1 /* $NetBSD: dksubr.c,v 1.100 2017/10/29 09:44:17 mlelstv Exp $ */ 2 3 /*- 4 * Copyright (c) 1996, 1997, 1998, 1999, 2002, 2008 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * This code is derived from software contributed to The NetBSD Foundation 8 * by Jason R. Thorpe and Roland C. Dowdeswell. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 20 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 21 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 22 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 23 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 24 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 25 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 26 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 27 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 28 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 29 * POSSIBILITY OF SUCH DAMAGE. 30 */ 31 32 #include <sys/cdefs.h> 33 __KERNEL_RCSID(0, "$NetBSD: dksubr.c,v 1.100 2017/10/29 09:44:17 mlelstv Exp $"); 34 35 #include <sys/param.h> 36 #include <sys/systm.h> 37 #include <sys/stat.h> 38 #include <sys/proc.h> 39 #include <sys/ioctl.h> 40 #include <sys/device.h> 41 #include <sys/disk.h> 42 #include <sys/disklabel.h> 43 #include <sys/buf.h> 44 #include <sys/bufq.h> 45 #include <sys/vnode.h> 46 #include <sys/fcntl.h> 47 #include <sys/namei.h> 48 #include <sys/module.h> 49 #include <sys/syslog.h> 50 51 #include <dev/dkvar.h> 52 #include <miscfs/specfs/specdev.h> /* for v_rdev */ 53 54 int dkdebug = 0; 55 56 #ifdef DEBUG 57 #define DKDB_FOLLOW 0x1 58 #define DKDB_INIT 0x2 59 #define DKDB_VNODE 0x4 60 #define DKDB_DUMP 0x8 61 62 #define IFDEBUG(x,y) if (dkdebug & (x)) y 63 #define DPRINTF(x,y) IFDEBUG(x, printf y) 64 #define DPRINTF_FOLLOW(y) DPRINTF(DKDB_FOLLOW, y) 65 #else 66 #define IFDEBUG(x,y) 67 #define DPRINTF(x,y) 68 #define DPRINTF_FOLLOW(y) 69 #endif 70 71 #define DKF_READYFORDUMP (DKF_INITED|DKF_TAKEDUMP) 72 73 static int dk_subr_modcmd(modcmd_t, void *); 74 75 #define DKLABELDEV(dev) \ 76 (MAKEDISKDEV(major((dev)), DISKUNIT((dev)), RAW_PART)) 77 78 static void dk_makedisklabel(struct dk_softc *); 79 static int dk_translate(struct dk_softc *, struct buf *); 80 static void dk_done1(struct dk_softc *, struct buf *, bool); 81 82 void 83 dk_init(struct dk_softc *dksc, device_t dev, int dtype) 84 { 85 86 memset(dksc, 0x0, sizeof(*dksc)); 87 dksc->sc_dtype = dtype; 88 dksc->sc_dev = dev; 89 90 strlcpy(dksc->sc_xname, device_xname(dev), DK_XNAME_SIZE); 91 dksc->sc_dkdev.dk_name = dksc->sc_xname; 92 } 93 94 void 95 dk_attach(struct dk_softc *dksc) 96 { 97 KASSERT(dksc->sc_dev != NULL); 98 99 mutex_init(&dksc->sc_iolock, MUTEX_DEFAULT, IPL_VM); 100 dksc->sc_flags |= DKF_READYFORDUMP; 101 #ifdef DIAGNOSTIC 102 dksc->sc_flags |= DKF_WARNLABEL | DKF_LABELSANITY; 103 #endif 104 105 if ((dksc->sc_flags & DKF_NO_RND) == 0) { 106 /* Attach the device into the rnd source list. */ 107 rnd_attach_source(&dksc->sc_rnd_source, dksc->sc_xname, 108 RND_TYPE_DISK, RND_FLAG_DEFAULT); 109 } 110 } 111 112 void 113 dk_detach(struct dk_softc *dksc) 114 { 115 if ((dksc->sc_flags & DKF_NO_RND) == 0) { 116 /* Unhook the entropy source. */ 117 rnd_detach_source(&dksc->sc_rnd_source); 118 } 119 120 dksc->sc_flags &= ~DKF_READYFORDUMP; 121 mutex_destroy(&dksc->sc_iolock); 122 } 123 124 /* ARGSUSED */ 125 int 126 dk_open(struct dk_softc *dksc, dev_t dev, 127 int flags, int fmt, struct lwp *l) 128 { 129 const struct dkdriver *dkd = dksc->sc_dkdev.dk_driver; 130 struct disklabel *lp = dksc->sc_dkdev.dk_label; 131 int part = DISKPART(dev); 132 int pmask = 1 << part; 133 int ret = 0; 134 struct disk *dk = &dksc->sc_dkdev; 135 136 DPRINTF_FOLLOW(("%s(%s, %p, 0x%"PRIx64", 0x%x)\n", __func__, 137 dksc->sc_xname, dksc, dev, flags)); 138 139 mutex_enter(&dk->dk_openlock); 140 141 /* 142 * If there are wedges, and this is not RAW_PART, then we 143 * need to fail. 144 */ 145 if (dk->dk_nwedges != 0 && part != RAW_PART) { 146 ret = EBUSY; 147 goto done; 148 } 149 150 /* 151 * initialize driver for the first opener 152 */ 153 if (dk->dk_openmask == 0 && dkd->d_firstopen != NULL) { 154 ret = (*dkd->d_firstopen)(dksc->sc_dev, dev, flags, fmt); 155 if (ret) 156 goto done; 157 } 158 159 /* 160 * If we're init'ed and there are no other open partitions then 161 * update the in-core disklabel. 162 */ 163 if ((dksc->sc_flags & DKF_INITED)) { 164 if ((dksc->sc_flags & DKF_VLABEL) == 0) { 165 dksc->sc_flags |= DKF_VLABEL; 166 dk_getdisklabel(dksc, dev); 167 } 168 } 169 170 /* Fail if we can't find the partition. */ 171 if (part != RAW_PART && 172 ((dksc->sc_flags & DKF_VLABEL) == 0 || 173 part >= lp->d_npartitions || 174 lp->d_partitions[part].p_fstype == FS_UNUSED)) { 175 ret = ENXIO; 176 goto done; 177 } 178 179 /* Mark our unit as open. */ 180 switch (fmt) { 181 case S_IFCHR: 182 dk->dk_copenmask |= pmask; 183 break; 184 case S_IFBLK: 185 dk->dk_bopenmask |= pmask; 186 break; 187 } 188 189 dk->dk_openmask = dk->dk_copenmask | dk->dk_bopenmask; 190 191 done: 192 mutex_exit(&dk->dk_openlock); 193 return ret; 194 } 195 196 /* ARGSUSED */ 197 int 198 dk_close(struct dk_softc *dksc, dev_t dev, 199 int flags, int fmt, struct lwp *l) 200 { 201 const struct dkdriver *dkd = dksc->sc_dkdev.dk_driver; 202 int part = DISKPART(dev); 203 int pmask = 1 << part; 204 struct disk *dk = &dksc->sc_dkdev; 205 206 DPRINTF_FOLLOW(("%s(%s, %p, 0x%"PRIx64", 0x%x)\n", __func__, 207 dksc->sc_xname, dksc, dev, flags)); 208 209 mutex_enter(&dk->dk_openlock); 210 211 switch (fmt) { 212 case S_IFCHR: 213 dk->dk_copenmask &= ~pmask; 214 break; 215 case S_IFBLK: 216 dk->dk_bopenmask &= ~pmask; 217 break; 218 } 219 dk->dk_openmask = dk->dk_copenmask | dk->dk_bopenmask; 220 221 if (dk->dk_openmask == 0) { 222 if (dkd->d_lastclose != NULL) 223 (*dkd->d_lastclose)(dksc->sc_dev); 224 if ((dksc->sc_flags & DKF_KLABEL) == 0) 225 dksc->sc_flags &= ~DKF_VLABEL; 226 } 227 228 mutex_exit(&dk->dk_openlock); 229 return 0; 230 } 231 232 static int 233 dk_translate(struct dk_softc *dksc, struct buf *bp) 234 { 235 int part; 236 int wlabel; 237 daddr_t blkno; 238 struct disklabel *lp; 239 struct disk *dk; 240 uint64_t numsecs; 241 unsigned secsize; 242 243 lp = dksc->sc_dkdev.dk_label; 244 dk = &dksc->sc_dkdev; 245 246 part = DISKPART(bp->b_dev); 247 numsecs = dk->dk_geom.dg_secperunit; 248 secsize = dk->dk_geom.dg_secsize; 249 250 /* 251 * The transfer must be a whole number of blocks and the offset must 252 * not be negative. 253 */ 254 if ((bp->b_bcount % secsize) != 0 || bp->b_blkno < 0) { 255 bp->b_error = EINVAL; 256 goto done; 257 } 258 259 /* If there is nothing to do, then we are done */ 260 if (bp->b_bcount == 0) 261 goto done; 262 263 wlabel = dksc->sc_flags & (DKF_WLABEL|DKF_LABELLING); 264 if (part == RAW_PART) { 265 uint64_t numblocks = btodb(numsecs * secsize); 266 if (bounds_check_with_mediasize(bp, DEV_BSIZE, numblocks) <= 0) 267 goto done; 268 } else { 269 if (bounds_check_with_label(&dksc->sc_dkdev, bp, wlabel) <= 0) 270 goto done; 271 } 272 273 /* 274 * Convert the block number to absolute and put it in terms 275 * of the device's logical block size. 276 */ 277 if (secsize >= DEV_BSIZE) 278 blkno = bp->b_blkno / (secsize / DEV_BSIZE); 279 else 280 blkno = bp->b_blkno * (DEV_BSIZE / secsize); 281 282 if (part != RAW_PART) 283 blkno += lp->d_partitions[DISKPART(bp->b_dev)].p_offset; 284 bp->b_rawblkno = blkno; 285 286 return -1; 287 288 done: 289 bp->b_resid = bp->b_bcount; 290 return bp->b_error; 291 } 292 293 static int 294 dk_strategy1(struct dk_softc *dksc, struct buf *bp) 295 { 296 int error; 297 298 DPRINTF_FOLLOW(("%s(%s, %p, %p)\n", __func__, 299 dksc->sc_xname, dksc, bp)); 300 301 if (!(dksc->sc_flags & DKF_INITED)) { 302 DPRINTF_FOLLOW(("%s: not inited\n", __func__)); 303 bp->b_error = ENXIO; 304 bp->b_resid = bp->b_bcount; 305 biodone(bp); 306 return 1; 307 } 308 309 error = dk_translate(dksc, bp); 310 if (error >= 0) { 311 biodone(bp); 312 return 1; 313 } 314 315 return 0; 316 } 317 318 void 319 dk_strategy(struct dk_softc *dksc, struct buf *bp) 320 { 321 int error; 322 323 error = dk_strategy1(dksc, bp); 324 if (error) 325 return; 326 327 /* 328 * Queue buffer and start unit 329 */ 330 dk_start(dksc, bp); 331 } 332 333 int 334 dk_strategy_defer(struct dk_softc *dksc, struct buf *bp) 335 { 336 int error; 337 338 error = dk_strategy1(dksc, bp); 339 if (error) 340 return error; 341 342 /* 343 * Queue buffer only 344 */ 345 mutex_enter(&dksc->sc_iolock); 346 disk_wait(&dksc->sc_dkdev); 347 bufq_put(dksc->sc_bufq, bp); 348 mutex_exit(&dksc->sc_iolock); 349 350 return 0; 351 } 352 353 int 354 dk_strategy_pending(struct dk_softc *dksc) 355 { 356 struct buf *bp; 357 358 if (!(dksc->sc_flags & DKF_INITED)) { 359 DPRINTF_FOLLOW(("%s: not inited\n", __func__)); 360 return 0; 361 } 362 363 mutex_enter(&dksc->sc_iolock); 364 bp = bufq_peek(dksc->sc_bufq); 365 mutex_exit(&dksc->sc_iolock); 366 367 return bp != NULL; 368 } 369 370 void 371 dk_start(struct dk_softc *dksc, struct buf *bp) 372 { 373 const struct dkdriver *dkd = dksc->sc_dkdev.dk_driver; 374 int error; 375 376 if (!(dksc->sc_flags & DKF_INITED)) { 377 DPRINTF_FOLLOW(("%s: not inited\n", __func__)); 378 return; 379 } 380 381 mutex_enter(&dksc->sc_iolock); 382 383 if (bp != NULL) { 384 disk_wait(&dksc->sc_dkdev); 385 bufq_put(dksc->sc_bufq, bp); 386 } 387 388 /* 389 * If another thread is running the queue, increment 390 * busy counter to 2 so that the queue is retried, 391 * because the driver may now accept additional 392 * requests. 393 */ 394 if (dksc->sc_busy < 2) 395 dksc->sc_busy++; 396 if (dksc->sc_busy > 1) 397 goto done; 398 399 /* 400 * Peeking at the buffer queue and committing the operation 401 * only after success isn't atomic. 402 * 403 * So when a diskstart fails, the buffer is saved 404 * and tried again before the next buffer is fetched. 405 * dk_drain() handles flushing of a saved buffer. 406 * 407 * This keeps order of I/O operations, unlike bufq_put. 408 */ 409 410 while (dksc->sc_busy > 0) { 411 412 bp = dksc->sc_deferred; 413 dksc->sc_deferred = NULL; 414 415 if (bp == NULL) 416 bp = bufq_get(dksc->sc_bufq); 417 418 while (bp != NULL) { 419 420 disk_busy(&dksc->sc_dkdev); 421 mutex_exit(&dksc->sc_iolock); 422 error = dkd->d_diskstart(dksc->sc_dev, bp); 423 mutex_enter(&dksc->sc_iolock); 424 if (error == EAGAIN) { 425 dksc->sc_deferred = bp; 426 disk_unbusy(&dksc->sc_dkdev, 0, (bp->b_flags & B_READ)); 427 disk_wait(&dksc->sc_dkdev); 428 break; 429 } 430 431 if (error != 0) { 432 bp->b_error = error; 433 bp->b_resid = bp->b_bcount; 434 dk_done1(dksc, bp, false); 435 } 436 437 bp = bufq_get(dksc->sc_bufq); 438 } 439 440 dksc->sc_busy--; 441 } 442 done: 443 mutex_exit(&dksc->sc_iolock); 444 } 445 446 static void 447 dk_done1(struct dk_softc *dksc, struct buf *bp, bool lock) 448 { 449 struct disk *dk = &dksc->sc_dkdev; 450 451 if (bp->b_error != 0) { 452 struct cfdriver *cd = device_cfdriver(dksc->sc_dev); 453 454 diskerr(bp, cd->cd_name, "error", LOG_PRINTF, 0, 455 dk->dk_label); 456 printf("\n"); 457 } 458 459 if (lock) 460 mutex_enter(&dksc->sc_iolock); 461 disk_unbusy(dk, bp->b_bcount - bp->b_resid, (bp->b_flags & B_READ)); 462 if (lock) 463 mutex_exit(&dksc->sc_iolock); 464 465 if ((dksc->sc_flags & DKF_NO_RND) == 0) 466 rnd_add_uint32(&dksc->sc_rnd_source, bp->b_rawblkno); 467 468 biodone(bp); 469 } 470 471 void 472 dk_done(struct dk_softc *dksc, struct buf *bp) 473 { 474 dk_done1(dksc, bp, true); 475 } 476 477 void 478 dk_drain(struct dk_softc *dksc) 479 { 480 struct buf *bp; 481 482 mutex_enter(&dksc->sc_iolock); 483 bp = dksc->sc_deferred; 484 dksc->sc_deferred = NULL; 485 if (bp != NULL) { 486 bp->b_error = EIO; 487 bp->b_resid = bp->b_bcount; 488 biodone(bp); 489 } 490 bufq_drain(dksc->sc_bufq); 491 mutex_exit(&dksc->sc_iolock); 492 } 493 494 int 495 dk_discard(struct dk_softc *dksc, dev_t dev, off_t pos, off_t len) 496 { 497 const struct dkdriver *dkd = dksc->sc_dkdev.dk_driver; 498 unsigned secsize = dksc->sc_dkdev.dk_geom.dg_secsize; 499 struct buf tmp, *bp = &tmp; 500 int maxsz; 501 int error = 0; 502 503 KASSERT(len >= 0); 504 505 DPRINTF_FOLLOW(("%s(%s, %p, 0x"PRIx64", %jd, %jd)\n", __func__, 506 dksc->sc_xname, dksc, (intmax_t)pos, (intmax_t)len)); 507 508 if (!(dksc->sc_flags & DKF_INITED)) { 509 DPRINTF_FOLLOW(("%s: not inited\n", __func__)); 510 return ENXIO; 511 } 512 513 if (secsize == 0 || (pos % secsize) != 0 || (len % secsize) != 0) 514 return EINVAL; 515 516 /* largest value that b_bcount can store */ 517 maxsz = rounddown(INT_MAX, secsize); 518 519 while (len > 0) { 520 /* enough data to please the bounds checking code */ 521 bp->b_dev = dev; 522 bp->b_blkno = (daddr_t)(pos / secsize); 523 bp->b_bcount = min(len, maxsz); 524 bp->b_flags = B_WRITE; 525 526 error = dk_translate(dksc, bp); 527 if (error >= 0) 528 break; 529 530 error = dkd->d_discard(dksc->sc_dev, 531 (off_t)bp->b_rawblkno * secsize, 532 (off_t)bp->b_bcount); 533 if (error) 534 break; 535 536 pos += bp->b_bcount; 537 len -= bp->b_bcount; 538 } 539 540 return error; 541 } 542 543 int 544 dk_size(struct dk_softc *dksc, dev_t dev) 545 { 546 const struct dkdriver *dkd = dksc->sc_dkdev.dk_driver; 547 struct disklabel *lp; 548 int is_open; 549 int part; 550 int size; 551 552 if ((dksc->sc_flags & DKF_INITED) == 0) 553 return -1; 554 555 part = DISKPART(dev); 556 is_open = dksc->sc_dkdev.dk_openmask & (1 << part); 557 558 if (!is_open && dkd->d_open(dev, 0, S_IFBLK, curlwp)) 559 return -1; 560 561 lp = dksc->sc_dkdev.dk_label; 562 if (lp->d_partitions[part].p_fstype != FS_SWAP) 563 size = -1; 564 else 565 size = lp->d_partitions[part].p_size * 566 (lp->d_secsize / DEV_BSIZE); 567 568 if (!is_open && dkd->d_close(dev, 0, S_IFBLK, curlwp)) 569 return -1; 570 571 return size; 572 } 573 574 int 575 dk_ioctl(struct dk_softc *dksc, dev_t dev, 576 u_long cmd, void *data, int flag, struct lwp *l) 577 { 578 const struct dkdriver *dkd = dksc->sc_dkdev.dk_driver; 579 struct disklabel *lp; 580 struct disk *dk = &dksc->sc_dkdev; 581 #ifdef __HAVE_OLD_DISKLABEL 582 struct disklabel newlabel; 583 #endif 584 int error; 585 586 DPRINTF_FOLLOW(("%s(%s, %p, 0x%"PRIx64", 0x%lx)\n", __func__, 587 dksc->sc_xname, dksc, dev, cmd)); 588 589 /* ensure that the pseudo disk is open for writes for these commands */ 590 switch (cmd) { 591 case DIOCSDINFO: 592 case DIOCWDINFO: 593 #ifdef __HAVE_OLD_DISKLABEL 594 case ODIOCSDINFO: 595 case ODIOCWDINFO: 596 #endif 597 case DIOCKLABEL: 598 case DIOCWLABEL: 599 case DIOCAWEDGE: 600 case DIOCDWEDGE: 601 case DIOCSSTRATEGY: 602 if ((flag & FWRITE) == 0) 603 return EBADF; 604 } 605 606 /* ensure that the pseudo-disk is initialized for these */ 607 switch (cmd) { 608 case DIOCGDINFO: 609 case DIOCSDINFO: 610 case DIOCWDINFO: 611 case DIOCGPARTINFO: 612 case DIOCKLABEL: 613 case DIOCWLABEL: 614 case DIOCGDEFLABEL: 615 case DIOCAWEDGE: 616 case DIOCDWEDGE: 617 case DIOCLWEDGES: 618 case DIOCMWEDGES: 619 case DIOCCACHESYNC: 620 #ifdef __HAVE_OLD_DISKLABEL 621 case ODIOCGDINFO: 622 case ODIOCSDINFO: 623 case ODIOCWDINFO: 624 case ODIOCGDEFLABEL: 625 #endif 626 if ((dksc->sc_flags & DKF_INITED) == 0) 627 return ENXIO; 628 } 629 630 error = disk_ioctl(dk, dev, cmd, data, flag, l); 631 if (error != EPASSTHROUGH) 632 return error; 633 else 634 error = 0; 635 636 switch (cmd) { 637 case DIOCWDINFO: 638 case DIOCSDINFO: 639 #ifdef __HAVE_OLD_DISKLABEL 640 case ODIOCWDINFO: 641 case ODIOCSDINFO: 642 #endif 643 #ifdef __HAVE_OLD_DISKLABEL 644 if (cmd == ODIOCSDINFO || cmd == ODIOCWDINFO) { 645 memset(&newlabel, 0, sizeof newlabel); 646 memcpy(&newlabel, data, sizeof (struct olddisklabel)); 647 lp = &newlabel; 648 } else 649 #endif 650 lp = (struct disklabel *)data; 651 652 mutex_enter(&dk->dk_openlock); 653 dksc->sc_flags |= DKF_LABELLING; 654 655 error = setdisklabel(dksc->sc_dkdev.dk_label, 656 lp, 0, dksc->sc_dkdev.dk_cpulabel); 657 if (error == 0) { 658 if (cmd == DIOCWDINFO 659 #ifdef __HAVE_OLD_DISKLABEL 660 || cmd == ODIOCWDINFO 661 #endif 662 ) 663 error = writedisklabel(DKLABELDEV(dev), 664 dkd->d_strategy, dksc->sc_dkdev.dk_label, 665 dksc->sc_dkdev.dk_cpulabel); 666 } 667 668 dksc->sc_flags &= ~DKF_LABELLING; 669 mutex_exit(&dk->dk_openlock); 670 break; 671 672 case DIOCKLABEL: 673 if (*(int *)data != 0) 674 dksc->sc_flags |= DKF_KLABEL; 675 else 676 dksc->sc_flags &= ~DKF_KLABEL; 677 break; 678 679 case DIOCWLABEL: 680 if (*(int *)data != 0) 681 dksc->sc_flags |= DKF_WLABEL; 682 else 683 dksc->sc_flags &= ~DKF_WLABEL; 684 break; 685 686 case DIOCGDEFLABEL: 687 dk_getdefaultlabel(dksc, (struct disklabel *)data); 688 break; 689 690 #ifdef __HAVE_OLD_DISKLABEL 691 case ODIOCGDEFLABEL: 692 dk_getdefaultlabel(dksc, &newlabel); 693 if (newlabel.d_npartitions > OLDMAXPARTITIONS) 694 return ENOTTY; 695 memcpy(data, &newlabel, sizeof (struct olddisklabel)); 696 break; 697 #endif 698 699 case DIOCGSTRATEGY: 700 { 701 struct disk_strategy *dks = (void *)data; 702 703 mutex_enter(&dksc->sc_iolock); 704 if (dksc->sc_bufq != NULL) 705 strlcpy(dks->dks_name, 706 bufq_getstrategyname(dksc->sc_bufq), 707 sizeof(dks->dks_name)); 708 else 709 error = EINVAL; 710 mutex_exit(&dksc->sc_iolock); 711 dks->dks_paramlen = 0; 712 break; 713 } 714 715 case DIOCSSTRATEGY: 716 { 717 struct disk_strategy *dks = (void *)data; 718 struct bufq_state *new; 719 struct bufq_state *old; 720 721 if (dks->dks_param != NULL) { 722 return EINVAL; 723 } 724 dks->dks_name[sizeof(dks->dks_name) - 1] = 0; /* ensure term */ 725 error = bufq_alloc(&new, dks->dks_name, 726 BUFQ_EXACT|BUFQ_SORT_RAWBLOCK); 727 if (error) { 728 return error; 729 } 730 mutex_enter(&dksc->sc_iolock); 731 old = dksc->sc_bufq; 732 if (old) 733 bufq_move(new, old); 734 dksc->sc_bufq = new; 735 mutex_exit(&dksc->sc_iolock); 736 if (old) 737 bufq_free(old); 738 break; 739 } 740 741 default: 742 error = ENOTTY; 743 } 744 745 return error; 746 } 747 748 /* 749 * dk_dump dumps all of physical memory into the partition specified. 750 * This requires substantially more framework than {s,w}ddump, and hence 751 * is probably much more fragile. 752 * 753 */ 754 755 #define DKFF_READYFORDUMP(x) (((x) & DKF_READYFORDUMP) == DKF_READYFORDUMP) 756 static volatile int dk_dumping = 0; 757 758 /* ARGSUSED */ 759 int 760 dk_dump(struct dk_softc *dksc, dev_t dev, 761 daddr_t blkno, void *vav, size_t size) 762 { 763 const struct dkdriver *dkd = dksc->sc_dkdev.dk_driver; 764 char *va = vav; 765 struct disklabel *lp; 766 struct partition *p; 767 int part, towrt, nsects, sectoff, maxblkcnt, nblk; 768 int maxxfer, rv = 0; 769 770 /* 771 * ensure that we consider this device to be safe for dumping, 772 * and that the device is configured. 773 */ 774 if (!DKFF_READYFORDUMP(dksc->sc_flags)) { 775 DPRINTF(DKDB_DUMP, ("%s: bad dump flags 0x%x\n", __func__, 776 dksc->sc_flags)); 777 return ENXIO; 778 } 779 780 /* ensure that we are not already dumping */ 781 if (dk_dumping) 782 return EFAULT; 783 dk_dumping = 1; 784 785 if (dkd->d_dumpblocks == NULL) { 786 DPRINTF(DKDB_DUMP, ("%s: no dumpblocks\n", __func__)); 787 return ENXIO; 788 } 789 790 /* device specific max transfer size */ 791 maxxfer = MAXPHYS; 792 if (dkd->d_iosize != NULL) 793 (*dkd->d_iosize)(dksc->sc_dev, &maxxfer); 794 795 /* Convert to disk sectors. Request must be a multiple of size. */ 796 part = DISKPART(dev); 797 lp = dksc->sc_dkdev.dk_label; 798 if ((size % lp->d_secsize) != 0) { 799 DPRINTF(DKDB_DUMP, ("%s: odd size %zu\n", __func__, size)); 800 return EFAULT; 801 } 802 towrt = size / lp->d_secsize; 803 blkno = dbtob(blkno) / lp->d_secsize; /* blkno in secsize units */ 804 805 p = &lp->d_partitions[part]; 806 if (p->p_fstype != FS_SWAP) { 807 DPRINTF(DKDB_DUMP, ("%s: bad fstype %d\n", __func__, 808 p->p_fstype)); 809 return ENXIO; 810 } 811 nsects = p->p_size; 812 sectoff = p->p_offset; 813 814 /* Check transfer bounds against partition size. */ 815 if ((blkno < 0) || ((blkno + towrt) > nsects)) { 816 DPRINTF(DKDB_DUMP, ("%s: out of bounds blkno=%jd, towrt=%d, " 817 "nsects=%d\n", __func__, (intmax_t)blkno, towrt, nsects)); 818 return EINVAL; 819 } 820 821 /* Offset block number to start of partition. */ 822 blkno += sectoff; 823 824 /* Start dumping and return when done. */ 825 maxblkcnt = howmany(maxxfer, lp->d_secsize); 826 while (towrt > 0) { 827 nblk = min(maxblkcnt, towrt); 828 829 if ((rv = (*dkd->d_dumpblocks)(dksc->sc_dev, va, blkno, nblk)) 830 != 0) { 831 DPRINTF(DKDB_DUMP, ("%s: dumpblocks %d\n", __func__, 832 rv)); 833 return rv; 834 } 835 836 towrt -= nblk; 837 blkno += nblk; 838 va += nblk * lp->d_secsize; 839 } 840 841 dk_dumping = 0; 842 843 return 0; 844 } 845 846 /* ARGSUSED */ 847 void 848 dk_getdefaultlabel(struct dk_softc *dksc, struct disklabel *lp) 849 { 850 const struct dkdriver *dkd = dksc->sc_dkdev.dk_driver; 851 struct disk_geom *dg = &dksc->sc_dkdev.dk_geom; 852 853 memset(lp, 0, sizeof(*lp)); 854 855 if (dg->dg_secperunit > UINT32_MAX) 856 lp->d_secperunit = UINT32_MAX; 857 else 858 lp->d_secperunit = dg->dg_secperunit; 859 lp->d_secsize = dg->dg_secsize; 860 lp->d_nsectors = dg->dg_nsectors; 861 lp->d_ntracks = dg->dg_ntracks; 862 lp->d_ncylinders = dg->dg_ncylinders; 863 lp->d_secpercyl = lp->d_ntracks * lp->d_nsectors; 864 865 strlcpy(lp->d_typename, dksc->sc_xname, sizeof(lp->d_typename)); 866 lp->d_type = dksc->sc_dtype; 867 strlcpy(lp->d_packname, "fictitious", sizeof(lp->d_packname)); 868 lp->d_rpm = 3600; 869 lp->d_interleave = 1; 870 lp->d_flags = 0; 871 872 lp->d_partitions[RAW_PART].p_offset = 0; 873 lp->d_partitions[RAW_PART].p_size = lp->d_secperunit; 874 lp->d_partitions[RAW_PART].p_fstype = FS_UNUSED; 875 lp->d_npartitions = RAW_PART + 1; 876 877 lp->d_magic = DISKMAGIC; 878 lp->d_magic2 = DISKMAGIC; 879 880 if (dkd->d_label) 881 dkd->d_label(dksc->sc_dev, lp); 882 883 lp->d_checksum = dkcksum(lp); 884 } 885 886 /* ARGSUSED */ 887 void 888 dk_getdisklabel(struct dk_softc *dksc, dev_t dev) 889 { 890 const struct dkdriver *dkd = dksc->sc_dkdev.dk_driver; 891 struct disklabel *lp = dksc->sc_dkdev.dk_label; 892 struct cpu_disklabel *clp = dksc->sc_dkdev.dk_cpulabel; 893 struct disk_geom *dg = &dksc->sc_dkdev.dk_geom; 894 struct partition *pp; 895 int i; 896 const char *errstring; 897 898 memset(clp, 0x0, sizeof(*clp)); 899 dk_getdefaultlabel(dksc, lp); 900 errstring = readdisklabel(DKLABELDEV(dev), dkd->d_strategy, 901 dksc->sc_dkdev.dk_label, dksc->sc_dkdev.dk_cpulabel); 902 if (errstring) { 903 dk_makedisklabel(dksc); 904 if (dksc->sc_flags & DKF_WARNLABEL) 905 printf("%s: %s\n", dksc->sc_xname, errstring); 906 return; 907 } 908 909 if ((dksc->sc_flags & DKF_LABELSANITY) == 0) 910 return; 911 912 /* Sanity check */ 913 if (lp->d_secperunit > dg->dg_secperunit) 914 printf("WARNING: %s: total sector size in disklabel (%ju) " 915 "!= the size of %s (%ju)\n", dksc->sc_xname, 916 (uintmax_t)lp->d_secperunit, dksc->sc_xname, 917 (uintmax_t)dg->dg_secperunit); 918 else if (lp->d_secperunit < UINT32_MAX && 919 lp->d_secperunit < dg->dg_secperunit) 920 printf("%s: %ju trailing sectors not covered by disklabel\n", 921 dksc->sc_xname, 922 (uintmax_t)dg->dg_secperunit - lp->d_secperunit); 923 924 for (i=0; i < lp->d_npartitions; i++) { 925 pp = &lp->d_partitions[i]; 926 if (pp->p_offset + pp->p_size > dg->dg_secperunit) 927 printf("WARNING: %s: end of partition `%c' exceeds " 928 "the size of %s (%ju)\n", dksc->sc_xname, 929 'a' + i, dksc->sc_xname, 930 (uintmax_t)dg->dg_secperunit); 931 } 932 } 933 934 /* 935 * Heuristic to conjure a disklabel if reading a disklabel failed. 936 * 937 * This is to allow the raw partition to be used for a filesystem 938 * without caring about the write protected label sector. 939 * 940 * If the driver provides it's own callback, use that instead. 941 */ 942 /* ARGSUSED */ 943 static void 944 dk_makedisklabel(struct dk_softc *dksc) 945 { 946 const struct dkdriver *dkd = dksc->sc_dkdev.dk_driver; 947 struct disklabel *lp = dksc->sc_dkdev.dk_label; 948 949 strlcpy(lp->d_packname, "default label", sizeof(lp->d_packname)); 950 951 if (dkd->d_label) 952 dkd->d_label(dksc->sc_dev, lp); 953 else 954 lp->d_partitions[RAW_PART].p_fstype = FS_BSDFFS; 955 956 lp->d_checksum = dkcksum(lp); 957 } 958 959 /* This function is taken from ccd.c:1.76 --rcd */ 960 961 /* 962 * XXX this function looks too generic for dksubr.c, shouldn't we 963 * put it somewhere better? 964 */ 965 966 /* 967 * Lookup the provided name in the filesystem. If the file exists, 968 * is a valid block device, and isn't being used by anyone else, 969 * set *vpp to the file's vnode. 970 */ 971 int 972 dk_lookup(struct pathbuf *pb, struct lwp *l, struct vnode **vpp) 973 { 974 struct nameidata nd; 975 struct vnode *vp; 976 int error; 977 978 if (l == NULL) 979 return ESRCH; /* Is ESRCH the best choice? */ 980 981 NDINIT(&nd, LOOKUP, FOLLOW, pb); 982 if ((error = vn_open(&nd, FREAD | FWRITE, 0)) != 0) { 983 DPRINTF((DKDB_FOLLOW|DKDB_INIT), 984 ("%s: vn_open error = %d\n", __func__, error)); 985 return error; 986 } 987 988 vp = nd.ni_vp; 989 if (vp->v_type != VBLK) { 990 error = ENOTBLK; 991 goto out; 992 } 993 994 /* Reopen as anonymous vnode to protect against forced unmount. */ 995 if ((error = bdevvp(vp->v_rdev, vpp)) != 0) 996 goto out; 997 VOP_UNLOCK(vp); 998 if ((error = vn_close(vp, FREAD | FWRITE, l->l_cred)) != 0) { 999 vrele(*vpp); 1000 return error; 1001 } 1002 if ((error = VOP_OPEN(*vpp, FREAD | FWRITE, l->l_cred)) != 0) { 1003 vrele(*vpp); 1004 return error; 1005 } 1006 mutex_enter((*vpp)->v_interlock); 1007 (*vpp)->v_writecount++; 1008 mutex_exit((*vpp)->v_interlock); 1009 1010 IFDEBUG(DKDB_VNODE, vprint("dk_lookup: vnode info", *vpp)); 1011 1012 return 0; 1013 out: 1014 VOP_UNLOCK(vp); 1015 (void) vn_close(vp, FREAD | FWRITE, l->l_cred); 1016 return error; 1017 } 1018 1019 MODULE(MODULE_CLASS_MISC, dk_subr, NULL); 1020 1021 static int 1022 dk_subr_modcmd(modcmd_t cmd, void *arg) 1023 { 1024 switch (cmd) { 1025 case MODULE_CMD_INIT: 1026 case MODULE_CMD_FINI: 1027 return 0; 1028 case MODULE_CMD_STAT: 1029 case MODULE_CMD_AUTOUNLOAD: 1030 default: 1031 return ENOTTY; 1032 } 1033 } 1034