1 /* $NetBSD: subr_disk.c,v 1.100 2010/10/14 00:47:16 mrg 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 2010/10/14 00:47:16 mrg 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/disk.h> 79 #include <sys/sysctl.h> 80 #include <lib/libkern/libkern.h> 81 82 /* 83 * Compute checksum for disk label. 84 */ 85 u_int 86 dkcksum(struct disklabel *lp) 87 { 88 89 return dkcksum_sized(lp, lp->d_npartitions); 90 } 91 92 u_int 93 dkcksum_sized(struct disklabel *lp, size_t npartitions) 94 { 95 uint16_t *start, *end; 96 uint16_t sum = 0; 97 98 start = (uint16_t *)lp; 99 end = (uint16_t *)&lp->d_partitions[npartitions]; 100 while (start < end) 101 sum ^= *start++; 102 return sum; 103 } 104 105 /* 106 * Disk error is the preface to plaintive error messages 107 * about failing disk transfers. It prints messages of the form 108 109 hp0g: hard error reading fsbn 12345 of 12344-12347 (hp0 bn %d cn %d tn %d sn %d) 110 111 * if the offset of the error in the transfer and a disk label 112 * are both available. blkdone should be -1 if the position of the error 113 * is unknown; the disklabel pointer may be null from drivers that have not 114 * been converted to use them. The message is printed with printf 115 * if pri is LOG_PRINTF, otherwise it uses log at the specified priority. 116 * The message should be completed (with at least a newline) with printf 117 * or addlog, respectively. There is no trailing space. 118 */ 119 #ifndef PRIdaddr 120 #define PRIdaddr PRId64 121 #endif 122 void 123 diskerr(const struct buf *bp, const char *dname, const char *what, int pri, 124 int blkdone, const struct disklabel *lp) 125 { 126 int unit = DISKUNIT(bp->b_dev), part = DISKPART(bp->b_dev); 127 void (*pr)(const char *, ...); 128 char partname = 'a' + part; 129 daddr_t sn; 130 131 if (/*CONSTCOND*/0) 132 /* Compiler will error this is the format is wrong... */ 133 printf("%" PRIdaddr, bp->b_blkno); 134 135 if (pri != LOG_PRINTF) { 136 static const char fmt[] = ""; 137 log(pri, fmt); 138 pr = addlog; 139 } else 140 pr = printf; 141 (*pr)("%s%d%c: %s %sing fsbn ", dname, unit, partname, what, 142 bp->b_flags & B_READ ? "read" : "writ"); 143 sn = bp->b_blkno; 144 if (bp->b_bcount <= DEV_BSIZE) 145 (*pr)("%" PRIdaddr, sn); 146 else { 147 if (blkdone >= 0) { 148 sn += blkdone; 149 (*pr)("%" PRIdaddr " of ", sn); 150 } 151 (*pr)("%" PRIdaddr "-%" PRIdaddr "", bp->b_blkno, 152 bp->b_blkno + (bp->b_bcount - 1) / DEV_BSIZE); 153 } 154 if (lp && (blkdone >= 0 || bp->b_bcount <= lp->d_secsize)) { 155 sn += lp->d_partitions[part].p_offset; 156 (*pr)(" (%s%d bn %" PRIdaddr "; cn %" PRIdaddr "", 157 dname, unit, sn, sn / lp->d_secpercyl); 158 sn %= lp->d_secpercyl; 159 (*pr)(" tn %" PRIdaddr " sn %" PRIdaddr ")", 160 sn / lp->d_nsectors, sn % lp->d_nsectors); 161 } 162 } 163 164 /* 165 * Searches the iostatlist for the disk corresponding to the 166 * name provided. 167 */ 168 struct disk * 169 disk_find(const char *name) 170 { 171 struct io_stats *stat; 172 173 stat = iostat_find(name); 174 175 if ((stat != NULL) && (stat->io_type == IOSTAT_DISK)) 176 return stat->io_parent; 177 178 return (NULL); 179 } 180 181 void 182 disk_init(struct disk *diskp, const char *name, const struct dkdriver *driver) 183 { 184 185 /* 186 * Initialize the wedge-related locks and other fields. 187 */ 188 mutex_init(&diskp->dk_rawlock, MUTEX_DEFAULT, IPL_NONE); 189 mutex_init(&diskp->dk_openlock, MUTEX_DEFAULT, IPL_NONE); 190 LIST_INIT(&diskp->dk_wedges); 191 diskp->dk_nwedges = 0; 192 diskp->dk_labelsector = LABELSECTOR; 193 disk_blocksize(diskp, DEV_BSIZE); 194 diskp->dk_name = name; 195 diskp->dk_driver = driver; 196 } 197 198 /* 199 * Attach a disk. 200 */ 201 void 202 disk_attach(struct disk *diskp) 203 { 204 205 /* 206 * Allocate and initialize the disklabel structures. 207 */ 208 diskp->dk_label = kmem_zalloc(sizeof(struct disklabel), KM_SLEEP); 209 diskp->dk_cpulabel = kmem_zalloc(sizeof(struct cpu_disklabel), 210 KM_SLEEP); 211 if ((diskp->dk_label == NULL) || (diskp->dk_cpulabel == NULL)) 212 panic("disk_attach: can't allocate storage for disklabel"); 213 214 /* 215 * Set up the stats collection. 216 */ 217 diskp->dk_stats = iostat_alloc(IOSTAT_DISK, diskp, diskp->dk_name); 218 } 219 220 int 221 disk_begindetach(struct disk *dk, int (*lastclose)(device_t), 222 device_t self, int flags) 223 { 224 int rc; 225 226 rc = 0; 227 mutex_enter(&dk->dk_openlock); 228 if (dk->dk_openmask == 0) 229 ; /* nothing to do */ 230 else if ((flags & DETACH_FORCE) == 0) 231 rc = EBUSY; 232 else if (lastclose != NULL) 233 rc = (*lastclose)(self); 234 mutex_exit(&dk->dk_openlock); 235 236 return rc; 237 } 238 239 /* 240 * Detach a disk. 241 */ 242 void 243 disk_detach(struct disk *diskp) 244 { 245 246 /* 247 * Remove from the drivelist. 248 */ 249 iostat_free(diskp->dk_stats); 250 251 /* 252 * Release the disk-info dictionary. 253 */ 254 if (diskp->dk_info) { 255 prop_object_release(diskp->dk_info); 256 diskp->dk_info = NULL; 257 } 258 259 /* 260 * Free the space used by the disklabel structures. 261 */ 262 kmem_free(diskp->dk_label, sizeof(*diskp->dk_label)); 263 kmem_free(diskp->dk_cpulabel, sizeof(*diskp->dk_cpulabel)); 264 } 265 266 void 267 disk_destroy(struct disk *diskp) 268 { 269 270 mutex_destroy(&diskp->dk_openlock); 271 mutex_destroy(&diskp->dk_rawlock); 272 } 273 274 /* 275 * Mark the disk as busy for metrics collection. 276 */ 277 void 278 disk_busy(struct disk *diskp) 279 { 280 281 iostat_busy(diskp->dk_stats); 282 } 283 284 /* 285 * Finished disk operations, gather metrics. 286 */ 287 void 288 disk_unbusy(struct disk *diskp, long bcount, int read) 289 { 290 291 iostat_unbusy(diskp->dk_stats, bcount, read); 292 } 293 294 /* 295 * Return true if disk has an I/O operation in flight. 296 */ 297 bool 298 disk_isbusy(struct disk *diskp) 299 { 300 301 return iostat_isbusy(diskp->dk_stats); 302 } 303 304 /* 305 * Set the physical blocksize of a disk, in bytes. 306 * Only necessary if blocksize != DEV_BSIZE. 307 */ 308 void 309 disk_blocksize(struct disk *diskp, int blocksize) 310 { 311 312 diskp->dk_blkshift = DK_BSIZE2BLKSHIFT(blocksize); 313 diskp->dk_byteshift = DK_BSIZE2BYTESHIFT(blocksize); 314 } 315 316 /* 317 * Bounds checking against the media size, used for the raw partition. 318 * secsize, mediasize and b_blkno must all be the same units. 319 * Possibly this has to be DEV_BSIZE (512). 320 */ 321 int 322 bounds_check_with_mediasize(struct buf *bp, int secsize, uint64_t mediasize) 323 { 324 int64_t sz; 325 326 sz = howmany(bp->b_bcount, secsize); 327 328 if (bp->b_blkno + sz > mediasize) { 329 sz = mediasize - bp->b_blkno; 330 if (sz == 0) { 331 /* If exactly at end of disk, return EOF. */ 332 bp->b_resid = bp->b_bcount; 333 return 0; 334 } 335 if (sz < 0) { 336 /* If past end of disk, return EINVAL. */ 337 bp->b_error = EINVAL; 338 return 0; 339 } 340 /* Otherwise, truncate request. */ 341 bp->b_bcount = sz * secsize; 342 } 343 344 return 1; 345 } 346 347 /* 348 * Determine the size of the transfer, and make sure it is 349 * within the boundaries of the partition. Adjust transfer 350 * if needed, and signal errors or early completion. 351 */ 352 int 353 bounds_check_with_label(struct disk *dk, struct buf *bp, int wlabel) 354 { 355 struct disklabel *lp = dk->dk_label; 356 struct partition *p = lp->d_partitions + DISKPART(bp->b_dev); 357 uint64_t p_size, p_offset, labelsector; 358 int64_t sz; 359 360 /* Protect against division by zero. XXX: Should never happen?!?! */ 361 if (lp->d_secpercyl == 0) { 362 bp->b_error = EINVAL; 363 return -1; 364 } 365 366 p_size = (uint64_t)p->p_size << dk->dk_blkshift; 367 p_offset = (uint64_t)p->p_offset << dk->dk_blkshift; 368 #if RAW_PART == 3 369 labelsector = lp->d_partitions[2].p_offset; 370 #else 371 labelsector = lp->d_partitions[RAW_PART].p_offset; 372 #endif 373 labelsector = (labelsector + dk->dk_labelsector) << dk->dk_blkshift; 374 375 sz = howmany(bp->b_bcount, DEV_BSIZE); 376 if ((bp->b_blkno + sz) > p_size) { 377 sz = p_size - bp->b_blkno; 378 if (sz == 0) { 379 /* If exactly at end of disk, return EOF. */ 380 bp->b_resid = bp->b_bcount; 381 return 0; 382 } 383 if (sz < 0) { 384 /* If past end of disk, return EINVAL. */ 385 bp->b_error = EINVAL; 386 return -1; 387 } 388 /* Otherwise, truncate request. */ 389 bp->b_bcount = sz << DEV_BSHIFT; 390 } 391 392 /* Overwriting disk label? */ 393 if (bp->b_blkno + p_offset <= labelsector && 394 bp->b_blkno + p_offset + sz > labelsector && 395 (bp->b_flags & B_READ) == 0 && !wlabel) { 396 bp->b_error = EROFS; 397 return -1; 398 } 399 400 /* calculate cylinder for disksort to order transfers with */ 401 bp->b_cylinder = (bp->b_blkno + p->p_offset) / 402 (lp->d_secsize / DEV_BSIZE) / lp->d_secpercyl; 403 return 1; 404 } 405 406 int 407 disk_read_sectors(void (*strat)(struct buf *), const struct disklabel *lp, 408 struct buf *bp, unsigned int sector, int count) 409 { 410 bp->b_blkno = sector; 411 bp->b_bcount = count * lp->d_secsize; 412 bp->b_flags = (bp->b_flags & ~B_WRITE) | B_READ; 413 bp->b_oflags &= ~BO_DONE; 414 bp->b_cylinder = sector / lp->d_secpercyl; 415 (*strat)(bp); 416 return biowait(bp); 417 } 418 419 const char * 420 convertdisklabel(struct disklabel *lp, void (*strat)(struct buf *), 421 struct buf *bp, uint32_t secperunit) 422 { 423 struct partition rp, *altp, *p; 424 int geom_ok; 425 426 memset(&rp, 0, sizeof(rp)); 427 rp.p_size = secperunit; 428 rp.p_fstype = FS_UNUSED; 429 430 /* If we can seek to d_secperunit - 1, believe the disk geometry. */ 431 if (secperunit != 0 && 432 disk_read_sectors(strat, lp, bp, secperunit - 1, 1) == 0) 433 geom_ok = 1; 434 else 435 geom_ok = 0; 436 437 #if 0 438 printf("%s: secperunit (%" PRIu32 ") %s\n", __func__, 439 secperunit, geom_ok ? "ok" : "not ok"); 440 #endif 441 442 p = &lp->d_partitions[RAW_PART]; 443 if (RAW_PART == 'c' - 'a') 444 altp = &lp->d_partitions['d' - 'a']; 445 else 446 altp = &lp->d_partitions['c' - 'a']; 447 448 if (lp->d_npartitions > RAW_PART && p->p_offset == 0 && p->p_size != 0) 449 ; /* already a raw partition */ 450 else if (lp->d_npartitions > MAX('c', 'd') - 'a' && 451 altp->p_offset == 0 && altp->p_size != 0) { 452 /* alternate partition ('c' or 'd') is suitable for raw slot, 453 * swap with 'd' or 'c'. 454 */ 455 rp = *p; 456 *p = *altp; 457 *altp = rp; 458 } else if (lp->d_npartitions <= RAW_PART && 459 lp->d_npartitions > 'c' - 'a') { 460 /* No raw partition is present, but the alternate is present. 461 * Copy alternate to raw partition. 462 */ 463 lp->d_npartitions = RAW_PART + 1; 464 *p = *altp; 465 } else if (!geom_ok) 466 return "no raw partition and disk reports bad geometry"; 467 else if (lp->d_npartitions <= RAW_PART) { 468 memset(&lp->d_partitions[lp->d_npartitions], 0, 469 sizeof(struct partition) * (RAW_PART - lp->d_npartitions)); 470 *p = rp; 471 lp->d_npartitions = RAW_PART + 1; 472 } else if (lp->d_npartitions < MAXPARTITIONS) { 473 memmove(p + 1, p, 474 sizeof(struct partition) * (lp->d_npartitions - RAW_PART)); 475 *p = rp; 476 lp->d_npartitions++; 477 } else 478 return "no raw partition and partition table is full"; 479 return NULL; 480 } 481 482 /* 483 * disk_ioctl -- 484 * Generic disk ioctl handling. 485 */ 486 int 487 disk_ioctl(struct disk *diskp, u_long cmd, void *data, int flag, 488 struct lwp *l) 489 { 490 int error; 491 492 switch (cmd) { 493 case DIOCGDISKINFO: 494 { 495 struct plistref *pref = (struct plistref *) data; 496 497 if (diskp->dk_info == NULL) 498 error = ENOTSUP; 499 else 500 error = prop_dictionary_copyout_ioctl(pref, cmd, 501 diskp->dk_info); 502 break; 503 } 504 505 default: 506 error = EPASSTHROUGH; 507 } 508 509 return (error); 510 } 511