subr_disk.c revision 1.100.18.5
1 /* $NetBSD: subr_disk.c,v 1.100.18.5 2013/06/23 06:18:58 tls Exp $ */ 2 3 /*- 4 * Copyright (c) 1996, 1997, 1999, 2000, 2009 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 of the Numerical Aerospace Simulation Facility, 9 * NASA Ames Research Center. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 22 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 23 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 24 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 25 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 26 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 27 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 28 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 29 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 30 * POSSIBILITY OF SUCH DAMAGE. 31 */ 32 33 /* 34 * Copyright (c) 1982, 1986, 1988, 1993 35 * The Regents of the University of California. All rights reserved. 36 * (c) UNIX System Laboratories, Inc. 37 * All or some portions of this file are derived from material licensed 38 * to the University of California by American Telephone and Telegraph 39 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 40 * the permission of UNIX System Laboratories, Inc. 41 * 42 * Redistribution and use in source and binary forms, with or without 43 * modification, are permitted provided that the following conditions 44 * are met: 45 * 1. Redistributions of source code must retain the above copyright 46 * notice, this list of conditions and the following disclaimer. 47 * 2. Redistributions in binary form must reproduce the above copyright 48 * notice, this list of conditions and the following disclaimer in the 49 * documentation and/or other materials provided with the distribution. 50 * 3. Neither the name of the University nor the names of its contributors 51 * may be used to endorse or promote products derived from this software 52 * without specific prior written permission. 53 * 54 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 55 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 56 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 57 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 58 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 59 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 60 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 61 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 62 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 63 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 64 * SUCH DAMAGE. 65 * 66 * @(#)ufs_disksubr.c 8.5 (Berkeley) 1/21/94 67 */ 68 69 #include <sys/cdefs.h> 70 __KERNEL_RCSID(0, "$NetBSD: subr_disk.c,v 1.100.18.5 2013/06/23 06:18:58 tls Exp $"); 71 72 #include <sys/param.h> 73 #include <sys/kernel.h> 74 #include <sys/kmem.h> 75 #include <sys/buf.h> 76 #include <sys/syslog.h> 77 #include <sys/disklabel.h> 78 #include <sys/conf.h> 79 #include <sys/disk.h> 80 #include <sys/sysctl.h> 81 #include <lib/libkern/libkern.h> 82 83 unsigned int disk_serial; 84 85 /* 86 * Compute checksum for disk label. 87 */ 88 u_int 89 dkcksum(struct disklabel *lp) 90 { 91 92 return dkcksum_sized(lp, lp->d_npartitions); 93 } 94 95 u_int 96 dkcksum_sized(struct disklabel *lp, size_t npartitions) 97 { 98 uint16_t *start, *end; 99 uint16_t sum = 0; 100 101 start = (uint16_t *)lp; 102 end = (uint16_t *)&lp->d_partitions[npartitions]; 103 while (start < end) 104 sum ^= *start++; 105 return sum; 106 } 107 108 /* 109 * Disk error is the preface to plaintive error messages 110 * about failing disk transfers. It prints messages of the form 111 112 hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d) 113 114 * if the offset of the error in the transfer and a disk label 115 * are both available. blkdone should be -1 if the position of the error 116 * is unknown; the disklabel pointer may be null from drivers that have not 117 * been converted to use them. The message is printed with printf 118 * if pri is LOG_PRINTF, otherwise it uses log at the specified priority. 119 * The message should be completed (with at least a newline) with printf 120 * or addlog, respectively. There is no trailing space. 121 */ 122 #ifndef PRIdaddr 123 #define PRIdaddr PRId64 124 #endif 125 void 126 diskerr(const struct buf *bp, const char *dname, const char *what, int pri, 127 int blkdone, const struct disklabel *lp) 128 { 129 int unit = DISKUNIT(bp->b_dev), part = DISKPART(bp->b_dev); 130 void (*pr)(const char *, ...) __printflike(1, 2); 131 char partname = 'a' + part; 132 daddr_t sn; 133 134 if (/*CONSTCOND*/0) 135 /* Compiler will error this is the format is wrong... */ 136 printf("%" PRIdaddr, bp->b_blkno); 137 138 if (pri != LOG_PRINTF) { 139 static const char fmt[] = ""; 140 log(pri, fmt); 141 pr = addlog; 142 } else 143 pr = printf; 144 (*pr)("%s%d%c: %s %sing fsbn ", dname, unit, partname, what, 145 bp->b_flags & B_READ ? "read" : "writ"); 146 sn = bp->b_blkno; 147 if (bp->b_bcount <= DEV_BSIZE) 148 (*pr)("%" PRIdaddr, sn); 149 else { 150 if (blkdone >= 0) { 151 sn += blkdone; 152 (*pr)("%" PRIdaddr " of ", sn); 153 } 154 (*pr)("%" PRIdaddr "-%" PRIdaddr "", bp->b_blkno, 155 bp->b_blkno + (bp->b_bcount - 1) / DEV_BSIZE); 156 } 157 if (lp && (blkdone >= 0 || bp->b_bcount <= lp->d_secsize)) { 158 sn += lp->d_partitions[part].p_offset; 159 (*pr)(" (%s%d bn %" PRIdaddr "; cn %" PRIdaddr "", 160 dname, unit, sn, sn / lp->d_secpercyl); 161 sn %= lp->d_secpercyl; 162 (*pr)(" tn %" PRIdaddr " sn %" PRIdaddr ")", 163 sn / lp->d_nsectors, sn % lp->d_nsectors); 164 } 165 } 166 167 /* 168 * Searches the iostatlist for the disk corresponding to the 169 * name provided. 170 */ 171 struct disk * 172 disk_find(const char *name) 173 { 174 struct io_stats *stat; 175 176 stat = iostat_find(name); 177 178 if ((stat != NULL) && (stat->io_type == IOSTAT_DISK)) 179 return stat->io_parent; 180 181 return (NULL); 182 } 183 184 /* 185 * Searches for the disk corresponding to a supplied block device 186 * using major, minor, unit. 187 */ 188 struct disk * 189 disk_find_blk(dev_t dev) 190 { 191 devmajor_t major; 192 int unit; 193 char name[16]; /* XXX */ 194 const char *swname; 195 196 major = major(dev); 197 unit = DISKUNIT(dev); 198 199 if ((swname = devsw_blk2name(major)) == NULL) { 200 return NULL; 201 } 202 203 if (snprintf(name, sizeof(name), "%s%d", 204 swname, unit) > sizeof(name)) { 205 return NULL; 206 } 207 208 return disk_find(name); 209 } 210 211 int disk_maxphys(const struct disk *const diskp) 212 { 213 struct buf b = { b_bcount: MACHINE_MAXPHYS }; 214 215 diskp->dk_driver->d_minphys(&b); 216 217 return b.b_bcount; 218 } 219 220 void 221 disk_init(struct disk *diskp, const char *name, const struct dkdriver *driver) 222 { 223 KASSERT(diskp != NULL); 224 KASSERT(name != NULL); 225 KASSERT(driver != NULL); 226 227 /* 228 * Initialize the wedge-related locks and other fields. 229 */ 230 mutex_init(&diskp->dk_rawlock, MUTEX_DEFAULT, IPL_NONE); 231 mutex_init(&diskp->dk_openlock, MUTEX_DEFAULT, IPL_NONE); 232 LIST_INIT(&diskp->dk_wedges); 233 diskp->dk_nwedges = 0; 234 diskp->dk_labelsector = LABELSECTOR; 235 disk_blocksize(diskp, DEV_BSIZE); 236 diskp->dk_name = name; 237 diskp->dk_driver = driver; 238 } 239 240 /* 241 * Attach a disk. 242 */ 243 void 244 disk_attach(struct disk *diskp) 245 { 246 247 /* 248 * Allocate and initialize the disklabel structures. 249 */ 250 diskp->dk_label = kmem_zalloc(sizeof(struct disklabel), KM_SLEEP); 251 diskp->dk_cpulabel = kmem_zalloc(sizeof(struct cpu_disklabel), 252 KM_SLEEP); 253 if ((diskp->dk_label == NULL) || (diskp->dk_cpulabel == NULL)) 254 panic("disk_attach: can't allocate storage for disklabel"); 255 256 /* 257 * Set up the stats collection. 258 */ 259 diskp->dk_stats = iostat_alloc(IOSTAT_DISK, diskp, diskp->dk_name); 260 } 261 262 int 263 disk_begindetach(struct disk *dk, int (*lastclose)(device_t), 264 device_t self, int flags) 265 { 266 int rc; 267 268 rc = 0; 269 mutex_enter(&dk->dk_openlock); 270 if (dk->dk_openmask == 0) 271 ; /* nothing to do */ 272 else if ((flags & DETACH_FORCE) == 0) 273 rc = EBUSY; 274 else if (lastclose != NULL) 275 rc = (*lastclose)(self); 276 mutex_exit(&dk->dk_openlock); 277 278 return rc; 279 } 280 281 /* 282 * Detach a disk. 283 */ 284 void 285 disk_detach(struct disk *diskp) 286 { 287 288 /* 289 * Remove from the drivelist. 290 */ 291 iostat_free(diskp->dk_stats); 292 293 /* 294 * Release the disk-info dictionary. 295 */ 296 if (diskp->dk_info) { 297 prop_object_release(diskp->dk_info); 298 diskp->dk_info = NULL; 299 } 300 301 /* 302 * Free the space used by the disklabel structures. 303 */ 304 kmem_free(diskp->dk_label, sizeof(*diskp->dk_label)); 305 kmem_free(diskp->dk_cpulabel, sizeof(*diskp->dk_cpulabel)); 306 } 307 308 void 309 disk_destroy(struct disk *diskp) 310 { 311 312 mutex_destroy(&diskp->dk_openlock); 313 mutex_destroy(&diskp->dk_rawlock); 314 } 315 316 /* 317 * Mark the disk as busy for metrics collection. 318 */ 319 void 320 disk_busy(struct disk *diskp) 321 { 322 323 iostat_busy(diskp->dk_stats); 324 } 325 326 /* 327 * Finished disk operations, gather metrics. 328 */ 329 void 330 disk_unbusy(struct disk *diskp, long bcount, int read) 331 { 332 333 iostat_unbusy(diskp->dk_stats, bcount, read); 334 } 335 336 /* 337 * Return true if disk has an I/O operation in flight. 338 */ 339 bool 340 disk_isbusy(struct disk *diskp) 341 { 342 343 return iostat_isbusy(diskp->dk_stats); 344 } 345 346 /* 347 * Set the physical blocksize of a disk, in bytes. 348 * Only necessary if blocksize != DEV_BSIZE. 349 */ 350 void 351 disk_blocksize(struct disk *diskp, int blocksize) 352 { 353 354 diskp->dk_blkshift = DK_BSIZE2BLKSHIFT(blocksize); 355 diskp->dk_byteshift = DK_BSIZE2BYTESHIFT(blocksize); 356 } 357 358 /* 359 * Bounds checking against the media size, used for the raw partition. 360 * secsize, mediasize and b_blkno must all be the same units. 361 * Possibly this has to be DEV_BSIZE (512). 362 */ 363 int 364 bounds_check_with_mediasize(struct buf *bp, int secsize, uint64_t mediasize) 365 { 366 int64_t sz; 367 368 sz = howmany(bp->b_bcount, secsize); 369 370 if (bp->b_blkno + sz > mediasize) { 371 sz = mediasize - bp->b_blkno; 372 if (sz == 0) { 373 /* If exactly at end of disk, return EOF. */ 374 bp->b_resid = bp->b_bcount; 375 return 0; 376 } 377 if (sz < 0) { 378 /* If past end of disk, return EINVAL. */ 379 bp->b_error = EINVAL; 380 return 0; 381 } 382 /* Otherwise, truncate request. */ 383 bp->b_bcount = sz * secsize; 384 } 385 386 return 1; 387 } 388 389 /* 390 * Determine the size of the transfer, and make sure it is 391 * within the boundaries of the partition. Adjust transfer 392 * if needed, and signal errors or early completion. 393 */ 394 int 395 bounds_check_with_label(struct disk *dk, struct buf *bp, int wlabel) 396 { 397 struct disklabel *lp = dk->dk_label; 398 struct partition *p = lp->d_partitions + DISKPART(bp->b_dev); 399 uint64_t p_size, p_offset, labelsector; 400 int64_t sz; 401 402 /* Protect against division by zero. XXX: Should never happen?!?! */ 403 if (lp->d_secpercyl == 0) { 404 bp->b_error = EINVAL; 405 return -1; 406 } 407 408 p_size = (uint64_t)p->p_size << dk->dk_blkshift; 409 p_offset = (uint64_t)p->p_offset << dk->dk_blkshift; 410 #if RAW_PART == 3 411 labelsector = lp->d_partitions[2].p_offset; 412 #else 413 labelsector = lp->d_partitions[RAW_PART].p_offset; 414 #endif 415 labelsector = (labelsector + dk->dk_labelsector) << dk->dk_blkshift; 416 417 sz = howmany(bp->b_bcount, DEV_BSIZE); 418 if ((bp->b_blkno + sz) > p_size) { 419 sz = p_size - bp->b_blkno; 420 if (sz == 0) { 421 /* If exactly at end of disk, return EOF. */ 422 bp->b_resid = bp->b_bcount; 423 return 0; 424 } 425 if (sz < 0) { 426 /* If past end of disk, return EINVAL. */ 427 bp->b_error = EINVAL; 428 return -1; 429 } 430 /* Otherwise, truncate request. */ 431 bp->b_bcount = sz << DEV_BSHIFT; 432 } 433 434 /* Overwriting disk label? */ 435 if (bp->b_blkno + p_offset <= labelsector && 436 bp->b_blkno + p_offset + sz > labelsector && 437 (bp->b_flags & B_READ) == 0 && !wlabel) { 438 bp->b_error = EROFS; 439 return -1; 440 } 441 442 /* calculate cylinder for disksort to order transfers with */ 443 bp->b_cylinder = (bp->b_blkno + p->p_offset) / 444 (lp->d_secsize / DEV_BSIZE) / lp->d_secpercyl; 445 return 1; 446 } 447 448 int 449 disk_read_sectors(void (*strat)(struct buf *), const struct disklabel *lp, 450 struct buf *bp, unsigned int sector, int count) 451 { 452 bp->b_blkno = sector; 453 bp->b_bcount = count * lp->d_secsize; 454 bp->b_flags = (bp->b_flags & ~B_WRITE) | B_READ; 455 bp->b_oflags &= ~BO_DONE; 456 bp->b_cylinder = sector / lp->d_secpercyl; 457 (*strat)(bp); 458 return biowait(bp); 459 } 460 461 const char * 462 convertdisklabel(struct disklabel *lp, void (*strat)(struct buf *), 463 struct buf *bp, uint32_t secperunit) 464 { 465 struct partition rp, *altp, *p; 466 int geom_ok; 467 468 memset(&rp, 0, sizeof(rp)); 469 rp.p_size = secperunit; 470 rp.p_fstype = FS_UNUSED; 471 472 /* If we can seek to d_secperunit - 1, believe the disk geometry. */ 473 if (secperunit != 0 && 474 disk_read_sectors(strat, lp, bp, secperunit - 1, 1) == 0) 475 geom_ok = 1; 476 else 477 geom_ok = 0; 478 479 #if 0 480 printf("%s: secperunit (%" PRIu32 ") %s\n", __func__, 481 secperunit, geom_ok ? "ok" : "not ok"); 482 #endif 483 484 p = &lp->d_partitions[RAW_PART]; 485 if (RAW_PART == 'c' - 'a') 486 altp = &lp->d_partitions['d' - 'a']; 487 else 488 altp = &lp->d_partitions['c' - 'a']; 489 490 if (lp->d_npartitions > RAW_PART && p->p_offset == 0 && p->p_size != 0) 491 ; /* already a raw partition */ 492 else if (lp->d_npartitions > MAX('c', 'd') - 'a' && 493 altp->p_offset == 0 && altp->p_size != 0) { 494 /* alternate partition ('c' or 'd') is suitable for raw slot, 495 * swap with 'd' or 'c'. 496 */ 497 rp = *p; 498 *p = *altp; 499 *altp = rp; 500 } else if (lp->d_npartitions <= RAW_PART && 501 lp->d_npartitions > 'c' - 'a') { 502 /* No raw partition is present, but the alternate is present. 503 * Copy alternate to raw partition. 504 */ 505 lp->d_npartitions = RAW_PART + 1; 506 *p = *altp; 507 } else if (!geom_ok) 508 return "no raw partition and disk reports bad geometry"; 509 else if (lp->d_npartitions <= RAW_PART) { 510 memset(&lp->d_partitions[lp->d_npartitions], 0, 511 sizeof(struct partition) * (RAW_PART - lp->d_npartitions)); 512 *p = rp; 513 lp->d_npartitions = RAW_PART + 1; 514 } else if (lp->d_npartitions < MAXPARTITIONS) { 515 memmove(p + 1, p, 516 sizeof(struct partition) * (lp->d_npartitions - RAW_PART)); 517 *p = rp; 518 lp->d_npartitions++; 519 } else 520 return "no raw partition and partition table is full"; 521 return NULL; 522 } 523 524 /* 525 * disk_ioctl -- 526 * Generic disk ioctl handling. 527 */ 528 int 529 disk_ioctl(struct disk *diskp, u_long cmd, void *data, int flag, 530 struct lwp *l) 531 { 532 int error; 533 534 switch (cmd) { 535 case DIOCGDISKINFO: 536 { 537 struct plistref *pref = (struct plistref *) data; 538 539 if (diskp->dk_info == NULL) 540 error = ENOTSUP; 541 else 542 error = prop_dictionary_copyout_ioctl(pref, cmd, 543 diskp->dk_info); 544 break; 545 } 546 547 default: 548 error = EPASSTHROUGH; 549 } 550 551 return (error); 552 } 553 554 void 555 disk_set_info(device_t dev, struct disk *dk, const char *type) 556 { 557 struct disk_geom *dg = &dk->dk_geom; 558 559 if (dg->dg_secperunit == 0 && dg->dg_ncylinders == 0) { 560 #ifdef DIAGNOSTIC 561 printf("%s: secperunit and ncylinders are zero\n", dk->dk_name); 562 #endif 563 return; 564 } 565 566 if (dg->dg_secperunit == 0) { 567 if (dg->dg_nsectors == 0 || dg->dg_ntracks == 0) { 568 #ifdef DIAGNOSTIC 569 printf("%s: secperunit and (sectors or tracks) " 570 "are zero\n", dk->dk_name); 571 #endif 572 return; 573 } 574 dg->dg_secperunit = dg->dg_nsectors * 575 dg->dg_ntracks * dg->dg_ncylinders; 576 } 577 578 if (dg->dg_ncylinders == 0) { 579 if (dg->dg_ntracks && dg->dg_nsectors) 580 dg->dg_ncylinders = dg->dg_secperunit / 581 (dg->dg_ntracks * dg->dg_nsectors); 582 } 583 584 if (dg->dg_secsize == 0) { 585 #ifdef DIAGNOSTIC 586 printf("%s: fixing 0 sector size\n", dk->dk_name); 587 #endif 588 dg->dg_secsize = DEV_BSIZE; 589 } 590 591 prop_dictionary_t disk_info, odisk_info, geom; 592 593 disk_info = prop_dictionary_create(); 594 geom = prop_dictionary_create(); 595 596 prop_dictionary_set_uint64(geom, "sectors-per-unit", 597 dg->dg_secperunit); 598 599 prop_dictionary_set_uint32(geom, "sector-size", dg->dg_secsize); 600 601 if (dg->dg_nsectors) 602 prop_dictionary_set_uint16(geom, "sectors-per-track", 603 dg->dg_nsectors); 604 605 if (dg->dg_ntracks) 606 prop_dictionary_set_uint16(geom, "tracks-per-cylinder", 607 dg->dg_ntracks); 608 609 if (dg->dg_ncylinders) 610 prop_dictionary_set_uint64(geom, "cylinders-per-unit", 611 dg->dg_ncylinders); 612 613 prop_dictionary_set(disk_info, "geometry", geom); 614 615 if (type) 616 prop_dictionary_set_cstring_nocopy(disk_info, "type", type); 617 618 prop_object_release(geom); 619 620 odisk_info = dk->dk_info; 621 dk->dk_info = disk_info; 622 623 if (dev) 624 prop_dictionary_set(device_properties(dev), "disk-info", 625 disk_info); 626 627 /* 628 * Don't release disk_info here; we keep a reference to it. 629 * disk_detach() will release it when we go away. 630 */ 631 if (odisk_info) 632 prop_object_release(odisk_info); 633 } 634
Indexes created Thu Sep 25 16:09:42 GMT 2025