spec_vnops.c revision 1.112
1 /* $NetBSD: spec_vnops.c,v 1.112 2008/01/24 17:32:55 ad Exp $ */ 2 3 /*- 4 * Copyright (c) 2008 The NetBSD Foundation, Inc. 5 * 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. All advertising materials mentioning features or use of this software 16 * must display the following acknowledgement: 17 * This product includes software developed by the NetBSD 18 * Foundation, Inc. and its contributors. 19 * 4. Neither the name of The NetBSD Foundation nor the names of its 20 * contributors may be used to endorse or promote products derived 21 * from this software without specific prior written permission. 22 * 23 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 24 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 25 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 26 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 27 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 28 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 29 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 30 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 31 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 32 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 33 * POSSIBILITY OF SUCH DAMAGE. 34 */ 35 36 /* 37 * Copyright (c) 1989, 1993 38 * The Regents of the University of California. All rights reserved. 39 * 40 * Redistribution and use in source and binary forms, with or without 41 * modification, are permitted provided that the following conditions 42 * are met: 43 * 1. Redistributions of source code must retain the above copyright 44 * notice, this list of conditions and the following disclaimer. 45 * 2. Redistributions in binary form must reproduce the above copyright 46 * notice, this list of conditions and the following disclaimer in the 47 * documentation and/or other materials provided with the distribution. 48 * 3. Neither the name of the University nor the names of its contributors 49 * may be used to endorse or promote products derived from this software 50 * without specific prior written permission. 51 * 52 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 53 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 54 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 55 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 56 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 57 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 58 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 59 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 60 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 61 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 62 * SUCH DAMAGE. 63 * 64 * @(#)spec_vnops.c 8.15 (Berkeley) 7/14/95 65 */ 66 67 #include <sys/cdefs.h> 68 __KERNEL_RCSID(0, "$NetBSD: spec_vnops.c,v 1.112 2008/01/24 17:32:55 ad Exp $"); 69 70 #include <sys/param.h> 71 #include <sys/proc.h> 72 #include <sys/systm.h> 73 #include <sys/kernel.h> 74 #include <sys/conf.h> 75 #include <sys/buf.h> 76 #include <sys/mount.h> 77 #include <sys/namei.h> 78 #include <sys/vnode.h> 79 #include <sys/stat.h> 80 #include <sys/errno.h> 81 #include <sys/ioctl.h> 82 #include <sys/poll.h> 83 #include <sys/file.h> 84 #include <sys/disklabel.h> 85 #include <sys/lockf.h> 86 #include <sys/tty.h> 87 #include <sys/kauth.h> 88 #include <sys/fstrans.h> 89 90 #include <miscfs/genfs/genfs.h> 91 #include <miscfs/specfs/specdev.h> 92 93 /* symbolic sleep message strings for devices */ 94 const char devopn[] = "devopn"; 95 const char devio[] = "devio"; 96 const char devwait[] = "devwait"; 97 const char devin[] = "devin"; 98 const char devout[] = "devout"; 99 const char devioc[] = "devioc"; 100 const char devcls[] = "devcls"; 101 102 vnode_t *specfs_hash[SPECHSZ]; 103 kmutex_t specfs_lock; 104 105 /* 106 * This vnode operations vector is used for special device nodes 107 * created from whole cloth by the kernel. For the ops vector for 108 * vnodes built from special devices found in a filesystem, see (e.g) 109 * ffs_specop_entries[] in ffs_vnops.c or the equivalent for other 110 * filesystems. 111 */ 112 113 int (**spec_vnodeop_p)(void *); 114 const struct vnodeopv_entry_desc spec_vnodeop_entries[] = { 115 { &vop_default_desc, vn_default_error }, 116 { &vop_lookup_desc, spec_lookup }, /* lookup */ 117 { &vop_create_desc, spec_create }, /* create */ 118 { &vop_mknod_desc, spec_mknod }, /* mknod */ 119 { &vop_open_desc, spec_open }, /* open */ 120 { &vop_close_desc, spec_close }, /* close */ 121 { &vop_access_desc, spec_access }, /* access */ 122 { &vop_getattr_desc, spec_getattr }, /* getattr */ 123 { &vop_setattr_desc, spec_setattr }, /* setattr */ 124 { &vop_read_desc, spec_read }, /* read */ 125 { &vop_write_desc, spec_write }, /* write */ 126 { &vop_lease_desc, spec_lease_check }, /* lease */ 127 { &vop_fcntl_desc, spec_fcntl }, /* fcntl */ 128 { &vop_ioctl_desc, spec_ioctl }, /* ioctl */ 129 { &vop_poll_desc, spec_poll }, /* poll */ 130 { &vop_kqfilter_desc, spec_kqfilter }, /* kqfilter */ 131 { &vop_revoke_desc, spec_revoke }, /* revoke */ 132 { &vop_mmap_desc, spec_mmap }, /* mmap */ 133 { &vop_fsync_desc, spec_fsync }, /* fsync */ 134 { &vop_seek_desc, spec_seek }, /* seek */ 135 { &vop_remove_desc, spec_remove }, /* remove */ 136 { &vop_link_desc, spec_link }, /* link */ 137 { &vop_rename_desc, spec_rename }, /* rename */ 138 { &vop_mkdir_desc, spec_mkdir }, /* mkdir */ 139 { &vop_rmdir_desc, spec_rmdir }, /* rmdir */ 140 { &vop_symlink_desc, spec_symlink }, /* symlink */ 141 { &vop_readdir_desc, spec_readdir }, /* readdir */ 142 { &vop_readlink_desc, spec_readlink }, /* readlink */ 143 { &vop_abortop_desc, spec_abortop }, /* abortop */ 144 { &vop_inactive_desc, spec_inactive }, /* inactive */ 145 { &vop_reclaim_desc, spec_reclaim }, /* reclaim */ 146 { &vop_lock_desc, spec_lock }, /* lock */ 147 { &vop_unlock_desc, spec_unlock }, /* unlock */ 148 { &vop_bmap_desc, spec_bmap }, /* bmap */ 149 { &vop_strategy_desc, spec_strategy }, /* strategy */ 150 { &vop_print_desc, spec_print }, /* print */ 151 { &vop_islocked_desc, spec_islocked }, /* islocked */ 152 { &vop_pathconf_desc, spec_pathconf }, /* pathconf */ 153 { &vop_advlock_desc, spec_advlock }, /* advlock */ 154 { &vop_bwrite_desc, spec_bwrite }, /* bwrite */ 155 { &vop_getpages_desc, spec_getpages }, /* getpages */ 156 { &vop_putpages_desc, spec_putpages }, /* putpages */ 157 { NULL, NULL } 158 }; 159 const struct vnodeopv_desc spec_vnodeop_opv_desc = 160 { &spec_vnodeop_p, spec_vnodeop_entries }; 161 162 /* 163 * Returns true if dev is /dev/mem or /dev/kmem. 164 */ 165 int 166 iskmemdev(dev_t dev) 167 { 168 /* mem_no is emitted by config(8) to generated devsw.c */ 169 extern const int mem_no; 170 171 /* minor 14 is /dev/io on i386 with COMPAT_10 */ 172 return (major(dev) == mem_no && (minor(dev) < 2 || minor(dev) == 14)); 173 } 174 175 /* 176 * Initialize a vnode that represents a device. 177 */ 178 void 179 spec_node_init(vnode_t *vp, dev_t rdev) 180 { 181 specnode_t *sn; 182 specdev_t *sd; 183 vnode_t *vp2; 184 vnode_t **vpp; 185 186 KASSERT(vp->v_type == VBLK || vp->v_type == VCHR); 187 KASSERT(vp->v_specnode == NULL); 188 189 /* 190 * Search the hash table for this device. If known, add a 191 * reference to the device structure. If not known, create 192 * a new entry to represent the device. In all cases add 193 * the vnode to the hash table. 194 */ 195 sn = kmem_alloc(sizeof(*sn), KM_SLEEP); 196 if (sn == NULL) { 197 /* XXX */ 198 panic("spec_node_init: unable to allocate memory"); 199 } 200 sd = kmem_alloc(sizeof(*sd), KM_SLEEP); 201 if (sd == NULL) { 202 /* XXX */ 203 panic("spec_node_init: unable to allocate memory"); 204 } 205 mutex_enter(&specfs_lock); 206 vpp = &specfs_hash[SPECHASH(rdev)]; 207 for (vp2 = *vpp; vp2 != NULL; vp2 = vp2->v_specnext) { 208 KASSERT(vp2->v_specnode != NULL); 209 if (rdev == vp2->v_rdev && vp->v_type == vp2->v_type) { 210 break; 211 } 212 } 213 if (vp2 == NULL) { 214 /* No existing record, create a new one. */ 215 sd->sd_rdev = rdev; 216 sd->sd_mountpoint = NULL; 217 sd->sd_lockf = NULL; 218 sd->sd_refcnt = 1; 219 sd->sd_opencnt = 0; 220 sd->sd_bdevvp = NULL; 221 sn->sn_dev = sd; 222 sd = NULL; 223 } else { 224 /* Use the existing record. */ 225 sn->sn_dev = vp2->v_specnode->sn_dev; 226 sn->sn_dev->sd_refcnt++; 227 } 228 /* Insert vnode into the hash chain. */ 229 sn->sn_opencnt = 0; 230 sn->sn_rdev = rdev; 231 sn->sn_gone = false; 232 vp->v_specnode = sn; 233 vp->v_specnext = *vpp; 234 *vpp = vp; 235 mutex_exit(&specfs_lock); 236 237 /* Free the record we allocated if unused. */ 238 if (sd != NULL) { 239 kmem_free(sd, sizeof(*sd)); 240 } 241 } 242 243 /* 244 * A vnode representing a special device is going away. Close 245 * the device if the vnode holds it open. 246 */ 247 void 248 spec_node_revoke(vnode_t *vp) 249 { 250 specnode_t *sn; 251 specdev_t *sd; 252 253 sn = vp->v_specnode; 254 sd = sn->sn_dev; 255 256 KASSERT(vp->v_type == VBLK || vp->v_type == VCHR); 257 KASSERT(vp->v_specnode != NULL); 258 KASSERT((vp->v_iflag & VI_XLOCK) != 0); 259 KASSERT(sn->sn_gone == false); 260 261 mutex_enter(&specfs_lock); 262 KASSERT(sn->sn_opencnt <= sd->sd_opencnt); 263 if (sn->sn_opencnt != 0) { 264 sd->sd_opencnt -= (sn->sn_opencnt - 1); 265 sn->sn_opencnt = 1; 266 sn->sn_gone = true; 267 mutex_exit(&specfs_lock); 268 269 VOP_CLOSE(vp, FNONBLOCK, NOCRED); 270 271 mutex_enter(&specfs_lock); 272 KASSERT(sn->sn_opencnt == 0); 273 } 274 mutex_exit(&specfs_lock); 275 } 276 277 /* 278 * A vnode representing a special device is being recycled. 279 * Destroy the specfs component. 280 */ 281 void 282 spec_node_destroy(vnode_t *vp) 283 { 284 specnode_t *sn; 285 specdev_t *sd; 286 vnode_t **vpp, *vp2; 287 int refcnt; 288 289 sn = vp->v_specnode; 290 sd = sn->sn_dev; 291 292 KASSERT(vp->v_type == VBLK || vp->v_type == VCHR); 293 KASSERT(vp->v_specnode != NULL); 294 KASSERT(sn->sn_opencnt == 0); 295 296 mutex_enter(&specfs_lock); 297 /* Remove from the hash and destroy the node. */ 298 vpp = &specfs_hash[SPECHASH(vp->v_rdev)]; 299 for (vp2 = *vpp;; vp2 = vp2->v_specnext) { 300 if (vp2 == NULL) { 301 panic("spec_node_destroy: corrupt hash"); 302 } 303 if (vp2 == vp) { 304 KASSERT(vp == *vpp); 305 *vpp = vp->v_specnext; 306 break; 307 } 308 if (vp2->v_specnext == vp) { 309 vp2->v_specnext = vp->v_specnext; 310 break; 311 } 312 } 313 sn = vp->v_specnode; 314 vp->v_specnode = NULL; 315 refcnt = sd->sd_refcnt--; 316 KASSERT(refcnt > 0); 317 mutex_exit(&specfs_lock); 318 319 /* If the device is no longer in use, destroy our record. */ 320 if (refcnt == 1) { 321 KASSERT(sd->sd_opencnt == 0); 322 KASSERT(sd->sd_bdevvp == NULL); 323 kmem_free(sd, sizeof(*sd)); 324 } 325 kmem_free(sn, sizeof(*sn)); 326 } 327 328 /* 329 * Trivial lookup routine that always fails. 330 */ 331 int 332 spec_lookup(void *v) 333 { 334 struct vop_lookup_args /* { 335 struct vnode *a_dvp; 336 struct vnode **a_vpp; 337 struct componentname *a_cnp; 338 } */ *ap = v; 339 340 *ap->a_vpp = NULL; 341 return (ENOTDIR); 342 } 343 344 /* 345 * Open a special file. 346 */ 347 /* ARGSUSED */ 348 int 349 spec_open(void *v) 350 { 351 struct vop_open_args /* { 352 struct vnode *a_vp; 353 int a_mode; 354 kauth_cred_t a_cred; 355 } */ *ap = v; 356 struct lwp *l; 357 struct vnode *vp; 358 dev_t dev; 359 int error; 360 struct partinfo pi; 361 enum kauth_device_req req; 362 specnode_t *sn; 363 specdev_t *sd; 364 365 l = curlwp; 366 vp = ap->a_vp; 367 dev = vp->v_rdev; 368 sn = vp->v_specnode; 369 sd = sn->sn_dev; 370 371 /* 372 * Don't allow open if fs is mounted -nodev. 373 */ 374 if (vp->v_mount && (vp->v_mount->mnt_flag & MNT_NODEV)) 375 return (ENXIO); 376 377 switch (ap->a_mode & (FREAD | FWRITE)) { 378 case FREAD | FWRITE: 379 req = KAUTH_REQ_DEVICE_RAWIO_SPEC_RW; 380 break; 381 case FWRITE: 382 req = KAUTH_REQ_DEVICE_RAWIO_SPEC_WRITE; 383 break; 384 default: 385 req = KAUTH_REQ_DEVICE_RAWIO_SPEC_READ; 386 break; 387 } 388 389 switch (vp->v_type) { 390 case VCHR: 391 error = kauth_authorize_device_spec(ap->a_cred, req, vp); 392 if (error != 0) 393 return (error); 394 395 /* 396 * Character devices can accept opens from multiple 397 * vnodes. 398 */ 399 mutex_enter(&specfs_lock); 400 if (sn->sn_gone) { 401 mutex_exit(&specfs_lock); 402 return (EBADF); 403 } 404 sd->sd_opencnt++; 405 sn->sn_opencnt++; 406 mutex_exit(&specfs_lock); 407 if (cdev_type(dev) == D_TTY) 408 vp->v_vflag |= VV_ISTTY; 409 VOP_UNLOCK(vp, 0); 410 error = cdev_open(dev, ap->a_mode, S_IFCHR, l); 411 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 412 break; 413 414 case VBLK: 415 error = kauth_authorize_device_spec(ap->a_cred, req, vp); 416 if (error != 0) 417 return (error); 418 419 /* 420 * For block devices, permit only one open. The buffer 421 * cache cannot remain self-consistent with multiple 422 * vnodes holding a block device open. 423 */ 424 mutex_enter(&specfs_lock); 425 if (sn->sn_gone) { 426 mutex_exit(&specfs_lock); 427 return (EBADF); 428 } 429 if (sd->sd_opencnt != 0) { 430 mutex_exit(&specfs_lock); 431 return EBUSY; 432 } 433 sn->sn_opencnt = 1; 434 sd->sd_opencnt = 1; 435 sd->sd_bdevvp = vp; 436 mutex_exit(&specfs_lock); 437 438 error = bdev_open(dev, ap->a_mode, S_IFBLK, l); 439 break; 440 441 case VNON: 442 case VLNK: 443 case VDIR: 444 case VREG: 445 case VBAD: 446 case VFIFO: 447 case VSOCK: 448 default: 449 return 0; 450 } 451 452 mutex_enter(&specfs_lock); 453 if (sn->sn_gone) { 454 if (error == 0) 455 error = EBADF; 456 } else if (error != 0) { 457 sd->sd_opencnt--; 458 sn->sn_opencnt--; 459 } 460 mutex_exit(&specfs_lock); 461 462 if (cdev_type(dev) != D_DISK || error != 0) 463 return error; 464 465 if (vp->v_type == VCHR) 466 error = cdev_ioctl(vp->v_rdev, DIOCGPART, &pi, FREAD, curlwp); 467 else 468 error = bdev_ioctl(vp->v_rdev, DIOCGPART, &pi, FREAD, curlwp); 469 if (error == 0) 470 uvm_vnp_setsize(vp, 471 (voff_t)pi.disklab->d_secsize * pi.part->p_size); 472 return 0; 473 } 474 475 /* 476 * Vnode op for read 477 */ 478 /* ARGSUSED */ 479 int 480 spec_read(void *v) 481 { 482 struct vop_read_args /* { 483 struct vnode *a_vp; 484 struct uio *a_uio; 485 int a_ioflag; 486 kauth_cred_t a_cred; 487 } */ *ap = v; 488 struct vnode *vp = ap->a_vp; 489 struct uio *uio = ap->a_uio; 490 struct lwp *l = curlwp; 491 struct buf *bp; 492 daddr_t bn; 493 int bsize, bscale; 494 struct partinfo dpart; 495 int n, on; 496 int error = 0; 497 498 #ifdef DIAGNOSTIC 499 if (uio->uio_rw != UIO_READ) 500 panic("spec_read mode"); 501 if (&uio->uio_vmspace->vm_map != kernel_map && 502 uio->uio_vmspace != curproc->p_vmspace) 503 panic("spec_read proc"); 504 #endif 505 if (uio->uio_resid == 0) 506 return (0); 507 508 switch (vp->v_type) { 509 510 case VCHR: 511 VOP_UNLOCK(vp, 0); 512 error = cdev_read(vp->v_rdev, uio, ap->a_ioflag); 513 vn_lock(vp, LK_SHARED | LK_RETRY); 514 return (error); 515 516 case VBLK: 517 KASSERT(vp == vp->v_specnode->sn_dev->sd_bdevvp); 518 if (uio->uio_offset < 0) 519 return (EINVAL); 520 bsize = BLKDEV_IOSIZE; 521 if (bdev_ioctl(vp->v_rdev, DIOCGPART, &dpart, FREAD, l) == 0) { 522 if (dpart.part->p_fstype == FS_BSDFFS && 523 dpart.part->p_frag != 0 && dpart.part->p_fsize != 0) 524 bsize = dpart.part->p_frag * 525 dpart.part->p_fsize; 526 } 527 bscale = bsize >> DEV_BSHIFT; 528 do { 529 bn = (uio->uio_offset >> DEV_BSHIFT) &~ (bscale - 1); 530 on = uio->uio_offset % bsize; 531 n = min((unsigned)(bsize - on), uio->uio_resid); 532 error = bread(vp, bn, bsize, NOCRED, &bp); 533 n = min(n, bsize - bp->b_resid); 534 if (error) { 535 brelse(bp, 0); 536 return (error); 537 } 538 error = uiomove((char *)bp->b_data + on, n, uio); 539 brelse(bp, 0); 540 } while (error == 0 && uio->uio_resid > 0 && n != 0); 541 return (error); 542 543 default: 544 panic("spec_read type"); 545 } 546 /* NOTREACHED */ 547 } 548 549 /* 550 * Vnode op for write 551 */ 552 /* ARGSUSED */ 553 int 554 spec_write(void *v) 555 { 556 struct vop_write_args /* { 557 struct vnode *a_vp; 558 struct uio *a_uio; 559 int a_ioflag; 560 kauth_cred_t a_cred; 561 } */ *ap = v; 562 struct vnode *vp = ap->a_vp; 563 struct uio *uio = ap->a_uio; 564 struct lwp *l = curlwp; 565 struct buf *bp; 566 daddr_t bn; 567 int bsize, bscale; 568 struct partinfo dpart; 569 int n, on; 570 int error = 0; 571 572 #ifdef DIAGNOSTIC 573 if (uio->uio_rw != UIO_WRITE) 574 panic("spec_write mode"); 575 if (&uio->uio_vmspace->vm_map != kernel_map && 576 uio->uio_vmspace != curproc->p_vmspace) 577 panic("spec_write proc"); 578 #endif 579 580 switch (vp->v_type) { 581 582 case VCHR: 583 VOP_UNLOCK(vp, 0); 584 error = cdev_write(vp->v_rdev, uio, ap->a_ioflag); 585 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 586 return (error); 587 588 case VBLK: 589 KASSERT(vp == vp->v_specnode->sn_dev->sd_bdevvp); 590 if (uio->uio_resid == 0) 591 return (0); 592 if (uio->uio_offset < 0) 593 return (EINVAL); 594 bsize = BLKDEV_IOSIZE; 595 if (bdev_ioctl(vp->v_rdev, DIOCGPART, &dpart, FREAD, l) == 0) { 596 if (dpart.part->p_fstype == FS_BSDFFS && 597 dpart.part->p_frag != 0 && dpart.part->p_fsize != 0) 598 bsize = dpart.part->p_frag * 599 dpart.part->p_fsize; 600 } 601 bscale = bsize >> DEV_BSHIFT; 602 do { 603 bn = (uio->uio_offset >> DEV_BSHIFT) &~ (bscale - 1); 604 on = uio->uio_offset % bsize; 605 n = min((unsigned)(bsize - on), uio->uio_resid); 606 if (n == bsize) 607 bp = getblk(vp, bn, bsize, 0, 0); 608 else 609 error = bread(vp, bn, bsize, NOCRED, &bp); 610 if (error) { 611 brelse(bp, 0); 612 return (error); 613 } 614 n = min(n, bsize - bp->b_resid); 615 error = uiomove((char *)bp->b_data + on, n, uio); 616 if (error) 617 brelse(bp, 0); 618 else { 619 if (n + on == bsize) 620 bawrite(bp); 621 else 622 bdwrite(bp); 623 error = bp->b_error; 624 } 625 } while (error == 0 && uio->uio_resid > 0 && n != 0); 626 return (error); 627 628 default: 629 panic("spec_write type"); 630 } 631 /* NOTREACHED */ 632 } 633 634 /* 635 * Device ioctl operation. 636 */ 637 /* ARGSUSED */ 638 int 639 spec_ioctl(void *v) 640 { 641 struct vop_ioctl_args /* { 642 struct vnode *a_vp; 643 u_long a_command; 644 void *a_data; 645 int a_fflag; 646 kauth_cred_t a_cred; 647 } */ *ap = v; 648 struct vnode *vp; 649 dev_t dev; 650 651 /* 652 * Extract all the info we need from the vnode, taking care to 653 * avoid a race with VOP_REVOKE(). 654 */ 655 656 vp = ap->a_vp; 657 dev = NODEV; 658 mutex_enter(&vp->v_interlock); 659 if ((vp->v_iflag & VI_XLOCK) == 0 && vp->v_specnode) { 660 dev = vp->v_rdev; 661 } 662 mutex_exit(&vp->v_interlock); 663 if (dev == NODEV) { 664 return ENXIO; 665 } 666 667 switch (vp->v_type) { 668 669 case VCHR: 670 return cdev_ioctl(dev, ap->a_command, ap->a_data, 671 ap->a_fflag, curlwp); 672 673 case VBLK: 674 KASSERT(vp == vp->v_specnode->sn_dev->sd_bdevvp); 675 return bdev_ioctl(dev, ap->a_command, ap->a_data, 676 ap->a_fflag, curlwp); 677 678 default: 679 panic("spec_ioctl"); 680 /* NOTREACHED */ 681 } 682 } 683 684 /* ARGSUSED */ 685 int 686 spec_poll(void *v) 687 { 688 struct vop_poll_args /* { 689 struct vnode *a_vp; 690 int a_events; 691 } */ *ap = v; 692 struct vnode *vp; 693 dev_t dev; 694 695 /* 696 * Extract all the info we need from the vnode, taking care to 697 * avoid a race with VOP_REVOKE(). 698 */ 699 700 vp = ap->a_vp; 701 dev = NODEV; 702 mutex_enter(&vp->v_interlock); 703 if ((vp->v_iflag & VI_XLOCK) == 0 && vp->v_specnode) { 704 dev = vp->v_rdev; 705 } 706 mutex_exit(&vp->v_interlock); 707 if (dev == NODEV) { 708 return POLLERR; 709 } 710 711 switch (vp->v_type) { 712 713 case VCHR: 714 return cdev_poll(dev, ap->a_events, curlwp); 715 716 default: 717 return (genfs_poll(v)); 718 } 719 } 720 721 /* ARGSUSED */ 722 int 723 spec_kqfilter(void *v) 724 { 725 struct vop_kqfilter_args /* { 726 struct vnode *a_vp; 727 struct proc *a_kn; 728 } */ *ap = v; 729 dev_t dev; 730 731 switch (ap->a_vp->v_type) { 732 733 case VCHR: 734 dev = ap->a_vp->v_rdev; 735 return cdev_kqfilter(dev, ap->a_kn); 736 default: 737 /* 738 * Block devices don't support kqfilter, and refuse it 739 * for any other files (like those vflush()ed) too. 740 */ 741 return (EOPNOTSUPP); 742 } 743 } 744 745 /* 746 * Allow mapping of only D_DISK. This is called only for VBLK. 747 */ 748 int 749 spec_mmap(void *v) 750 { 751 struct vop_mmap_args /* { 752 struct vnode *a_vp; 753 vm_prot_t a_prot; 754 kauth_cred_t a_cred; 755 } */ *ap = v; 756 struct vnode *vp = ap->a_vp; 757 758 KASSERT(vp->v_type == VBLK); 759 if (bdev_type(vp->v_rdev) != D_DISK) 760 return EINVAL; 761 762 return 0; 763 } 764 765 /* 766 * Synch buffers associated with a block device 767 */ 768 /* ARGSUSED */ 769 int 770 spec_fsync(void *v) 771 { 772 struct vop_fsync_args /* { 773 struct vnode *a_vp; 774 kauth_cred_t a_cred; 775 int a_flags; 776 off_t offlo; 777 off_t offhi; 778 } */ *ap = v; 779 struct vnode *vp = ap->a_vp; 780 781 if (vp->v_type == VBLK) { 782 KASSERT(vp == vp->v_specnode->sn_dev->sd_bdevvp); 783 vflushbuf(vp, (ap->a_flags & FSYNC_WAIT) != 0); 784 } 785 return (0); 786 } 787 788 /* 789 * Just call the device strategy routine 790 */ 791 int 792 spec_strategy(void *v) 793 { 794 struct vop_strategy_args /* { 795 struct vnode *a_vp; 796 struct buf *a_bp; 797 } */ *ap = v; 798 struct vnode *vp = ap->a_vp; 799 struct buf *bp = ap->a_bp; 800 int error; 801 802 KASSERT(vp == vp->v_specnode->sn_dev->sd_bdevvp); 803 804 error = 0; 805 bp->b_dev = vp->v_rdev; 806 if (!(bp->b_flags & B_READ) && 807 (LIST_FIRST(&bp->b_dep)) != NULL && bioopsp) 808 bioopsp->io_start(bp); 809 810 if (!(bp->b_flags & B_READ)) 811 error = fscow_run(bp, false); 812 813 if (error) { 814 bp->b_error = error; 815 biodone(bp); 816 return (error); 817 } 818 819 bdev_strategy(bp); 820 821 return (0); 822 } 823 824 int 825 spec_inactive(void *v) 826 { 827 struct vop_inactive_args /* { 828 struct vnode *a_vp; 829 struct proc *a_l; 830 } */ *ap = v; 831 832 VOP_UNLOCK(ap->a_vp, 0); 833 return (0); 834 } 835 836 /* 837 * This is a noop, simply returning what one has been given. 838 */ 839 int 840 spec_bmap(void *v) 841 { 842 struct vop_bmap_args /* { 843 struct vnode *a_vp; 844 daddr_t a_bn; 845 struct vnode **a_vpp; 846 daddr_t *a_bnp; 847 int *a_runp; 848 } */ *ap = v; 849 850 if (ap->a_vpp != NULL) 851 *ap->a_vpp = ap->a_vp; 852 if (ap->a_bnp != NULL) 853 *ap->a_bnp = ap->a_bn; 854 if (ap->a_runp != NULL) 855 *ap->a_runp = (MAXBSIZE >> DEV_BSHIFT) - 1; 856 return (0); 857 } 858 859 /* 860 * Device close routine 861 */ 862 /* ARGSUSED */ 863 int 864 spec_close(void *v) 865 { 866 struct vop_close_args /* { 867 struct vnode *a_vp; 868 int a_fflag; 869 kauth_cred_t a_cred; 870 } */ *ap = v; 871 struct vnode *vp = ap->a_vp; 872 struct session *sess; 873 dev_t dev = vp->v_rdev; 874 int mode, error, flags, flags1, count; 875 specnode_t *sn; 876 specdev_t *sd; 877 878 flags = vp->v_iflag; 879 sn = vp->v_specnode; 880 sd = sn->sn_dev; 881 882 switch (vp->v_type) { 883 884 case VCHR: 885 /* 886 * Hack: a tty device that is a controlling terminal 887 * has a reference from the session structure. We 888 * cannot easily tell that a character device is a 889 * controlling terminal, unless it is the closing 890 * process' controlling terminal. In that case, if the 891 * open count is 1 release the reference from the 892 * session. Also, remove the link from the tty back to 893 * the session and pgrp. 894 * 895 * XXX V. fishy. 896 */ 897 mutex_enter(&proclist_lock); 898 sess = curlwp->l_proc->p_session; 899 if (sn->sn_opencnt == 1 && vp == sess->s_ttyvp) { 900 mutex_spin_enter(&tty_lock); 901 sess->s_ttyvp = NULL; 902 if (sess->s_ttyp->t_session != NULL) { 903 sess->s_ttyp->t_pgrp = NULL; 904 sess->s_ttyp->t_session = NULL; 905 mutex_spin_exit(&tty_lock); 906 SESSRELE(sess); 907 mutex_exit(&proclist_lock); 908 } else { 909 mutex_spin_exit(&tty_lock); 910 if (sess->s_ttyp->t_pgrp != NULL) 911 panic("spec_close: spurious pgrp ref"); 912 mutex_exit(&proclist_lock); 913 } 914 vrele(vp); 915 } else 916 mutex_exit(&proclist_lock); 917 918 /* 919 * If the vnode is locked, then we are in the midst 920 * of forcably closing the device, otherwise we only 921 * close on last reference. 922 */ 923 mode = S_IFCHR; 924 break; 925 926 case VBLK: 927 KASSERT(vp == vp->v_specnode->sn_dev->sd_bdevvp); 928 /* 929 * On last close of a block device (that isn't mounted) 930 * we must invalidate any in core blocks, so that 931 * we can, for instance, change floppy disks. 932 */ 933 error = vinvalbuf(vp, V_SAVE, ap->a_cred, curlwp, 0, 0); 934 if (error) 935 return (error); 936 /* 937 * We do not want to really close the device if it 938 * is still in use unless we are trying to close it 939 * forcibly. Since every use (buffer, vnode, swap, cmap) 940 * holds a reference to the vnode, and because we mark 941 * any other vnodes that alias this device, when the 942 * sum of the reference counts on all the aliased 943 * vnodes descends to one, we are on last close. 944 */ 945 mode = S_IFBLK; 946 break; 947 948 default: 949 panic("spec_close: not special"); 950 } 951 952 mutex_enter(&specfs_lock); 953 sn->sn_opencnt--; 954 count = --sd->sd_opencnt; 955 if (vp->v_type == VBLK) 956 sd->sd_bdevvp = NULL; 957 mutex_exit(&specfs_lock); 958 959 if (count != 0) 960 return 0; 961 962 flags1 = ap->a_fflag; 963 964 /* 965 * if VI_XLOCK is set, then we're going away soon, so make this 966 * non-blocking. Also ensures that we won't wedge in vn_lock below. 967 */ 968 if (flags & VI_XLOCK) 969 flags1 |= FNONBLOCK; 970 971 /* 972 * If we're able to block, release the vnode lock & reacquire. We 973 * might end up sleeping for someone else who wants our queues. They 974 * won't get them if we hold the vnode locked. Also, if VI_XLOCK is 975 * set, don't release the lock as we won't be able to regain it. 976 */ 977 if (!(flags1 & FNONBLOCK)) 978 VOP_UNLOCK(vp, 0); 979 980 if (vp->v_type == VBLK) 981 error = bdev_close(dev, flags1, mode, curlwp); 982 else 983 error = cdev_close(dev, flags1, mode, curlwp); 984 985 if (!(flags1 & FNONBLOCK)) 986 vn_lock(vp, LK_EXCLUSIVE | LK_RETRY); 987 988 return (error); 989 } 990 991 /* 992 * Print out the contents of a special device vnode. 993 */ 994 int 995 spec_print(void *v) 996 { 997 struct vop_print_args /* { 998 struct vnode *a_vp; 999 } */ *ap = v; 1000 1001 printf("dev %d, %d\n", major(ap->a_vp->v_rdev), 1002 minor(ap->a_vp->v_rdev)); 1003 return 0; 1004 } 1005 1006 /* 1007 * Return POSIX pathconf information applicable to special devices. 1008 */ 1009 int 1010 spec_pathconf(void *v) 1011 { 1012 struct vop_pathconf_args /* { 1013 struct vnode *a_vp; 1014 int a_name; 1015 register_t *a_retval; 1016 } */ *ap = v; 1017 1018 switch (ap->a_name) { 1019 case _PC_LINK_MAX: 1020 *ap->a_retval = LINK_MAX; 1021 return (0); 1022 case _PC_MAX_CANON: 1023 *ap->a_retval = MAX_CANON; 1024 return (0); 1025 case _PC_MAX_INPUT: 1026 *ap->a_retval = MAX_INPUT; 1027 return (0); 1028 case _PC_PIPE_BUF: 1029 *ap->a_retval = PIPE_BUF; 1030 return (0); 1031 case _PC_CHOWN_RESTRICTED: 1032 *ap->a_retval = 1; 1033 return (0); 1034 case _PC_VDISABLE: 1035 *ap->a_retval = _POSIX_VDISABLE; 1036 return (0); 1037 case _PC_SYNC_IO: 1038 *ap->a_retval = 1; 1039 return (0); 1040 default: 1041 return (EINVAL); 1042 } 1043 /* NOTREACHED */ 1044 } 1045 1046 /* 1047 * Advisory record locking support. 1048 */ 1049 int 1050 spec_advlock(void *v) 1051 { 1052 struct vop_advlock_args /* { 1053 struct vnode *a_vp; 1054 void *a_id; 1055 int a_op; 1056 struct flock *a_fl; 1057 int a_flags; 1058 } */ *ap = v; 1059 struct vnode *vp = ap->a_vp; 1060 1061 return lf_advlock(ap, &vp->v_speclockf, (off_t)0); 1062 } 1063
Indexes created Wed Sep 24 22:09:59 GMT 2025