spec_vnops.c revision 1.94
1 /* $NetBSD: spec_vnops.c,v 1.94 2006/11/01 09:37:28 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 1.94 2006/11/01 09:37:28 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 *vp = ap->a_vp; 183 const struct bdevsw *bdev; 184 const struct cdevsw *cdev; 185 dev_t 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; 212 213 /* XXX we're holding a vnode lock here */ 214 if (iskmemdev(dev)) { 215 error = kauth_authorize_system(ap->a_cred, 216 KAUTH_SYSTEM_RAWIO, 217 KAUTH_REQ_SYSTEM_RAWIO_MEMORY, 218 (void *)rw, NULL, NULL); 219 } else { 220 error = kauth_authorize_system(ap->a_cred, 221 KAUTH_SYSTEM_RAWIO, 222 KAUTH_REQ_SYSTEM_RAWIO_DISK, 223 (void *)rw, vp, (void *)(u_long)dev); 224 } 225 226 if (error) 227 return (error); 228 229 #if NVERIEXEC > 0 230 if (veriexec_strict >= VERIEXEC_IPS && 231 iskmemdev(dev) && (ap->a_mode & FWRITE)) { 232 return (error); 233 } else { 234 struct vnode *bvp; 235 dev_t blkdev; 236 237 blkdev = devsw_chr2blk(dev); 238 if (blkdev != NODEV) { 239 bvp = NULL; 240 vfinddev(blkdev, VBLK, &bvp); 241 error = veriexec_rawchk(bvp); 242 if (error) 243 return (error); 244 } 245 } 246 #endif /* NVERIEXEC > 0 */ 247 } 248 249 if (cdev->d_type == D_TTY) 250 vp->v_flag |= VISTTY; 251 VOP_UNLOCK(vp, 0); 252 error = (*cdev->d_open)(dev, ap->a_mode, S_IFCHR, l); 253 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 254 if (cdev->d_type != D_DISK) 255 return error; 256 d_ioctl = cdev->d_ioctl; 257 break; 258 259 case VBLK: 260 bdev = bdevsw_lookup(dev); 261 if (bdev == NULL) 262 return (ENXIO); 263 264 /* 265 * When running in very secure mode, do not allow 266 * opens for writing of any disk block devices. 267 */ 268 if (ap->a_cred != FSCRED) { 269 u_long rw; 270 271 if ((error = vfs_mountedon(vp)) != 0) 272 return (error); 273 274 rw = M2K(ap->a_mode); 275 276 error = kauth_authorize_system(ap->a_cred, 277 KAUTH_SYSTEM_RAWIO, 278 KAUTH_REQ_SYSTEM_RAWIO_DISK, 279 (void *)rw, vp, (void *)(u_long)dev); 280 if (error) 281 return (error); 282 } 283 284 #if NVERIEXEC > 0 285 error = veriexec_rawchk(vp); 286 if (error) 287 return (error); 288 #endif /* NVERIEXEC > 0 */ 289 290 error = (*bdev->d_open)(dev, ap->a_mode, S_IFBLK, l); 291 d_ioctl = bdev->d_ioctl; 292 break; 293 294 case VNON: 295 case VLNK: 296 case VDIR: 297 case VREG: 298 case VBAD: 299 case VFIFO: 300 case VSOCK: 301 default: 302 return 0; 303 } 304 305 #undef M2K 306 307 if (error) 308 return error; 309 if (!(*d_ioctl)(vp->v_rdev, DIOCGPART, (caddr_t)&pi, FREAD, curlwp)) 310 vp->v_size = (voff_t)pi.disklab->d_secsize * pi.part->p_size; 311 return 0; 312 } 313 314 /* 315 * Vnode op for read 316 */ 317 /* ARGSUSED */ 318 int 319 spec_read(v) 320 void *v; 321 { 322 struct vop_read_args /* { 323 struct vnode *a_vp; 324 struct uio *a_uio; 325 int a_ioflag; 326 kauth_cred_t a_cred; 327 } */ *ap = v; 328 struct vnode *vp = ap->a_vp; 329 struct uio *uio = ap->a_uio; 330 struct lwp *l = curlwp; 331 struct buf *bp; 332 const struct bdevsw *bdev; 333 const struct cdevsw *cdev; 334 daddr_t bn; 335 int bsize, bscale; 336 struct partinfo dpart; 337 int n, on; 338 int error = 0; 339 340 #ifdef DIAGNOSTIC 341 if (uio->uio_rw != UIO_READ) 342 panic("spec_read mode"); 343 if (&uio->uio_vmspace->vm_map != kernel_map && 344 uio->uio_vmspace != curproc->p_vmspace) 345 panic("spec_read proc"); 346 #endif 347 if (uio->uio_resid == 0) 348 return (0); 349 350 switch (vp->v_type) { 351 352 case VCHR: 353 VOP_UNLOCK(vp, 0); 354 cdev = cdevsw_lookup(vp->v_rdev); 355 if (cdev != NULL) 356 error = (*cdev->d_read)(vp->v_rdev, uio, ap->a_ioflag); 357 else 358 error = ENXIO; 359 vn_lock(vp, LK_SHARED | LK_RETRY); 360 return (error); 361 362 case VBLK: 363 if (uio->uio_offset < 0) 364 return (EINVAL); 365 bsize = BLKDEV_IOSIZE; 366 bdev = bdevsw_lookup(vp->v_rdev); 367 if (bdev != NULL && 368 (*bdev->d_ioctl)(vp->v_rdev, DIOCGPART, (caddr_t)&dpart, 369 FREAD, l) == 0) { 370 if (dpart.part->p_fstype == FS_BSDFFS && 371 dpart.part->p_frag != 0 && dpart.part->p_fsize != 0) 372 bsize = dpart.part->p_frag * 373 dpart.part->p_fsize; 374 } 375 bscale = bsize >> DEV_BSHIFT; 376 do { 377 bn = (uio->uio_offset >> DEV_BSHIFT) &~ (bscale - 1); 378 on = uio->uio_offset % bsize; 379 n = min((unsigned)(bsize - on), uio->uio_resid); 380 error = bread(vp, bn, bsize, NOCRED, &bp); 381 n = min(n, bsize - bp->b_resid); 382 if (error) { 383 brelse(bp); 384 return (error); 385 } 386 error = uiomove((char *)bp->b_data + on, n, uio); 387 brelse(bp); 388 } while (error == 0 && uio->uio_resid > 0 && n != 0); 389 return (error); 390 391 default: 392 panic("spec_read type"); 393 } 394 /* NOTREACHED */ 395 } 396 397 /* 398 * Vnode op for write 399 */ 400 /* ARGSUSED */ 401 int 402 spec_write(v) 403 void *v; 404 { 405 struct vop_write_args /* { 406 struct vnode *a_vp; 407 struct uio *a_uio; 408 int a_ioflag; 409 kauth_cred_t a_cred; 410 } */ *ap = v; 411 struct vnode *vp = ap->a_vp; 412 struct uio *uio = ap->a_uio; 413 struct lwp *l = curlwp; 414 struct buf *bp; 415 const struct bdevsw *bdev; 416 const struct cdevsw *cdev; 417 daddr_t bn; 418 int bsize, bscale; 419 struct partinfo dpart; 420 int n, on; 421 int error = 0; 422 423 #ifdef DIAGNOSTIC 424 if (uio->uio_rw != UIO_WRITE) 425 panic("spec_write mode"); 426 if (&uio->uio_vmspace->vm_map != kernel_map && 427 uio->uio_vmspace != curproc->p_vmspace) 428 panic("spec_write proc"); 429 #endif 430 431 switch (vp->v_type) { 432 433 case VCHR: 434 VOP_UNLOCK(vp, 0); 435 cdev = cdevsw_lookup(vp->v_rdev); 436 if (cdev != NULL) 437 error = (*cdev->d_write)(vp->v_rdev, uio, ap->a_ioflag); 438 else 439 error = ENXIO; 440 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 441 return (error); 442 443 case VBLK: 444 if (uio->uio_resid == 0) 445 return (0); 446 if (uio->uio_offset < 0) 447 return (EINVAL); 448 bsize = BLKDEV_IOSIZE; 449 bdev = bdevsw_lookup(vp->v_rdev); 450 if (bdev != NULL && 451 (*bdev->d_ioctl)(vp->v_rdev, DIOCGPART, (caddr_t)&dpart, 452 FREAD, l) == 0) { 453 if (dpart.part->p_fstype == FS_BSDFFS && 454 dpart.part->p_frag != 0 && dpart.part->p_fsize != 0) 455 bsize = dpart.part->p_frag * 456 dpart.part->p_fsize; 457 } 458 bscale = bsize >> DEV_BSHIFT; 459 do { 460 bn = (uio->uio_offset >> DEV_BSHIFT) &~ (bscale - 1); 461 on = uio->uio_offset % bsize; 462 n = min((unsigned)(bsize - on), uio->uio_resid); 463 if (n == bsize) 464 bp = getblk(vp, bn, bsize, 0, 0); 465 else 466 error = bread(vp, bn, bsize, NOCRED, &bp); 467 if (error) { 468 brelse(bp); 469 return (error); 470 } 471 n = min(n, bsize - bp->b_resid); 472 error = uiomove((char *)bp->b_data + on, n, uio); 473 if (error) 474 brelse(bp); 475 else { 476 if (n + on == bsize) 477 bawrite(bp); 478 else 479 bdwrite(bp); 480 if (bp->b_flags & B_ERROR) 481 error = bp->b_error; 482 } 483 } while (error == 0 && uio->uio_resid > 0 && n != 0); 484 return (error); 485 486 default: 487 panic("spec_write type"); 488 } 489 /* NOTREACHED */ 490 } 491 492 /* 493 * Device ioctl operation. 494 */ 495 /* ARGSUSED */ 496 int 497 spec_ioctl(v) 498 void *v; 499 { 500 struct vop_ioctl_args /* { 501 struct vnode *a_vp; 502 u_long a_command; 503 void *a_data; 504 int a_fflag; 505 kauth_cred_t a_cred; 506 struct lwp *a_l; 507 } */ *ap = v; 508 const struct bdevsw *bdev; 509 const struct cdevsw *cdev; 510 struct vnode *vp; 511 dev_t dev; 512 513 /* 514 * Extract all the info we need from the vnode, taking care to 515 * avoid a race with VOP_REVOKE(). 516 */ 517 518 vp = ap->a_vp; 519 dev = NODEV; 520 simple_lock(&vp->v_interlock); 521 if ((vp->v_flag & VXLOCK) == 0 && vp->v_specinfo) { 522 dev = vp->v_rdev; 523 } 524 simple_unlock(&vp->v_interlock); 525 if (dev == NODEV) { 526 return ENXIO; 527 } 528 529 switch (vp->v_type) { 530 531 case VCHR: 532 cdev = cdevsw_lookup(dev); 533 if (cdev == NULL) 534 return (ENXIO); 535 return ((*cdev->d_ioctl)(dev, ap->a_command, ap->a_data, 536 ap->a_fflag, ap->a_l)); 537 538 case VBLK: 539 bdev = bdevsw_lookup(dev); 540 if (bdev == NULL) 541 return (ENXIO); 542 if (ap->a_command == 0 && (long)ap->a_data == B_TAPE) { 543 if (bdev->d_type == D_TAPE) 544 return (0); 545 else 546 return (1); 547 } 548 return ((*bdev->d_ioctl)(dev, ap->a_command, ap->a_data, 549 ap->a_fflag, ap->a_l)); 550 551 default: 552 panic("spec_ioctl"); 553 /* NOTREACHED */ 554 } 555 } 556 557 /* ARGSUSED */ 558 int 559 spec_poll(v) 560 void *v; 561 { 562 struct vop_poll_args /* { 563 struct vnode *a_vp; 564 int a_events; 565 struct lwp *a_l; 566 } */ *ap = v; 567 const struct cdevsw *cdev; 568 struct vnode *vp; 569 dev_t dev; 570 571 /* 572 * Extract all the info we need from the vnode, taking care to 573 * avoid a race with VOP_REVOKE(). 574 */ 575 576 vp = ap->a_vp; 577 dev = NODEV; 578 simple_lock(&vp->v_interlock); 579 if ((vp->v_flag & VXLOCK) == 0 && vp->v_specinfo) { 580 dev = vp->v_rdev; 581 } 582 simple_unlock(&vp->v_interlock); 583 if (dev == NODEV) { 584 return POLLERR; 585 } 586 587 switch (vp->v_type) { 588 589 case VCHR: 590 cdev = cdevsw_lookup(dev); 591 if (cdev == NULL) 592 return (POLLERR); 593 return (*cdev->d_poll)(dev, ap->a_events, ap->a_l); 594 595 default: 596 return (genfs_poll(v)); 597 } 598 } 599 600 /* ARGSUSED */ 601 int 602 spec_kqfilter(v) 603 void *v; 604 { 605 struct vop_kqfilter_args /* { 606 struct vnode *a_vp; 607 struct proc *a_kn; 608 } */ *ap = v; 609 const struct cdevsw *cdev; 610 dev_t dev; 611 612 switch (ap->a_vp->v_type) { 613 614 case VCHR: 615 dev = ap->a_vp->v_rdev; 616 cdev = cdevsw_lookup(dev); 617 if (cdev == NULL) 618 return (ENXIO); 619 return (*cdev->d_kqfilter)(dev, ap->a_kn); 620 default: 621 /* 622 * Block devices don't support kqfilter, and refuse it 623 * for any other files (like those vflush()ed) too. 624 */ 625 return (EOPNOTSUPP); 626 } 627 } 628 629 /* 630 * Synch buffers associated with a block device 631 */ 632 /* ARGSUSED */ 633 int 634 spec_fsync(v) 635 void *v; 636 { 637 struct vop_fsync_args /* { 638 struct vnode *a_vp; 639 kauth_cred_t a_cred; 640 int a_flags; 641 off_t offlo; 642 off_t offhi; 643 struct lwp *a_l; 644 } */ *ap = v; 645 struct vnode *vp = ap->a_vp; 646 647 if (vp->v_type == VBLK) 648 vflushbuf(vp, (ap->a_flags & FSYNC_WAIT) != 0); 649 return (0); 650 } 651 652 /* 653 * Just call the device strategy routine 654 */ 655 int 656 spec_strategy(v) 657 void *v; 658 { 659 struct vop_strategy_args /* { 660 struct vnode *a_vp; 661 struct buf *a_bp; 662 } */ *ap = v; 663 struct vnode *vp = ap->a_vp; 664 struct buf *bp = ap->a_bp; 665 int error, s; 666 struct spec_cow_entry *e; 667 668 error = 0; 669 bp->b_dev = vp->v_rdev; 670 if (!(bp->b_flags & B_READ) && 671 (LIST_FIRST(&bp->b_dep)) != NULL && bioops.io_start) 672 (*bioops.io_start)(bp); 673 674 if (!(bp->b_flags & B_READ) && !SLIST_EMPTY(&vp->v_spec_cow_head)) { 675 SPEC_COW_LOCK(vp->v_specinfo, s); 676 while (vp->v_spec_cow_req > 0) 677 ltsleep(&vp->v_spec_cow_req, PRIBIO, "cowlist", 0, 678 &vp->v_spec_cow_slock); 679 vp->v_spec_cow_count++; 680 SPEC_COW_UNLOCK(vp->v_specinfo, s); 681 682 SLIST_FOREACH(e, &vp->v_spec_cow_head, ce_list) { 683 if ((error = (*e->ce_func)(e->ce_cookie, bp)) != 0) 684 break; 685 } 686 687 SPEC_COW_LOCK(vp->v_specinfo, s); 688 vp->v_spec_cow_count--; 689 if (vp->v_spec_cow_req && vp->v_spec_cow_count == 0) 690 wakeup(&vp->v_spec_cow_req); 691 SPEC_COW_UNLOCK(vp->v_specinfo, s); 692 } 693 694 if (error) { 695 bp->b_error = error; 696 bp->b_flags |= B_ERROR; 697 biodone(bp); 698 return (error); 699 } 700 701 DEV_STRATEGY(bp); 702 703 return (0); 704 } 705 706 int 707 spec_inactive(v) 708 void *v; 709 { 710 struct vop_inactive_args /* { 711 struct vnode *a_vp; 712 struct proc *a_l; 713 } */ *ap = v; 714 715 VOP_UNLOCK(ap->a_vp, 0); 716 return (0); 717 } 718 719 /* 720 * This is a noop, simply returning what one has been given. 721 */ 722 int 723 spec_bmap(v) 724 void *v; 725 { 726 struct vop_bmap_args /* { 727 struct vnode *a_vp; 728 daddr_t a_bn; 729 struct vnode **a_vpp; 730 daddr_t *a_bnp; 731 int *a_runp; 732 } */ *ap = v; 733 734 if (ap->a_vpp != NULL) 735 *ap->a_vpp = ap->a_vp; 736 if (ap->a_bnp != NULL) 737 *ap->a_bnp = ap->a_bn; 738 if (ap->a_runp != NULL) 739 *ap->a_runp = (MAXBSIZE >> DEV_BSHIFT) - 1; 740 return (0); 741 } 742 743 /* 744 * Device close routine 745 */ 746 /* ARGSUSED */ 747 int 748 spec_close(v) 749 void *v; 750 { 751 struct vop_close_args /* { 752 struct vnode *a_vp; 753 int a_fflag; 754 kauth_cred_t a_cred; 755 struct lwp *a_l; 756 } */ *ap = v; 757 struct vnode *vp = ap->a_vp; 758 const struct bdevsw *bdev; 759 const struct cdevsw *cdev; 760 struct session *sess; 761 dev_t dev = vp->v_rdev; 762 int (*devclose)(dev_t, int, int, struct lwp *); 763 int mode, error, count, flags, flags1; 764 765 count = vcount(vp); 766 flags = vp->v_flag; 767 768 switch (vp->v_type) { 769 770 case VCHR: 771 /* 772 * Hack: a tty device that is a controlling terminal 773 * has a reference from the session structure. 774 * We cannot easily tell that a character device is 775 * a controlling terminal, unless it is the closing 776 * process' controlling terminal. In that case, 777 * if the reference count is 2 (this last descriptor 778 * plus the session), release the reference from the session. 779 * Also remove the link from the tty back to the session 780 * and pgrp - due to the way consoles are handled we cannot 781 * guarantee that the vrele() will do the final close on the 782 * actual tty device. 783 */ 784 if (count == 2 && ap->a_l && 785 vp == (sess = ap->a_l->l_proc->p_session)->s_ttyvp) { 786 sess->s_ttyvp = NULL; 787 if (sess->s_ttyp->t_session != NULL) { 788 sess->s_ttyp->t_pgrp = NULL; 789 sess->s_ttyp->t_session = NULL; 790 SESSRELE(sess); 791 } else if (sess->s_ttyp->t_pgrp != NULL) 792 panic("spec_close: spurious pgrp ref"); 793 vrele(vp); 794 count--; 795 } 796 /* 797 * If the vnode is locked, then we are in the midst 798 * of forcably closing the device, otherwise we only 799 * close on last reference. 800 */ 801 if (count > 1 && (flags & VXLOCK) == 0) 802 return (0); 803 cdev = cdevsw_lookup(dev); 804 if (cdev != NULL) 805 devclose = cdev->d_close; 806 else 807 devclose = NULL; 808 mode = S_IFCHR; 809 break; 810 811 case VBLK: 812 /* 813 * On last close of a block device (that isn't mounted) 814 * we must invalidate any in core blocks, so that 815 * we can, for instance, change floppy disks. 816 */ 817 error = vinvalbuf(vp, V_SAVE, ap->a_cred, ap->a_l, 0, 0); 818 if (error) 819 return (error); 820 /* 821 * We do not want to really close the device if it 822 * is still in use unless we are trying to close it 823 * forcibly. Since every use (buffer, vnode, swap, cmap) 824 * holds a reference to the vnode, and because we mark 825 * any other vnodes that alias this device, when the 826 * sum of the reference counts on all the aliased 827 * vnodes descends to one, we are on last close. 828 */ 829 if (count > 1 && (flags & VXLOCK) == 0) 830 return (0); 831 bdev = bdevsw_lookup(dev); 832 if (bdev != NULL) 833 devclose = bdev->d_close; 834 else 835 devclose = NULL; 836 mode = S_IFBLK; 837 break; 838 839 default: 840 panic("spec_close: not special"); 841 } 842 843 flags1 = ap->a_fflag; 844 845 /* 846 * if VXLOCK is set, then we're going away soon, so make this 847 * non-blocking. Also ensures that we won't wedge in vn_lock below. 848 */ 849 if (flags & VXLOCK) 850 flags1 |= FNONBLOCK; 851 852 /* 853 * If we're able to block, release the vnode lock & reacquire. We 854 * might end up sleeping for someone else who wants our queues. They 855 * won't get them if we hold the vnode locked. Also, if VXLOCK is set, 856 * don't release the lock as we won't be able to regain it. 857 */ 858 if (!(flags1 & FNONBLOCK)) 859 VOP_UNLOCK(vp, 0); 860 861 if (devclose != NULL) 862 error = (*devclose)(dev, flags1, mode, ap->a_l); 863 else 864 error = ENXIO; 865 866 if (!(flags1 & FNONBLOCK)) 867 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 868 869 return (error); 870 } 871 872 /* 873 * Print out the contents of a special device vnode. 874 */ 875 int 876 spec_print(v) 877 void *v; 878 { 879 struct vop_print_args /* { 880 struct vnode *a_vp; 881 } */ *ap = v; 882 883 printf("tag VT_NON, dev %d, %d\n", major(ap->a_vp->v_rdev), 884 minor(ap->a_vp->v_rdev)); 885 return 0; 886 } 887 888 /* 889 * Return POSIX pathconf information applicable to special devices. 890 */ 891 int 892 spec_pathconf(v) 893 void *v; 894 { 895 struct vop_pathconf_args /* { 896 struct vnode *a_vp; 897 int a_name; 898 register_t *a_retval; 899 } */ *ap = v; 900 901 switch (ap->a_name) { 902 case _PC_LINK_MAX: 903 *ap->a_retval = LINK_MAX; 904 return (0); 905 case _PC_MAX_CANON: 906 *ap->a_retval = MAX_CANON; 907 return (0); 908 case _PC_MAX_INPUT: 909 *ap->a_retval = MAX_INPUT; 910 return (0); 911 case _PC_PIPE_BUF: 912 *ap->a_retval = PIPE_BUF; 913 return (0); 914 case _PC_CHOWN_RESTRICTED: 915 *ap->a_retval = 1; 916 return (0); 917 case _PC_VDISABLE: 918 *ap->a_retval = _POSIX_VDISABLE; 919 return (0); 920 case _PC_SYNC_IO: 921 *ap->a_retval = 1; 922 return (0); 923 default: 924 return (EINVAL); 925 } 926 /* NOTREACHED */ 927 } 928 929 /* 930 * Advisory record locking support. 931 */ 932 int 933 spec_advlock(v) 934 void *v; 935 { 936 struct vop_advlock_args /* { 937 struct vnode *a_vp; 938 void *a_id; 939 int a_op; 940 struct flock *a_fl; 941 int a_flags; 942 } */ *ap = v; 943 struct vnode *vp = ap->a_vp; 944 945 return lf_advlock(ap, &vp->v_speclockf, (off_t)0); 946 } 947
Indexes created Tue Sep 23 07:09:52 GMT 2025