spec_vnops.c revision 1.88
1 /* $NetBSD: spec_vnops.c,v 1.88 2006/08/11 19:17:47 christos Exp $ */ 2 3 /* 4 * Copyright (c) 1989, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the University nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 * 31 * @(#)spec_vnops.c 8.15 (Berkeley) 7/14/95 32 */ 33 34 #if defined(_KERNEL_OPT) 35 #include "veriexec.h" 36 #endif 37 38 #include <sys/cdefs.h> 39 __KERNEL_RCSID(0, "$NetBSD: spec_vnops.c,v 1.88 2006/08/11 19:17:47 christos Exp $"); 40 41 #include <sys/param.h> 42 #include <sys/proc.h> 43 #include <sys/systm.h> 44 #include <sys/kernel.h> 45 #include <sys/conf.h> 46 #include <sys/buf.h> 47 #include <sys/mount.h> 48 #include <sys/namei.h> 49 #include <sys/vnode.h> 50 #include <sys/stat.h> 51 #include <sys/errno.h> 52 #include <sys/ioctl.h> 53 #include <sys/poll.h> 54 #include <sys/file.h> 55 #include <sys/disklabel.h> 56 #include <sys/lockf.h> 57 #include <sys/tty.h> 58 #include <sys/kauth.h> 59 60 #include <miscfs/genfs/genfs.h> 61 #include <miscfs/specfs/specdev.h> 62 63 #if NVERIEXEC > 0 64 #include <sys/verified_exec.h> 65 #endif /* NVERIEXEC > 0 */ 66 67 /* symbolic sleep message strings for devices */ 68 const char devopn[] = "devopn"; 69 const char devio[] = "devio"; 70 const char devwait[] = "devwait"; 71 const char devin[] = "devin"; 72 const char devout[] = "devout"; 73 const char devioc[] = "devioc"; 74 const char devcls[] = "devcls"; 75 76 struct vnode *speclisth[SPECHSZ]; 77 78 /* 79 * This vnode operations vector is used for two things only: 80 * - special device nodes created from whole cloth by the kernel. 81 * - as a temporary vnodeops replacement for vnodes which were found to 82 * be aliased by callers of checkalias(). 83 * For the ops vector for vnodes built from special devices found in a 84 * filesystem, see (e.g) ffs_specop_entries[] in ffs_vnops.c or the 85 * equivalent for other filesystems. 86 */ 87 88 int (**spec_vnodeop_p)(void *); 89 const struct vnodeopv_entry_desc spec_vnodeop_entries[] = { 90 { &vop_default_desc, vn_default_error }, 91 { &vop_lookup_desc, spec_lookup }, /* lookup */ 92 { &vop_create_desc, spec_create }, /* create */ 93 { &vop_mknod_desc, spec_mknod }, /* mknod */ 94 { &vop_open_desc, spec_open }, /* open */ 95 { &vop_close_desc, spec_close }, /* close */ 96 { &vop_access_desc, spec_access }, /* access */ 97 { &vop_getattr_desc, spec_getattr }, /* getattr */ 98 { &vop_setattr_desc, spec_setattr }, /* setattr */ 99 { &vop_read_desc, spec_read }, /* read */ 100 { &vop_write_desc, spec_write }, /* write */ 101 { &vop_lease_desc, spec_lease_check }, /* lease */ 102 { &vop_fcntl_desc, spec_fcntl }, /* fcntl */ 103 { &vop_ioctl_desc, spec_ioctl }, /* ioctl */ 104 { &vop_poll_desc, spec_poll }, /* poll */ 105 { &vop_kqfilter_desc, spec_kqfilter }, /* kqfilter */ 106 { &vop_revoke_desc, spec_revoke }, /* revoke */ 107 { &vop_mmap_desc, spec_mmap }, /* mmap */ 108 { &vop_fsync_desc, spec_fsync }, /* fsync */ 109 { &vop_seek_desc, spec_seek }, /* seek */ 110 { &vop_remove_desc, spec_remove }, /* remove */ 111 { &vop_link_desc, spec_link }, /* link */ 112 { &vop_rename_desc, spec_rename }, /* rename */ 113 { &vop_mkdir_desc, spec_mkdir }, /* mkdir */ 114 { &vop_rmdir_desc, spec_rmdir }, /* rmdir */ 115 { &vop_symlink_desc, spec_symlink }, /* symlink */ 116 { &vop_readdir_desc, spec_readdir }, /* readdir */ 117 { &vop_readlink_desc, spec_readlink }, /* readlink */ 118 { &vop_abortop_desc, spec_abortop }, /* abortop */ 119 { &vop_inactive_desc, spec_inactive }, /* inactive */ 120 { &vop_reclaim_desc, spec_reclaim }, /* reclaim */ 121 { &vop_lock_desc, spec_lock }, /* lock */ 122 { &vop_unlock_desc, spec_unlock }, /* unlock */ 123 { &vop_bmap_desc, spec_bmap }, /* bmap */ 124 { &vop_strategy_desc, spec_strategy }, /* strategy */ 125 { &vop_print_desc, spec_print }, /* print */ 126 { &vop_islocked_desc, spec_islocked }, /* islocked */ 127 { &vop_pathconf_desc, spec_pathconf }, /* pathconf */ 128 { &vop_advlock_desc, spec_advlock }, /* advlock */ 129 { &vop_bwrite_desc, spec_bwrite }, /* bwrite */ 130 { &vop_getpages_desc, spec_getpages }, /* getpages */ 131 { &vop_putpages_desc, spec_putpages }, /* putpages */ 132 { NULL, NULL } 133 }; 134 const struct vnodeopv_desc spec_vnodeop_opv_desc = 135 { &spec_vnodeop_p, spec_vnodeop_entries }; 136 137 /* 138 * Trivial lookup routine that always fails. 139 */ 140 int 141 spec_lookup(v) 142 void *v; 143 { 144 struct vop_lookup_args /* { 145 struct vnode *a_dvp; 146 struct vnode **a_vpp; 147 struct componentname *a_cnp; 148 } */ *ap = v; 149 150 *ap->a_vpp = NULL; 151 return (ENOTDIR); 152 } 153 154 /* 155 * Returns true if dev is /dev/mem or /dev/kmem. 156 */ 157 static int 158 iskmemdev(dev_t dev) 159 { 160 /* mem_no is emitted by config(8) to generated devsw.c */ 161 extern const int mem_no; 162 163 /* minor 14 is /dev/io on i386 with COMPAT_10 */ 164 return (major(dev) == mem_no && (minor(dev) < 2 || minor(dev) == 14)); 165 } 166 167 /* 168 * Open a special file. 169 */ 170 /* ARGSUSED */ 171 int 172 spec_open(v) 173 void *v; 174 { 175 struct vop_open_args /* { 176 struct vnode *a_vp; 177 int a_mode; 178 kauth_cred_t a_cred; 179 struct lwp *a_l; 180 } */ *ap = v; 181 struct lwp *l = ap->a_l; 182 struct vnode *bvp, *vp = ap->a_vp; 183 const struct bdevsw *bdev; 184 const struct cdevsw *cdev; 185 dev_t blkdev, dev = (dev_t)vp->v_rdev; 186 int error; 187 struct partinfo pi; 188 int (*d_ioctl)(dev_t, u_long, caddr_t, int, struct lwp *); 189 190 /* 191 * Don't allow open if fs is mounted -nodev. 192 */ 193 if (vp->v_mount && (vp->v_mount->mnt_flag & MNT_NODEV)) 194 return (ENXIO); 195 196 switch (vp->v_type) { 197 198 case VCHR: 199 cdev = cdevsw_lookup(dev); 200 if (cdev == NULL) 201 return (ENXIO); 202 if (ap->a_cred != FSCRED && (ap->a_mode & FWRITE)) { 203 /* 204 * When running in very secure mode, do not allow 205 * opens for writing of any disk character devices. 206 */ 207 if (securelevel >= 2 && cdev->d_type == D_DISK) 208 return (EPERM); 209 /* 210 * When running in secure mode, do not allow opens 211 * for writing of /dev/mem, /dev/kmem, or character 212 * devices whose corresponding block devices are 213 * currently mounted. 214 */ 215 bvp = NULL; 216 if (securelevel >= 1) { 217 blkdev = devsw_chr2blk(dev); 218 if (blkdev != (dev_t)NODEV && 219 vfinddev(blkdev, VBLK, &bvp) && 220 (error = vfs_mountedon(bvp))) 221 return (error); 222 if (iskmemdev(dev)) 223 return (EPERM); 224 } 225 226 #if NVERIEXEC > 0 227 if (veriexec_strict >= VERIEXEC_IPS && iskmemdev(dev)) 228 return (error); 229 error = veriexec_rawchk(bvp); 230 if (error) 231 return (error); 232 #endif /* NVERIEXEC > 0 */ 233 } 234 if (cdev->d_type == D_TTY) 235 vp->v_flag |= VISTTY; 236 VOP_UNLOCK(vp, 0); 237 error = (*cdev->d_open)(dev, ap->a_mode, S_IFCHR, l); 238 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 239 if (cdev->d_type != D_DISK) 240 return error; 241 d_ioctl = cdev->d_ioctl; 242 break; 243 244 case VBLK: 245 bdev = bdevsw_lookup(dev); 246 if (bdev == NULL) 247 return (ENXIO); 248 /* 249 * When running in very secure mode, do not allow 250 * opens for writing of any disk block devices. 251 */ 252 if (securelevel >= 2 && ap->a_cred != FSCRED && 253 (ap->a_mode & FWRITE) && bdev->d_type == D_DISK) 254 return (EPERM); 255 /* 256 * Do not allow opens of block devices that are 257 * currently mounted. 258 */ 259 if ((error = vfs_mountedon(vp)) != 0) 260 return (error); 261 262 #if NVERIEXEC > 0 263 error = veriexec_rawchk(vp); 264 if (error) 265 return (error); 266 #endif /* NVERIEXEC > 0 */ 267 268 error = (*bdev->d_open)(dev, ap->a_mode, S_IFBLK, l); 269 d_ioctl = bdev->d_ioctl; 270 break; 271 272 case VNON: 273 case VLNK: 274 case VDIR: 275 case VREG: 276 case VBAD: 277 case VFIFO: 278 case VSOCK: 279 default: 280 return 0; 281 } 282 283 if (error) 284 return error; 285 if (!(*d_ioctl)(vp->v_rdev, DIOCGPART, (caddr_t)&pi, FREAD, curlwp)) 286 vp->v_size = (voff_t)pi.disklab->d_secsize * pi.part->p_size; 287 return 0; 288 } 289 290 /* 291 * Vnode op for read 292 */ 293 /* ARGSUSED */ 294 int 295 spec_read(v) 296 void *v; 297 { 298 struct vop_read_args /* { 299 struct vnode *a_vp; 300 struct uio *a_uio; 301 int a_ioflag; 302 kauth_cred_t a_cred; 303 } */ *ap = v; 304 struct vnode *vp = ap->a_vp; 305 struct uio *uio = ap->a_uio; 306 struct lwp *l = curlwp; 307 struct buf *bp; 308 const struct bdevsw *bdev; 309 const struct cdevsw *cdev; 310 daddr_t bn; 311 int bsize, bscale; 312 struct partinfo dpart; 313 int n, on; 314 int error = 0; 315 316 #ifdef DIAGNOSTIC 317 if (uio->uio_rw != UIO_READ) 318 panic("spec_read mode"); 319 if (&uio->uio_vmspace->vm_map != kernel_map && 320 uio->uio_vmspace != curproc->p_vmspace) 321 panic("spec_read proc"); 322 #endif 323 if (uio->uio_resid == 0) 324 return (0); 325 326 switch (vp->v_type) { 327 328 case VCHR: 329 VOP_UNLOCK(vp, 0); 330 cdev = cdevsw_lookup(vp->v_rdev); 331 if (cdev != NULL) 332 error = (*cdev->d_read)(vp->v_rdev, uio, ap->a_ioflag); 333 else 334 error = ENXIO; 335 vn_lock(vp, LK_SHARED | LK_RETRY); 336 return (error); 337 338 case VBLK: 339 if (uio->uio_offset < 0) 340 return (EINVAL); 341 bsize = BLKDEV_IOSIZE; 342 bdev = bdevsw_lookup(vp->v_rdev); 343 if (bdev != NULL && 344 (*bdev->d_ioctl)(vp->v_rdev, DIOCGPART, (caddr_t)&dpart, 345 FREAD, l) == 0) { 346 if (dpart.part->p_fstype == FS_BSDFFS && 347 dpart.part->p_frag != 0 && dpart.part->p_fsize != 0) 348 bsize = dpart.part->p_frag * 349 dpart.part->p_fsize; 350 } 351 bscale = bsize >> DEV_BSHIFT; 352 do { 353 bn = (uio->uio_offset >> DEV_BSHIFT) &~ (bscale - 1); 354 on = uio->uio_offset % bsize; 355 n = min((unsigned)(bsize - on), uio->uio_resid); 356 error = bread(vp, bn, bsize, NOCRED, &bp); 357 n = min(n, bsize - bp->b_resid); 358 if (error) { 359 brelse(bp); 360 return (error); 361 } 362 error = uiomove((char *)bp->b_data + on, n, uio); 363 brelse(bp); 364 } while (error == 0 && uio->uio_resid > 0 && n != 0); 365 return (error); 366 367 default: 368 panic("spec_read type"); 369 } 370 /* NOTREACHED */ 371 } 372 373 /* 374 * Vnode op for write 375 */ 376 /* ARGSUSED */ 377 int 378 spec_write(v) 379 void *v; 380 { 381 struct vop_write_args /* { 382 struct vnode *a_vp; 383 struct uio *a_uio; 384 int a_ioflag; 385 kauth_cred_t a_cred; 386 } */ *ap = v; 387 struct vnode *vp = ap->a_vp; 388 struct uio *uio = ap->a_uio; 389 struct lwp *l = curlwp; 390 struct buf *bp; 391 const struct bdevsw *bdev; 392 const struct cdevsw *cdev; 393 daddr_t bn; 394 int bsize, bscale; 395 struct partinfo dpart; 396 int n, on; 397 int error = 0; 398 399 #ifdef DIAGNOSTIC 400 if (uio->uio_rw != UIO_WRITE) 401 panic("spec_write mode"); 402 if (&uio->uio_vmspace->vm_map != kernel_map && 403 uio->uio_vmspace != curproc->p_vmspace) 404 panic("spec_write proc"); 405 #endif 406 407 switch (vp->v_type) { 408 409 case VCHR: 410 VOP_UNLOCK(vp, 0); 411 cdev = cdevsw_lookup(vp->v_rdev); 412 if (cdev != NULL) 413 error = (*cdev->d_write)(vp->v_rdev, uio, ap->a_ioflag); 414 else 415 error = ENXIO; 416 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 417 return (error); 418 419 case VBLK: 420 if (uio->uio_resid == 0) 421 return (0); 422 if (uio->uio_offset < 0) 423 return (EINVAL); 424 bsize = BLKDEV_IOSIZE; 425 bdev = bdevsw_lookup(vp->v_rdev); 426 if (bdev != NULL && 427 (*bdev->d_ioctl)(vp->v_rdev, DIOCGPART, (caddr_t)&dpart, 428 FREAD, l) == 0) { 429 if (dpart.part->p_fstype == FS_BSDFFS && 430 dpart.part->p_frag != 0 && dpart.part->p_fsize != 0) 431 bsize = dpart.part->p_frag * 432 dpart.part->p_fsize; 433 } 434 bscale = bsize >> DEV_BSHIFT; 435 do { 436 bn = (uio->uio_offset >> DEV_BSHIFT) &~ (bscale - 1); 437 on = uio->uio_offset % bsize; 438 n = min((unsigned)(bsize - on), uio->uio_resid); 439 if (n == bsize) 440 bp = getblk(vp, bn, bsize, 0, 0); 441 else 442 error = bread(vp, bn, bsize, NOCRED, &bp); 443 if (error) { 444 brelse(bp); 445 return (error); 446 } 447 n = min(n, bsize - bp->b_resid); 448 error = uiomove((char *)bp->b_data + on, n, uio); 449 if (error) 450 brelse(bp); 451 else { 452 if (n + on == bsize) 453 bawrite(bp); 454 else 455 bdwrite(bp); 456 if (bp->b_flags & B_ERROR) 457 error = bp->b_error; 458 } 459 } while (error == 0 && uio->uio_resid > 0 && n != 0); 460 return (error); 461 462 default: 463 panic("spec_write type"); 464 } 465 /* NOTREACHED */ 466 } 467 468 /* 469 * Device ioctl operation. 470 */ 471 /* ARGSUSED */ 472 int 473 spec_ioctl(v) 474 void *v; 475 { 476 struct vop_ioctl_args /* { 477 struct vnode *a_vp; 478 u_long a_command; 479 void *a_data; 480 int a_fflag; 481 kauth_cred_t a_cred; 482 struct lwp *a_l; 483 } */ *ap = v; 484 const struct bdevsw *bdev; 485 const struct cdevsw *cdev; 486 struct vnode *vp; 487 dev_t dev; 488 489 /* 490 * Extract all the info we need from the vnode, taking care to 491 * avoid a race with VOP_REVOKE(). 492 */ 493 494 vp = ap->a_vp; 495 dev = NODEV; 496 simple_lock(&vp->v_interlock); 497 if ((vp->v_flag & VXLOCK) == 0 && vp->v_specinfo) { 498 dev = vp->v_rdev; 499 } 500 simple_unlock(&vp->v_interlock); 501 if (dev == NODEV) { 502 return ENXIO; 503 } 504 505 switch (vp->v_type) { 506 507 case VCHR: 508 cdev = cdevsw_lookup(dev); 509 if (cdev == NULL) 510 return (ENXIO); 511 return ((*cdev->d_ioctl)(dev, ap->a_command, ap->a_data, 512 ap->a_fflag, ap->a_l)); 513 514 case VBLK: 515 bdev = bdevsw_lookup(dev); 516 if (bdev == NULL) 517 return (ENXIO); 518 if (ap->a_command == 0 && (long)ap->a_data == B_TAPE) { 519 if (bdev->d_type == D_TAPE) 520 return (0); 521 else 522 return (1); 523 } 524 return ((*bdev->d_ioctl)(dev, ap->a_command, ap->a_data, 525 ap->a_fflag, ap->a_l)); 526 527 default: 528 panic("spec_ioctl"); 529 /* NOTREACHED */ 530 } 531 } 532 533 /* ARGSUSED */ 534 int 535 spec_poll(v) 536 void *v; 537 { 538 struct vop_poll_args /* { 539 struct vnode *a_vp; 540 int a_events; 541 struct lwp *a_l; 542 } */ *ap = v; 543 const struct cdevsw *cdev; 544 dev_t dev; 545 546 switch (ap->a_vp->v_type) { 547 548 case VCHR: 549 dev = ap->a_vp->v_rdev; 550 cdev = cdevsw_lookup(dev); 551 if (cdev == NULL) 552 return (POLLERR); 553 return (*cdev->d_poll)(dev, ap->a_events, ap->a_l); 554 555 default: 556 return (genfs_poll(v)); 557 } 558 } 559 560 /* ARGSUSED */ 561 int 562 spec_kqfilter(v) 563 void *v; 564 { 565 struct vop_kqfilter_args /* { 566 struct vnode *a_vp; 567 struct proc *a_kn; 568 } */ *ap = v; 569 const struct cdevsw *cdev; 570 dev_t dev; 571 572 switch (ap->a_vp->v_type) { 573 574 case VCHR: 575 dev = ap->a_vp->v_rdev; 576 cdev = cdevsw_lookup(dev); 577 if (cdev == NULL) 578 return (ENXIO); 579 return (*cdev->d_kqfilter)(dev, ap->a_kn); 580 default: 581 /* 582 * Block devices don't support kqfilter, and refuse it 583 * for any other files (like those vflush()ed) too. 584 */ 585 return (EOPNOTSUPP); 586 } 587 } 588 589 /* 590 * Synch buffers associated with a block device 591 */ 592 /* ARGSUSED */ 593 int 594 spec_fsync(v) 595 void *v; 596 { 597 struct vop_fsync_args /* { 598 struct vnode *a_vp; 599 kauth_cred_t a_cred; 600 int a_flags; 601 off_t offlo; 602 off_t offhi; 603 struct lwp *a_l; 604 } */ *ap = v; 605 struct vnode *vp = ap->a_vp; 606 607 if (vp->v_type == VBLK) 608 vflushbuf(vp, (ap->a_flags & FSYNC_WAIT) != 0); 609 return (0); 610 } 611 612 /* 613 * Just call the device strategy routine 614 */ 615 int 616 spec_strategy(v) 617 void *v; 618 { 619 struct vop_strategy_args /* { 620 struct vnode *a_vp; 621 struct buf *a_bp; 622 } */ *ap = v; 623 struct vnode *vp = ap->a_vp; 624 struct buf *bp = ap->a_bp; 625 int error, s; 626 struct spec_cow_entry *e; 627 628 error = 0; 629 bp->b_dev = vp->v_rdev; 630 if (!(bp->b_flags & B_READ) && 631 (LIST_FIRST(&bp->b_dep)) != NULL && bioops.io_start) 632 (*bioops.io_start)(bp); 633 634 if (!(bp->b_flags & B_READ) && !SLIST_EMPTY(&vp->v_spec_cow_head)) { 635 SPEC_COW_LOCK(vp->v_specinfo, s); 636 while (vp->v_spec_cow_req > 0) 637 ltsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 638 &vp->v_spec_cow_slock); 639 vp->v_spec_cow_count++; 640 SPEC_COW_UNLOCK(vp->v_specinfo, s); 641 642 SLIST_FOREACH(e, &vp->v_spec_cow_head, ce_list) { 643 if ((error = (*e->ce_func)(e->ce_cookie, bp)) != 0) 644 break; 645 } 646 647 SPEC_COW_LOCK(vp->v_specinfo, s); 648 vp->v_spec_cow_count--; 649 if (vp->v_spec_cow_req && vp->v_spec_cow_count == 0) 650 wakeup(&vp->v_spec_cow_req); 651 SPEC_COW_UNLOCK(vp->v_specinfo, s); 652 } 653 654 if (error) { 655 bp->b_error = error; 656 bp->b_flags |= B_ERROR; 657 biodone(bp); 658 return (error); 659 } 660 661 DEV_STRATEGY(bp); 662 663 return (0); 664 } 665 666 int 667 spec_inactive(v) 668 void *v; 669 { 670 struct vop_inactive_args /* { 671 struct vnode *a_vp; 672 struct proc *a_l; 673 } */ *ap = v; 674 675 VOP_UNLOCK(ap->a_vp, 0); 676 return (0); 677 } 678 679 /* 680 * This is a noop, simply returning what one has been given. 681 */ 682 int 683 spec_bmap(v) 684 void *v; 685 { 686 struct vop_bmap_args /* { 687 struct vnode *a_vp; 688 daddr_t a_bn; 689 struct vnode **a_vpp; 690 daddr_t *a_bnp; 691 int *a_runp; 692 } */ *ap = v; 693 694 if (ap->a_vpp != NULL) 695 *ap->a_vpp = ap->a_vp; 696 if (ap->a_bnp != NULL) 697 *ap->a_bnp = ap->a_bn; 698 if (ap->a_runp != NULL) 699 *ap->a_runp = (MAXBSIZE >> DEV_BSHIFT) - 1; 700 return (0); 701 } 702 703 /* 704 * Device close routine 705 */ 706 /* ARGSUSED */ 707 int 708 spec_close(v) 709 void *v; 710 { 711 struct vop_close_args /* { 712 struct vnode *a_vp; 713 int a_fflag; 714 kauth_cred_t a_cred; 715 struct lwp *a_l; 716 } */ *ap = v; 717 struct vnode *vp = ap->a_vp; 718 const struct bdevsw *bdev; 719 const struct cdevsw *cdev; 720 struct session *sess; 721 dev_t dev = vp->v_rdev; 722 int (*devclose)(dev_t, int, int, struct lwp *); 723 int mode, error, count, flags, flags1; 724 725 count = vcount(vp); 726 flags = vp->v_flag; 727 728 switch (vp->v_type) { 729 730 case VCHR: 731 /* 732 * Hack: a tty device that is a controlling terminal 733 * has a reference from the session structure. 734 * We cannot easily tell that a character device is 735 * a controlling terminal, unless it is the closing 736 * process' controlling terminal. In that case, 737 * if the reference count is 2 (this last descriptor 738 * plus the session), release the reference from the session. 739 * Also remove the link from the tty back to the session 740 * and pgrp - due to the way consoles are handled we cannot 741 * guarantee that the vrele() will do the final close on the 742 * actual tty device. 743 */ 744 if (count == 2 && ap->a_l && 745 vp == (sess = ap->a_l->l_proc->p_session)->s_ttyvp) { 746 sess->s_ttyvp = NULL; 747 if (sess->s_ttyp->t_session != NULL) { 748 sess->s_ttyp->t_pgrp = NULL; 749 sess->s_ttyp->t_session = NULL; 750 SESSRELE(sess); 751 } else if (sess->s_ttyp->t_pgrp != NULL) 752 panic("spec_close: spurious pgrp ref"); 753 vrele(vp); 754 count--; 755 } 756 /* 757 * If the vnode is locked, then we are in the midst 758 * of forcably closing the device, otherwise we only 759 * close on last reference. 760 */ 761 if (count > 1 && (flags & VXLOCK) == 0) 762 return (0); 763 cdev = cdevsw_lookup(dev); 764 if (cdev != NULL) 765 devclose = cdev->d_close; 766 else 767 devclose = NULL; 768 mode = S_IFCHR; 769 break; 770 771 case VBLK: 772 /* 773 * On last close of a block device (that isn't mounted) 774 * we must invalidate any in core blocks, so that 775 * we can, for instance, change floppy disks. 776 */ 777 error = vinvalbuf(vp, V_SAVE, ap->a_cred, ap->a_l, 0, 0); 778 if (error) 779 return (error); 780 /* 781 * We do not want to really close the device if it 782 * is still in use unless we are trying to close it 783 * forcibly. Since every use (buffer, vnode, swap, cmap) 784 * holds a reference to the vnode, and because we mark 785 * any other vnodes that alias this device, when the 786 * sum of the reference counts on all the aliased 787 * vnodes descends to one, we are on last close. 788 */ 789 if (count > 1 && (flags & VXLOCK) == 0) 790 return (0); 791 bdev = bdevsw_lookup(dev); 792 if (bdev != NULL) 793 devclose = bdev->d_close; 794 else 795 devclose = NULL; 796 mode = S_IFBLK; 797 break; 798 799 default: 800 panic("spec_close: not special"); 801 } 802 803 flags1 = ap->a_fflag; 804 805 /* 806 * if VXLOCK is set, then we're going away soon, so make this 807 * non-blocking. Also ensures that we won't wedge in vn_lock below. 808 */ 809 if (flags & VXLOCK) 810 flags1 |= FNONBLOCK; 811 812 /* 813 * If we're able to block, release the vnode lock & reacquire. We 814 * might end up sleeping for someone else who wants our queues. They 815 * won't get them if we hold the vnode locked. Also, if VXLOCK is set, 816 * don't release the lock as we won't be able to regain it. 817 */ 818 if (!(flags1 & FNONBLOCK)) 819 VOP_UNLOCK(vp, 0); 820 821 if (devclose != NULL) 822 error = (*devclose)(dev, flags1, mode, ap->a_l); 823 else 824 error = ENXIO; 825 826 if (!(flags1 & FNONBLOCK)) 827 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 828 829 return (error); 830 } 831 832 /* 833 * Print out the contents of a special device vnode. 834 */ 835 int 836 spec_print(v) 837 void *v; 838 { 839 struct vop_print_args /* { 840 struct vnode *a_vp; 841 } */ *ap = v; 842 843 printf("tag VT_NON, dev %d, %d\n", major(ap->a_vp->v_rdev), 844 minor(ap->a_vp->v_rdev)); 845 return 0; 846 } 847 848 /* 849 * Return POSIX pathconf information applicable to special devices. 850 */ 851 int 852 spec_pathconf(v) 853 void *v; 854 { 855 struct vop_pathconf_args /* { 856 struct vnode *a_vp; 857 int a_name; 858 register_t *a_retval; 859 } */ *ap = v; 860 861 switch (ap->a_name) { 862 case _PC_LINK_MAX: 863 *ap->a_retval = LINK_MAX; 864 return (0); 865 case _PC_MAX_CANON: 866 *ap->a_retval = MAX_CANON; 867 return (0); 868 case _PC_MAX_INPUT: 869 *ap->a_retval = MAX_INPUT; 870 return (0); 871 case _PC_PIPE_BUF: 872 *ap->a_retval = PIPE_BUF; 873 return (0); 874 case _PC_CHOWN_RESTRICTED: 875 *ap->a_retval = 1; 876 return (0); 877 case _PC_VDISABLE: 878 *ap->a_retval = _POSIX_VDISABLE; 879 return (0); 880 case _PC_SYNC_IO: 881 *ap->a_retval = 1; 882 return (0); 883 default: 884 return (EINVAL); 885 } 886 /* NOTREACHED */ 887 } 888 889 /* 890 * Advisory record locking support. 891 */ 892 int 893 spec_advlock(v) 894 void *v; 895 { 896 struct vop_advlock_args /* { 897 struct vnode *a_vp; 898 void *a_id; 899 int a_op; 900 struct flock *a_fl; 901 int a_flags; 902 } */ *ap = v; 903 struct vnode *vp = ap->a_vp; 904 905 return lf_advlock(ap, &vp->v_speclockf, (off_t)0); 906 } 907
Indexes created Tue Sep 23 07:09:52 GMT 2025