlock); 737 error = ENOENT; 738 } else { 739 error = VOP_LOCK(vp, 740 (flags & ~LK_RETRY) | LK_INTERLOCK); 741 if (error == 0 || error == EDEADLK || error == EBUSY) 742 return (error); 743 } 744 flags &= ~LK_INTERLOCK; 745 } while (flags & LK_RETRY); 746 return (error); 747 } 748 749 /* 750 * File table vnode close routine. 751 */ 752 static int 753 vn_closefile(struct file *fp, struct proc *p) 754 { 755 756 return (vn_close(((struct vnode *)fp->f_data), fp->f_flag, 757 fp->f_cred, p)); 758 } 759 760 /* 761 * Enable LK_CANRECURSE on lock. Return prior status. 762 */ 763 u_int 764 vn_setrecurse(struct vnode *vp) 765 { 766 struct lock *lkp = &vp->v_lock; 767 u_int retval = lkp->lk_flags & LK_CANRECURSE; 768 769 lkp->lk_flags |= LK_CANRECURSE; 770 return retval; 771 } 772 773 /* 774 * Called when done with locksetrecurse. 775 */ 776 void 777 vn_restorerecurse(struct vnode *vp, u_int flags) 778 { 779 struct lock *lkp = &vp->v_lock; 780 781 lkp->lk_flags &= ~LK_CANRECURSE; 782 lkp->lk_flags |= flags; 783 } 784 785 int 786 vn_cow_establish(struct vnode *vp, 787 int (*func)(void *, struct buf *), void *cookie) 788 { 789 int s; 790 struct spec_cow_entry *e; 791 792 MALLOC(e, struct spec_cow_entry *, sizeof(struct spec_cow_entry), 793 M_DEVBUF, M_WAITOK); 794 e->ce_func = func; 795 e->ce_cookie = cookie; 796 797 SPEC_COW_LOCK(vp->v_specinfo, s); 798 vp->v_spec_cow_req++; 799 while (vp->v_spec_cow_count > 0) 800 ltsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 801 &vp->v_spec_cow_slock); 802 803 SLIST_INSERT_HEAD(&vp->v_spec_cow_head, e, ce_list); 804 805 vp->v_spec_cow_req--; 806 if (vp->v_spec_cow_req == 0) 807 wakeup(&vp->v_spec_cow_req); 808 SPEC_COW_UNLOCK(vp->v_specinfo, s); 809 810 return 0; 811 } 812 813 int 814 vn_cow_disestablish(struct vnode *vp, 815 int (*func)(void *, struct buf *), void *cookie) 816 { 817 int s; 818 struct spec_cow_entry *e; 819 820 SPEC_COW_LOCK(vp->v_specinfo, s); 821 vp->v_spec_cow_req++; 822 while (vp->v_spec_cow_count > 0) 823 ltsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 824 &vp->v_spec_cow_slock); 825 826 SLIST_FOREACH(e, &vp->v_spec_cow_head, ce_list) 827 if (e->ce_func == func && e->ce_cookie == cookie) { 828 SLIST_REMOVE(&vp->v_spec_cow_head, e, 829 spec_cow_entry, ce_list); 830 FREE(e, M_DEVBUF); 831 break; 832 } 833 834 vp->v_spec_cow_req--; 835 if (vp->v_spec_cow_req == 0) 836 wakeup(&vp->v_spec_cow_req); 837 SPEC_COW_UNLOCK(vp->v_specinfo, s); 838 839 return e ? 0 : EINVAL; 840 } 841 842 /* 843 * Simplified in-kernel wrapper calls for extended attribute access. 844 * Both calls pass in a NULL credential, authorizing a "kernel" access. 845 * Set IO_NODELOCKED in ioflg if the vnode is already locked. 846 */ 847 int 848 vn_extattr_get(struct vnode *vp, int ioflg, int attrnamespace, 849 const char *attrname, size_t *buflen, void *bf, struct proc *p) 850 { 851 struct uio auio; 852 struct iovec aiov; 853 int error; 854 855 aiov.iov_len = *buflen; 856 aiov.iov_base = bf; 857 858 auio.uio_iov = &aiov; 859 auio.uio_iovcnt = 1; 860 auio.uio_rw = UIO_READ; 861 auio.uio_segflg = UIO_SYSSPACE; 862 auio.uio_procp = p; 863 auio.uio_offset = 0; 864 auio.uio_resid = *buflen; 865 866 if ((ioflg & IO_NODELOCKED) == 0) 867 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 868 869 error = VOP_GETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, NULL, 870 p); 871 872 if ((ioflg & IO_NODELOCKED) == 0) 873 VOP_UNLOCK(vp, 0); 874 875 if (error == 0) 876 *buflen = *buflen - auio.uio_resid; 877 878 return (error); 879 } 880 881 /* 882 * XXX Failure mode if partially written? 883 */ 884 int 885 vn_extattr_set(struct vnode *vp, int ioflg, int attrnamespace, 886 const char *attrname, size_t buflen, const void *bf, struct proc *p) 887 { 888 struct uio auio; 889 struct iovec aiov; 890 struct mount *mp; 891 int error; 892 893 aiov.iov_len = buflen; 894 aiov.iov_base = __UNCONST(bf); /* XXXUNCONST kills const */ 895 896 auio.uio_iov = &aiov; 897 auio.uio_iovcnt = 1; 898 auio.uio_rw = UIO_WRITE; 899 auio.uio_segflg = UIO_SYSSPACE; 900 auio.uio_procp = p; 901 auio.uio_offset = 0; 902 auio.uio_resid = buflen; 903 904 if ((ioflg & IO_NODELOCKED) == 0) { 905 if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0) 906 return (error); 907 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 908 } 909 910 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, &auio, NULL, p); 911 912 if ((ioflg & IO_NODELOCKED) == 0) { 913 vn_finished_write(mp, 0); 914 VOP_UNLOCK(vp, 0); 915 } 916 917 return (error); 918 } 919 920 int 921 vn_extattr_rm(struct vnode *vp, int ioflg, int attrnamespace, 922 const char *attrname, struct proc *p) 923 { 924 struct mount *mp; 925 int error; 926 927 if ((ioflg & IO_NODELOCKED) == 0) { 928 if ((error = vn_start_write(vp, &mp, V_WAIT)) != 0) 929 return (error); 930 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 931 } 932 933 error = VOP_DELETEEXTATTR(vp, attrnamespace, attrname, NULL, p); 934 if (error == EOPNOTSUPP) 935 error = VOP_SETEXTATTR(vp, attrnamespace, attrname, NULL, 936 NULL, p); 937 938 if ((ioflg & IO_NODELOCKED) == 0) { 939 vn_finished_write(mp, 0); 940 VOP_UNLOCK(vp, 0); 941 } 942 943 return (error); 944 } 945 946 /* 947 * Preparing to start a filesystem write operation. If the operation is 948 * permitted, then we bump the count of operations in progress and 949 * proceed. If a suspend request is in progress, we wait until the 950 * suspension is over, and then proceed. 951 * V_PCATCH adds PCATCH to the tsleep flags. 952 * V_WAIT waits until suspension is over. Otherwise returns EWOULDBLOCK. 953 * V_SLEEPONLY wait, but do not bump the operations count. 954 * V_LOWER this is a lower level operation. No further vnodes should be 955 * locked. Otherwise it is a upper level operation. No vnodes 956 * should be locked. 957 */ 958 int 959 vn_start_write(struct vnode *vp, struct mount **mpp, int flags) 960 { 961 struct mount *mp; 962 int error, mask, prio; 963 964 /* 965 * If a vnode is provided, get and return the mount point that 966 * to which it will write. 967 */ 968 if (vp != NULL) { 969 *mpp = vp->v_mount; 970 } 971 if ((mp = *mpp) == NULL) 972 return (0); 973 mp = mp->mnt_leaf; 974 /* 975 * Check on status of suspension. 976 */ 977 prio = PUSER - 1; 978 if (flags & V_PCATCH) 979 prio |= PCATCH; 980 981 if ((flags & V_LOWER) == 0) 982 mask = IMNT_SUSPEND; 983 else 984 mask = IMNT_SUSPENDLOW; 985 986 while ((mp->mnt_iflag & mask) != 0) { 987 if ((flags & V_WAIT) == 0) 988 return (EWOULDBLOCK); 989 error = tsleep(&mp->mnt_flag, prio, "suspfs", 0); 990 if (error) 991 return (error); 992 } 993 if (flags & V_SLEEPONLY) 994 return (0); 995 simple_lock(&mp->mnt_slock); 996 if ((flags & V_LOWER) == 0) 997 mp->mnt_writeopcountupper++; 998 else 999 mp->mnt_writeopcountlower++; 1000 simple_unlock(&mp->mnt_slock); 1001 return (0); 1002 } 1003 1004 /* 1005 * Filesystem write operation has completed. If we are suspending and this 1006 * operation is the last one, notify the suspender that the suspension is 1007 * now in effect. 1008 */ 1009 void 1010 vn_finished_write(struct mount *mp, int flags) 1011 { 1012 if (mp == NULL) 1013 return; 1014 mp = mp->mnt_leaf; 1015 simple_lock(&mp->mnt_slock); 1016 if ((flags & V_LOWER) == 0) { 1017 mp->mnt_writeopcountupper--; 1018 if (mp->mnt_writeopcountupper < 0) 1019 printf("vn_finished_write: neg cnt upper=%d\n", 1020 mp->mnt_writeopcountupper); 1021 if ((mp->mnt_iflag & IMNT_SUSPEND) != 0 && 1022 mp->mnt_writeopcountupper <= 0) 1023 wakeup(&mp->mnt_writeopcountupper); 1024 } else { 1025 mp->mnt_writeopcountlower--; 1026 if (mp->mnt_writeopcountlower < 0) 1027 printf("vn_finished_write: neg cnt lower=%d\n", 1028 mp->mnt_writeopcountlower); 1029 if ((mp->mnt_iflag & IMNT_SUSPENDLOW) != 0 && 1030 mp->mnt_writeopcountupper <= 0) 1031 wakeup(&mp->mnt_writeopcountlower); 1032 } 1033 simple_unlock(&mp->mnt_slock); 1034 } 1035
Indexes created Sat Sep 20 22:09:52 GMT 2025