spec_vnops.c revision 1.89
1 /* $NetBSD: spec_vnops.c,v 1.89 2006/09/08 20:58:57 elad 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 1.89 2006/09/08 20:58:57 elad 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 #define M2K(m) (((m) & FREAD) && ((m) & FWRITE) ? KAUTH_REQ_SYSTEM_RAWIO_RW : \ 197 (m) & FWRITE ? KAUTH_REQ_SYSTEM_RAWIO_WRITE : \ 198 KAUTH_REQ_SYSTEM_RAWIO_READ) 199 200 switch (vp->v_type) { 201 202 case VCHR: 203 cdev = cdevsw_lookup(dev); 204 if (cdev == NULL) 205 return (ENXIO); 206 207 if (ap->a_cred != FSCRED) { 208 u_long rw; 209 210 rw = M2K(ap->a_mode); 211 error = 0; 212 bvp = NULL; 213 214 /* XXX we're holding a vnode lock here */ 215 if (iskmemdev(dev)) { 216 error = kauth_authorize_system(ap->a_cred, 217 KAUTH_SYSTEM_RAWIO, 218 KAUTH_REQ_SYSTEM_RAWIO_MEMORY, 219 (void *)rw, NULL, NULL); 220 } else { 221 blkdev = devsw_chr2blk(dev); 222 if (blkdev != (dev_t)NODEV) { 223 vfinddev(blkdev, VBLK, &bvp); 224 error = kauth_authorize_system(ap->a_cred, 225 KAUTH_SYSTEM_RAWIO, 226 KAUTH_REQ_SYSTEM_RAWIO_DISK, 227 (void *)rw, vp, (void *)(u_long)dev); 228 if (error) printf("nope.\n"); 229 } 230 } 231 232 if (error) 233 return (error); 234 235 #if NVERIEXEC > 0 236 if (veriexec_strict >= VERIEXEC_IPS && iskmemdev(dev)) 237 return (error); 238 error = veriexec_rawchk(bvp); 239 if (error) 240 return (error); 241 #endif /* NVERIEXEC > 0 */ 242 } 243 244 if (cdev->d_type == D_TTY) 245 vp->v_flag |= VISTTY; 246 VOP_UNLOCK(vp, 0); 247 error = (*cdev->d_open)(dev, ap->a_mode, S_IFCHR, l); 248 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 249 if (cdev->d_type != D_DISK) 250 return error; 251 d_ioctl = cdev->d_ioctl; 252 break; 253 254 case VBLK: 255 bdev = bdevsw_lookup(dev); 256 if (bdev == NULL) 257 return (ENXIO); 258 259 /* 260 * When running in very secure mode, do not allow 261 * opens for writing of any disk block devices. 262 */ 263 if (ap->a_cred != FSCRED) { 264 u_long rw; 265 266 rw = M2K(ap->a_mode); 267 268 error = kauth_authorize_system(ap->a_cred, 269 KAUTH_SYSTEM_RAWIO, 270 KAUTH_REQ_SYSTEM_RAWIO_DISK, 271 (void *)rw, vp, (void *)(u_long)dev); 272 if (error) 273 return (error); 274 } 275 276 #if NVERIEXEC > 0 277 error = veriexec_rawchk(vp); 278 if (error) 279 return (error); 280 #endif /* NVERIEXEC > 0 */ 281 282 error = (*bdev->d_open)(dev, ap->a_mode, S_IFBLK, l); 283 d_ioctl = bdev->d_ioctl; 284 break; 285 286 case VNON: 287 case VLNK: 288 case VDIR: 289 case VREG: 290 case VBAD: 291 case VFIFO: 292 case VSOCK: 293 default: 294 return 0; 295 } 296 297 #undef M2K 298 299 if (error) 300 return error; 301 if (!(*d_ioctl)(vp->v_rdev, DIOCGPART, (caddr_t)&pi, FREAD, curlwp)) 302 vp->v_size = (voff_t)pi.disklab->d_secsize * pi.part->p_size; 303 return 0; 304 } 305 306 /* 307 * Vnode op for read 308 */ 309 /* ARGSUSED */ 310 int 311 spec_read(v) 312 void *v; 313 { 314 struct vop_read_args /* { 315 struct vnode *a_vp; 316 struct uio *a_uio; 317 int a_ioflag; 318 kauth_cred_t a_cred; 319 } */ *ap = v; 320 struct vnode *vp = ap->a_vp; 321 struct uio *uio = ap->a_uio; 322 struct lwp *l = curlwp; 323 struct buf *bp; 324 const struct bdevsw *bdev; 325 const struct cdevsw *cdev; 326 daddr_t bn; 327 int bsize, bscale; 328 struct partinfo dpart; 329 int n, on; 330 int error = 0; 331 332 #ifdef DIAGNOSTIC 333 if (uio->uio_rw != UIO_READ) 334 panic("spec_read mode"); 335 if (&uio->uio_vmspace->vm_map != kernel_map && 336 uio->uio_vmspace != curproc->p_vmspace) 337 panic("spec_read proc"); 338 #endif 339 if (uio->uio_resid == 0) 340 return (0); 341 342 switch (vp->v_type) { 343 344 case VCHR: 345 VOP_UNLOCK(vp, 0); 346 cdev = cdevsw_lookup(vp->v_rdev); 347 if (cdev != NULL) 348 error = (*cdev->d_read)(vp->v_rdev, uio, ap->a_ioflag); 349 else 350 error = ENXIO; 351 vn_lock(vp, LK_SHARED | LK_RETRY); 352 return (error); 353 354 case VBLK: 355 if (uio->uio_offset < 0) 356 return (EINVAL); 357 bsize = BLKDEV_IOSIZE; 358 bdev = bdevsw_lookup(vp->v_rdev); 359 if (bdev != NULL && 360 (*bdev->d_ioctl)(vp->v_rdev, DIOCGPART, (caddr_t)&dpart, 361 FREAD, l) == 0) { 362 if (dpart.part->p_fstype == FS_BSDFFS && 363 dpart.part->p_frag != 0 && dpart.part->p_fsize != 0) 364 bsize = dpart.part->p_frag * 365 dpart.part->p_fsize; 366 } 367 bscale = bsize >> DEV_BSHIFT; 368 do { 369 bn = (uio->uio_offset >> DEV_BSHIFT) &~ (bscale - 1); 370 on = uio->uio_offset % bsize; 371 n = min((unsigned)(bsize - on), uio->uio_resid); 372 error = bread(vp, bn, bsize, NOCRED, &bp); 373 n = min(n, bsize - bp->b_resid); 374 if (error) { 375 brelse(bp); 376 return (error); 377 } 378 error = uiomove((char *)bp->b_data + on, n, uio); 379 brelse(bp); 380 } while (error == 0 && uio->uio_resid > 0 && n != 0); 381 return (error); 382 383 default: 384 panic("spec_read type"); 385 } 386 /* NOTREACHED */ 387 } 388 389 /* 390 * Vnode op for write 391 */ 392 /* ARGSUSED */ 393 int 394 spec_write(v) 395 void *v; 396 { 397 struct vop_write_args /* { 398 struct vnode *a_vp; 399 struct uio *a_uio; 400 int a_ioflag; 401 kauth_cred_t a_cred; 402 } */ *ap = v; 403 struct vnode *vp = ap->a_vp; 404 struct uio *uio = ap->a_uio; 405 struct lwp *l = curlwp; 406 struct buf *bp; 407 const struct bdevsw *bdev; 408 const struct cdevsw *cdev; 409 daddr_t bn; 410 int bsize, bscale; 411 struct partinfo dpart; 412 int n, on; 413 int error = 0; 414 415 #ifdef DIAGNOSTIC 416 if (uio->uio_rw != UIO_WRITE) 417 panic("spec_write mode"); 418 if (&uio->uio_vmspace->vm_map != kernel_map && 419 uio->uio_vmspace != curproc->p_vmspace) 420 panic("spec_write proc"); 421 #endif 422 423 switch (vp->v_type) { 424 425 case VCHR: 426 VOP_UNLOCK(vp, 0); 427 cdev = cdevsw_lookup(vp->v_rdev); 428 if (cdev != NULL) 429 error = (*cdev->d_write)(vp->v_rdev, uio, ap->a_ioflag); 430 else 431 error = ENXIO; 432 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 433 return (error); 434 435 case VBLK: 436 if (uio->uio_resid == 0) 437 return (0); 438 if (uio->uio_offset < 0) 439 return (EINVAL); 440 bsize = BLKDEV_IOSIZE; 441 bdev = bdevsw_lookup(vp->v_rdev); 442 if (bdev != NULL && 443 (*bdev->d_ioctl)(vp->v_rdev, DIOCGPART, (caddr_t)&dpart, 444 FREAD, l) == 0) { 445 if (dpart.part->p_fstype == FS_BSDFFS && 446 dpart.part->p_frag != 0 && dpart.part->p_fsize != 0) 447 bsize = dpart.part->p_frag * 448 dpart.part->p_fsize; 449 } 450 bscale = bsize >> DEV_BSHIFT; 451 do { 452 bn = (uio->uio_offset >> DEV_BSHIFT) &~ (bscale - 1); 453 on = uio->uio_offset % bsize; 454 n = min((unsigned)(bsize - on), uio->uio_resid); 455 if (n == bsize) 456 bp = getblk(vp, bn, bsize, 0, 0); 457 else 458 error = bread(vp, bn, bsize, NOCRED, &bp); 459 if (error) { 460 brelse(bp); 461 return (error); 462 } 463 n = min(n, bsize - bp->b_resid); 464 error = uiomove((char *)bp->b_data + on, n, uio); 465 if (error) 466 brelse(bp); 467 else { 468 if (n + on == bsize) 469 bawrite(bp); 470 else 471 bdwrite(bp); 472 if (bp->b_flags & B_ERROR) 473 error = bp->b_error; 474 } 475 } while (error == 0 && uio->uio_resid > 0 && n != 0); 476 return (error); 477 478 default: 479 panic("spec_write type"); 480 } 481 /* NOTREACHED */ 482 } 483 484 /* 485 * Device ioctl operation. 486 */ 487 /* ARGSUSED */ 488 int 489 spec_ioctl(v) 490 void *v; 491 { 492 struct vop_ioctl_args /* { 493 struct vnode *a_vp; 494 u_long a_command; 495 void *a_data; 496 int a_fflag; 497 kauth_cred_t a_cred; 498 struct lwp *a_l; 499 } */ *ap = v; 500 const struct bdevsw *bdev; 501 const struct cdevsw *cdev; 502 struct vnode *vp; 503 dev_t dev; 504 505 /* 506 * Extract all the info we need from the vnode, taking care to 507 * avoid a race with VOP_REVOKE(). 508 */ 509 510 vp = ap->a_vp; 511 dev = NODEV; 512 simple_lock(&vp->v_interlock); 513 if ((vp->v_flag & VXLOCK) == 0 && vp->v_specinfo) { 514 dev = vp->v_rdev; 515 } 516 simple_unlock(&vp->v_interlock); 517 if (dev == NODEV) { 518 return ENXIO; 519 } 520 521 switch (vp->v_type) { 522 523 case VCHR: 524 cdev = cdevsw_lookup(dev); 525 if (cdev == NULL) 526 return (ENXIO); 527 return ((*cdev->d_ioctl)(dev, ap->a_command, ap->a_data, 528 ap->a_fflag, ap->a_l)); 529 530 case VBLK: 531 bdev = bdevsw_lookup(dev); 532 if (bdev == NULL) 533 return (ENXIO); 534 if (ap->a_command == 0 && (long)ap->a_data == B_TAPE) { 535 if (bdev->d_type == D_TAPE) 536 return (0); 537 else 538 return (1); 539 } 540 return ((*bdev->d_ioctl)(dev, ap->a_command, ap->a_data, 541 ap->a_fflag, ap->a_l)); 542 543 default: 544 panic("spec_ioctl"); 545 /* NOTREACHED */ 546 } 547 } 548 549 /* ARGSUSED */ 550 int 551 spec_poll(v) 552 void *v; 553 { 554 struct vop_poll_args /* { 555 struct vnode *a_vp; 556 int a_events; 557 struct lwp *a_l; 558 } */ *ap = v; 559 const struct cdevsw *cdev; 560 dev_t dev; 561 562 switch (ap->a_vp->v_type) { 563 564 case VCHR: 565 dev = ap->a_vp->v_rdev; 566 cdev = cdevsw_lookup(dev); 567 if (cdev == NULL) 568 return (POLLERR); 569 return (*cdev->d_poll)(dev, ap->a_events, ap->a_l); 570 571 default: 572 return (genfs_poll(v)); 573 } 574 } 575 576 /* ARGSUSED */ 577 int 578 spec_kqfilter(v) 579 void *v; 580 { 581 struct vop_kqfilter_args /* { 582 struct vnode *a_vp; 583 struct proc *a_kn; 584 } */ *ap = v; 585 const struct cdevsw *cdev; 586 dev_t dev; 587 588 switch (ap->a_vp->v_type) { 589 590 case VCHR: 591 dev = ap->a_vp->v_rdev; 592 cdev = cdevsw_lookup(dev); 593 if (cdev == NULL) 594 return (ENXIO); 595 return (*cdev->d_kqfilter)(dev, ap->a_kn); 596 default: 597 /* 598 * Block devices don't support kqfilter, and refuse it 599 * for any other files (like those vflush()ed) too. 600 */ 601 return (EOPNOTSUPP); 602 } 603 } 604 605 /* 606 * Synch buffers associated with a block device 607 */ 608 /* ARGSUSED */ 609 int 610 spec_fsync(v) 611 void *v; 612 { 613 struct vop_fsync_args /* { 614 struct vnode *a_vp; 615 kauth_cred_t a_cred; 616 int a_flags; 617 off_t offlo; 618 off_t offhi; 619 struct lwp *a_l; 620 } */ *ap = v; 621 struct vnode *vp = ap->a_vp; 622 623 if (vp->v_type == VBLK) 624 vflushbuf(vp, (ap->a_flags & FSYNC_WAIT) != 0); 625 return (0); 626 } 627 628 /* 629 * Just call the device strategy routine 630 */ 631 int 632 spec_strategy(v) 633 void *v; 634 { 635 struct vop_strategy_args /* { 636 struct vnode *a_vp; 637 struct buf *a_bp; 638 } */ *ap = v; 639 struct vnode *vp = ap->a_vp; 640 struct buf *bp = ap->a_bp; 641 int error, s; 642 struct spec_cow_entry *e; 643 644 error = 0; 645 bp->b_dev = vp->v_rdev; 646 if (!(bp->b_flags & B_READ) && 647 (LIST_FIRST(&bp->b_dep)) != NULL && bioops.io_start) 648 (*bioops.io_start)(bp); 649 650 if (!(bp->b_flags & B_READ) && !SLIST_EMPTY(&vp->v_spec_cow_head)) { 651 SPEC_COW_LOCK(vp->v_specinfo, s); 652 while (vp->v_spec_cow_req > 0) 653 ltsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 654 &vp->v_spec_cow_slock); 655 vp->v_spec_cow_count++; 656 SPEC_COW_UNLOCK(vp->v_specinfo, s); 657 658 SLIST_FOREACH(e, &vp->v_spec_cow_head, ce_list) { 659 if ((error = (*e->ce_func)(e->ce_cookie, bp)) != 0) 660 break; 661 } 662 663 SPEC_COW_LOCK(vp->v_specinfo, s); 664 vp->v_spec_cow_count--; 665 if (vp->v_spec_cow_req && vp->v_spec_cow_count == 0) 666 wakeup(&vp->v_spec_cow_req); 667 SPEC_COW_UNLOCK(vp->v_specinfo, s); 668 } 669 670 if (error) { 671 bp->b_error = error; 672 bp->b_flags |= B_ERROR; 673 biodone(bp); 674 return (error); 675 } 676 677 DEV_STRATEGY(bp); 678 679 return (0); 680 } 681 682 int 683 spec_inactive(v) 684 void *v; 685 { 686 struct vop_inactive_args /* { 687 struct vnode *a_vp; 688 struct proc *a_l; 689 } */ *ap = v; 690 691 VOP_UNLOCK(ap->a_vp, 0); 692 return (0); 693 } 694 695 /* 696 * This is a noop, simply returning what one has been given. 697 */ 698 int 699 spec_bmap(v) 700 void *v; 701 { 702 struct vop_bmap_args /* { 703 struct vnode *a_vp; 704 daddr_t a_bn; 705 struct vnode **a_vpp; 706 daddr_t *a_bnp; 707 int *a_runp; 708 } */ *ap = v; 709 710 if (ap->a_vpp != NULL) 711 *ap->a_vpp = ap->a_vp; 712 if (ap->a_bnp != NULL) 713 *ap->a_bnp = ap->a_bn; 714 if (ap->a_runp != NULL) 715 *ap->a_runp = (MAXBSIZE >> DEV_BSHIFT) - 1; 716 return (0); 717 } 718 719 /* 720 * Device close routine 721 */ 722 /* ARGSUSED */ 723 int 724 spec_close(v) 725 void *v; 726 { 727 struct vop_close_args /* { 728 struct vnode *a_vp; 729 int a_fflag; 730 kauth_cred_t a_cred; 731 struct lwp *a_l; 732 } */ *ap = v; 733 struct vnode *vp = ap->a_vp; 734 const struct bdevsw *bdev; 735 const struct cdevsw *cdev; 736 struct session *sess; 737 dev_t dev = vp->v_rdev; 738 int (*devclose)(dev_t, int, int, struct lwp *); 739 int mode, error, count, flags, flags1; 740 741 count = vcount(vp); 742 flags = vp->v_flag; 743 744 switch (vp->v_type) { 745 746 case VCHR: 747 /* 748 * Hack: a tty device that is a controlling terminal 749 * has a reference from the session structure. 750 * We cannot easily tell that a character device is 751 * a controlling terminal, unless it is the closing 752 * process' controlling terminal. In that case, 753 * if the reference count is 2 (this last descriptor 754 * plus the session), release the reference from the session. 755 * Also remove the link from the tty back to the session 756 * and pgrp - due to the way consoles are handled we cannot 757 * guarantee that the vrele() will do the final close on the 758 * actual tty device. 759 */ 760 if (count == 2 && ap->a_l && 761 vp == (sess = ap->a_l->l_proc->p_session)->s_ttyvp) { 762 sess->s_ttyvp = NULL; 763 if (sess->s_ttyp->t_session != NULL) { 764 sess->s_ttyp->t_pgrp = NULL; 765 sess->s_ttyp->t_session = NULL; 766 SESSRELE(sess); 767 } else if (sess->s_ttyp->t_pgrp != NULL) 768 panic("spec_close: spurious pgrp ref"); 769 vrele(vp); 770 count--; 771 } 772 /* 773 * If the vnode is locked, then we are in the midst 774 * of forcably closing the device, otherwise we only 775 * close on last reference. 776 */ 777 if (count > 1 && (flags & VXLOCK) == 0) 778 return (0); 779 cdev = cdevsw_lookup(dev); 780 if (cdev != NULL) 781 devclose = cdev->d_close; 782 else 783 devclose = NULL; 784 mode = S_IFCHR; 785 break; 786 787 case VBLK: 788 /* 789 * On last close of a block device (that isn't mounted) 790 * we must invalidate any in core blocks, so that 791 * we can, for instance, change floppy disks. 792 */ 793 error = vinvalbuf(vp, V_SAVE, ap->a_cred, ap->a_l, 0, 0); 794 if (error) 795 return (error); 796 /* 797 * We do not want to really close the device if it 798 * is still in use unless we are trying to close it 799 * forcibly. Since every use (buffer, vnode, swap, cmap) 800 * holds a reference to the vnode, and because we mark 801 * any other vnodes that alias this device, when the 802 * sum of the reference counts on all the aliased 803 * vnodes descends to one, we are on last close. 804 */ 805 if (count > 1 && (flags & VXLOCK) == 0) 806 return (0); 807 bdev = bdevsw_lookup(dev); 808 if (bdev != NULL) 809 devclose = bdev->d_close; 810 else 811 devclose = NULL; 812 mode = S_IFBLK; 813 break; 814 815 default: 816 panic("spec_close: not special"); 817 } 818 819 flags1 = ap->a_fflag; 820 821 /* 822 * if VXLOCK is set, then we're going away soon, so make this 823 * non-blocking. Also ensures that we won't wedge in vn_lock below. 824 */ 825 if (flags & VXLOCK) 826 flags1 |= FNONBLOCK; 827 828 /* 829 * If we're able to block, release the vnode lock & reacquire. We 830 * might end up sleeping for someone else who wants our queues. They 831 * won't get them if we hold the vnode locked. Also, if VXLOCK is set, 832 * don't release the lock as we won't be able to regain it. 833 */ 834 if (!(flags1 & FNONBLOCK)) 835 VOP_UNLOCK(vp, 0); 836 837 if (devclose != NULL) 838 error = (*devclose)(dev, flags1, mode, ap->a_l); 839 else 840 error = ENXIO; 841 842 if (!(flags1 & FNONBLOCK)) 843 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 844 845 return (error); 846 } 847 848 /* 849 * Print out the contents of a special device vnode. 850 */ 851 int 852 spec_print(v) 853 void *v; 854 { 855 struct vop_print_args /* { 856 struct vnode *a_vp; 857 } */ *ap = v; 858 859 printf("tag VT_NON, dev %d, %d\n", major(ap->a_vp->v_rdev), 860 minor(ap->a_vp->v_rdev)); 861 return 0; 862 } 863 864 /* 865 * Return POSIX pathconf information applicable to special devices. 866 */ 867 int 868 spec_pathconf(v) 869 void *v; 870 { 871 struct vop_pathconf_args /* { 872 struct vnode *a_vp; 873 int a_name; 874 register_t *a_retval; 875 } */ *ap = v; 876 877 switch (ap->a_name) { 878 case _PC_LINK_MAX: 879 *ap->a_retval = LINK_MAX; 880 return (0); 881 case _PC_MAX_CANON: 882 *ap->a_retval = MAX_CANON; 883 return (0); 884 case _PC_MAX_INPUT: 885 *ap->a_retval = MAX_INPUT; 886 return (0); 887 case _PC_PIPE_BUF: 888 *ap->a_retval = PIPE_BUF; 889 return (0); 890 case _PC_CHOWN_RESTRICTED: 891 *ap->a_retval = 1; 892 return (0); 893 case _PC_VDISABLE: 894 *ap->a_retval = _POSIX_VDISABLE; 895 return (0); 896 case _PC_SYNC_IO: 897 *ap->a_retval = 1; 898 return (0); 899 default: 900 return (EINVAL); 901 } 902 /* NOTREACHED */ 903 } 904 905 /* 906 * Advisory record locking support. 907 */ 908 int 909 spec_advlock(v) 910 void *v; 911 { 912 struct vop_advlock_args /* { 913 struct vnode *a_vp; 914 void *a_id; 915 int a_op; 916 struct flock *a_fl; 917 int a_flags; 918 } */ *ap = v; 919 struct vnode *vp = ap->a_vp; 920 921 return lf_advlock(ap, &vp->v_speclockf, (off_t)0); 922 } 923
Indexes created Tue Sep 23 14:10:03 GMT 2025