1 /* 2 * Copyright (c) 2007 The DragonFly Project. All rights reserved. 3 * 4 * This code is derived from software contributed to The DragonFly Project 5 * by Matthew Dillon <dillon@backplane.com> 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 * 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in 15 * the documentation and/or other materials provided with the 16 * distribution. 17 * 3. Neither the name of The DragonFly Project nor the names of its 18 * contributors may be used to endorse or promote products derived 19 * from this software without specific, prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 22 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS 24 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE 25 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, 26 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING, 27 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; 28 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED 29 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 30 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT 31 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32 * SUCH DAMAGE. 33 * 34 * $DragonFly: src/sys/kern/subr_disklabel64.c,v 1.5 2007/07/20 17:21:51 dillon Exp $ 35 */ 36 37 #include <sys/param.h> 38 #include <sys/systm.h> 39 #include <sys/kernel.h> 40 #include <sys/conf.h> 41 #include <sys/disklabel.h> 42 #include <sys/disklabel64.h> 43 #include <sys/diskslice.h> 44 #include <sys/disk.h> 45 #include <sys/kern_syscall.h> 46 #include <sys/buf2.h> 47 48 /* 49 * Alignment against physical start (verses slice start). We use a megabyte 50 * here. Why do we use a megabyte? Because SSDs already use large 128K 51 * blocks internally (for MLC) and who the hell knows in the future. 52 * 53 * This way if the sysop picks sane values for partition sizes everything 54 * will be nicely aligned, particularly swap for e.g. swapcache, and 55 * clustered operations against larger physical sector sizes for newer HDs, 56 * and so forth. 57 */ 58 #define PALIGN_SIZE (1024 * 1024) 59 #define PALIGN_MASK (PALIGN_SIZE - 1) 60 61 /* 62 * Retrieve the partition start and extent, in blocks. Return 0 on success, 63 * EINVAL on error. 64 */ 65 static int 66 l64_getpartbounds(struct diskslices *ssp, disklabel_t lp, u_int32_t part, 67 u_int64_t *start, u_int64_t *blocks) 68 { 69 struct partition64 *pp; 70 71 if (part >= lp.lab64->d_npartitions) 72 return (EINVAL); 73 74 pp = &lp.lab64->d_partitions[part]; 75 76 if ((pp->p_boffset & (ssp->dss_secsize - 1)) || 77 (pp->p_bsize & (ssp->dss_secsize - 1))) { 78 return (EINVAL); 79 } 80 *start = pp->p_boffset / ssp->dss_secsize; 81 *blocks = pp->p_bsize / ssp->dss_secsize; 82 return(0); 83 } 84 85 /* 86 * Get the filesystem type XXX - diskslices code needs to use uuids 87 */ 88 static void 89 l64_loadpartinfo(disklabel_t lp, u_int32_t part, struct partinfo *dpart) 90 { 91 struct partition64 *pp; 92 const size_t uuid_size = sizeof(struct uuid); 93 94 if (part < lp.lab64->d_npartitions) { 95 pp = &lp.lab64->d_partitions[part]; 96 dpart->fstype_uuid = pp->p_type_uuid; 97 dpart->storage_uuid = pp->p_stor_uuid; 98 dpart->fstype = pp->p_fstype; 99 } else { 100 bzero(&dpart->fstype_uuid, uuid_size); 101 bzero(&dpart->storage_uuid, uuid_size); 102 dpart->fstype = 0; 103 } 104 } 105 106 /* 107 * Get the number of partitions 108 */ 109 static u_int32_t 110 l64_getnumparts(disklabel_t lp) 111 { 112 return(lp.lab64->d_npartitions); 113 } 114 115 static void 116 l64_freedisklabel(disklabel_t *lpp) 117 { 118 kfree((*lpp).lab64, M_DEVBUF); 119 (*lpp).lab64 = NULL; 120 } 121 122 /* 123 * Attempt to read a disk label from a device. 64 bit disklabels are 124 * sector-agnostic and begin at offset 0 on the device. 64 bit disklabels 125 * may only be used with GPT partitioning schemes. 126 * 127 * Returns NULL on sucess, and an error string on failure. 128 */ 129 static const char * 130 l64_readdisklabel(cdev_t dev, struct diskslice *sp, disklabel_t *lpp, 131 struct disk_info *info) 132 { 133 struct buf *bp; 134 struct disklabel64 *dlp; 135 const char *msg; 136 uint32_t savecrc; 137 size_t dlpcrcsize; 138 size_t bpsize; 139 int secsize; 140 141 /* 142 * XXX I/O size is subject to device DMA limitations 143 */ 144 secsize = info->d_media_blksize; 145 bpsize = (sizeof(*dlp) + secsize - 1) & ~(secsize - 1); 146 147 bp = geteblk(bpsize); 148 bp->b_bio1.bio_offset = 0; 149 bp->b_bio1.bio_done = biodone_sync; 150 bp->b_bio1.bio_flags |= BIO_SYNC; 151 bp->b_bcount = bpsize; 152 bp->b_flags &= ~B_INVAL; 153 bp->b_flags |= B_FAILONDIS; 154 bp->b_cmd = BUF_CMD_READ; 155 dev_dstrategy(dev, &bp->b_bio1); 156 157 if (biowait(&bp->b_bio1, "labrd")) { 158 msg = "I/O error"; 159 } else { 160 dlp = (struct disklabel64 *)bp->b_data; 161 dlpcrcsize = offsetof(struct disklabel64, 162 d_partitions[dlp->d_npartitions]) - 163 offsetof(struct disklabel64, d_magic); 164 savecrc = dlp->d_crc; 165 dlp->d_crc = 0; 166 if (dlp->d_magic != DISKMAGIC64) { 167 msg = "no disk label"; 168 } else if (dlp->d_npartitions > MAXPARTITIONS64) { 169 msg = "disklabel64 corrupted, too many partitions"; 170 } else if (savecrc != crc32(&dlp->d_magic, dlpcrcsize)) { 171 msg = "disklabel64 corrupted, bad CRC"; 172 } else { 173 dlp->d_crc = savecrc; 174 (*lpp).lab64 = kmalloc(sizeof(*dlp), 175 M_DEVBUF, M_WAITOK|M_ZERO); 176 *(*lpp).lab64 = *dlp; 177 msg = NULL; 178 } 179 } 180 bp->b_flags |= B_INVAL | B_AGE; 181 brelse(bp); 182 return (msg); 183 } 184 185 /* 186 * If everything is good, copy olpx to nlpx. Check to see if any 187 * open partitions would change. 188 */ 189 static int 190 l64_setdisklabel(disklabel_t olpx, disklabel_t nlpx, struct diskslices *ssp, 191 struct diskslice *sp, u_int32_t *openmask) 192 { 193 struct disklabel64 *olp, *nlp; 194 struct partition64 *opp, *npp; 195 uint32_t savecrc; 196 uint64_t slicebsize; 197 size_t nlpcrcsize; 198 int i; 199 200 olp = olpx.lab64; 201 nlp = nlpx.lab64; 202 203 slicebsize = (uint64_t)sp->ds_size * ssp->dss_secsize; 204 205 if (nlp->d_magic != DISKMAGIC64) 206 return (EINVAL); 207 if (nlp->d_npartitions > MAXPARTITIONS64) 208 return (EINVAL); 209 savecrc = nlp->d_crc; 210 nlp->d_crc = 0; 211 nlpcrcsize = offsetof(struct disklabel64, 212 d_partitions[nlp->d_npartitions]) - 213 offsetof(struct disklabel64, d_magic); 214 if (crc32(&nlp->d_magic, nlpcrcsize) != savecrc) { 215 nlp->d_crc = savecrc; 216 return (EINVAL); 217 } 218 nlp->d_crc = savecrc; 219 220 /* 221 * Check if open partitions have changed 222 */ 223 i = 0; 224 while (i < MAXPARTITIONS64) { 225 if (openmask[i >> 5] == 0) { 226 i += 32; 227 continue; 228 } 229 if ((openmask[i >> 5] & (1 << (i & 31))) == 0) { 230 ++i; 231 continue; 232 } 233 if (nlp->d_npartitions <= i) 234 return (EBUSY); 235 opp = &olp->d_partitions[i]; 236 npp = &nlp->d_partitions[i]; 237 if (npp->p_boffset != opp->p_boffset || 238 npp->p_bsize < opp->p_bsize) { 239 return (EBUSY); 240 } 241 242 /* 243 * Do not allow p_type_uuid or p_stor_uuid to change if 244 * the partition is currently open. 245 */ 246 if (bcmp(&npp->p_type_uuid, &opp->p_type_uuid, 247 sizeof(npp->p_type_uuid)) != 0) { 248 return (EBUSY); 249 } 250 if (bcmp(&npp->p_stor_uuid, &opp->p_stor_uuid, 251 sizeof(npp->p_stor_uuid)) != 0) { 252 return (EBUSY); 253 } 254 ++i; 255 } 256 257 /* 258 * Make sure the label and partition offsets and sizes are sane. 259 */ 260 if (nlp->d_total_size > slicebsize) 261 return (ENOSPC); 262 if (nlp->d_total_size & (ssp->dss_secsize - 1)) 263 return (EINVAL); 264 if (nlp->d_bbase & (ssp->dss_secsize - 1)) 265 return (EINVAL); 266 if (nlp->d_pbase & (ssp->dss_secsize - 1)) 267 return (EINVAL); 268 if (nlp->d_pstop & (ssp->dss_secsize - 1)) 269 return (EINVAL); 270 if (nlp->d_abase & (ssp->dss_secsize - 1)) 271 return (EINVAL); 272 273 for (i = 0; i < nlp->d_npartitions; ++i) { 274 npp = &nlp->d_partitions[i]; 275 if (npp->p_bsize == 0) { 276 if (npp->p_boffset != 0) 277 return (EINVAL); 278 continue; 279 } 280 if (npp->p_boffset & (ssp->dss_secsize - 1)) 281 return (EINVAL); 282 if (npp->p_bsize & (ssp->dss_secsize - 1)) 283 return (EINVAL); 284 if (npp->p_boffset < nlp->d_pbase) 285 return (ENOSPC); 286 if (npp->p_boffset + npp->p_bsize > nlp->d_total_size) 287 return (ENOSPC); 288 } 289 290 /* 291 * Structurally we may add code to make modifications above in the 292 * future, so regenerate the crc anyway. 293 */ 294 nlp->d_crc = 0; 295 nlp->d_crc = crc32(&nlp->d_magic, nlpcrcsize); 296 *olp = *nlp; 297 298 return (0); 299 } 300 301 /* 302 * Write disk label back to device after modification. 303 */ 304 static int 305 l64_writedisklabel(cdev_t dev, struct diskslices *ssp, 306 struct diskslice *sp, disklabel_t lpx) 307 { 308 struct disklabel64 *lp; 309 struct disklabel64 *dlp; 310 struct buf *bp; 311 int error = 0; 312 size_t bpsize; 313 int secsize; 314 315 lp = lpx.lab64; 316 317 /* 318 * XXX I/O size is subject to device DMA limitations 319 */ 320 secsize = ssp->dss_secsize; 321 bpsize = (sizeof(*lp) + secsize - 1) & ~(secsize - 1); 322 323 bp = geteblk(bpsize); 324 bp->b_bio1.bio_offset = 0; 325 bp->b_bio1.bio_done = biodone_sync; 326 bp->b_bio1.bio_flags |= BIO_SYNC; 327 bp->b_bcount = bpsize; 328 bp->b_flags |= B_FAILONDIS; 329 330 /* 331 * Because our I/O is larger then the label, and because we do not 332 * write the d_reserved0[] area, do a read-modify-write. 333 */ 334 bp->b_flags &= ~B_INVAL; 335 bp->b_cmd = BUF_CMD_READ; 336 KKASSERT(dkpart(dev) == WHOLE_SLICE_PART); 337 dev_dstrategy(dev, &bp->b_bio1); 338 error = biowait(&bp->b_bio1, "labrd"); 339 if (error) 340 goto done; 341 342 dlp = (void *)bp->b_data; 343 bcopy(&lp->d_magic, &dlp->d_magic, 344 sizeof(*lp) - offsetof(struct disklabel64, d_magic)); 345 bp->b_cmd = BUF_CMD_WRITE; 346 bp->b_bio1.bio_done = biodone_sync; 347 bp->b_bio1.bio_flags |= BIO_SYNC; 348 KKASSERT(dkpart(dev) == WHOLE_SLICE_PART); 349 dev_dstrategy(dev, &bp->b_bio1); 350 error = biowait(&bp->b_bio1, "labwr"); 351 done: 352 bp->b_flags |= B_INVAL | B_AGE; 353 brelse(bp); 354 return (error); 355 } 356 357 /* 358 * Create a disklabel based on a disk_info structure for the purposes of 359 * DSO_COMPATLABEL - cases where no real label exists on the storage medium. 360 * 361 * If a diskslice is passed, the label is truncated to the slice. 362 * 363 * NOTE! This is not a legal label because d_bbase and d_pbase are both 364 * set to 0. 365 */ 366 static disklabel_t 367 l64_clone_label(struct disk_info *info, struct diskslice *sp) 368 { 369 struct disklabel64 *lp; 370 disklabel_t res; 371 uint32_t blksize = info->d_media_blksize; 372 size_t lpcrcsize; 373 374 lp = kmalloc(sizeof *lp, M_DEVBUF, M_WAITOK | M_ZERO); 375 376 if (sp) 377 lp->d_total_size = (uint64_t)sp->ds_size * blksize; 378 else 379 lp->d_total_size = info->d_media_blocks * blksize; 380 381 lp->d_magic = DISKMAGIC64; 382 lp->d_align = blksize; 383 lp->d_npartitions = MAXPARTITIONS64; 384 lp->d_pstop = lp->d_total_size; 385 386 /* 387 * Create a dummy 'c' part and a dummy 'a' part (if requested). 388 * Note that the 'c' part is really a hack. 64 bit disklabels 389 * do not use 'c' to mean the raw partition. 390 */ 391 392 lp->d_partitions[2].p_boffset = 0; 393 lp->d_partitions[2].p_bsize = lp->d_total_size; 394 /* XXX SET FS TYPE */ 395 396 if (info->d_dsflags & DSO_COMPATPARTA) { 397 lp->d_partitions[0].p_boffset = 0; 398 lp->d_partitions[0].p_bsize = lp->d_total_size; 399 /* XXX SET FS TYPE */ 400 } 401 402 lpcrcsize = offsetof(struct disklabel64, 403 d_partitions[lp->d_npartitions]) - 404 offsetof(struct disklabel64, d_magic); 405 406 lp->d_crc = crc32(&lp->d_magic, lpcrcsize); 407 res.lab64 = lp; 408 return (res); 409 } 410 411 /* 412 * Create a virgin disklabel64 suitable for writing to the media. 413 * 414 * disklabel64 always reserves 32KB for a boot area and leaves room 415 * for up to RESPARTITIONS64 partitions. 416 */ 417 static void 418 l64_makevirginlabel(disklabel_t lpx, struct diskslices *ssp, 419 struct diskslice *sp, struct disk_info *info) 420 { 421 struct disklabel64 *lp = lpx.lab64; 422 struct partition64 *pp; 423 uint32_t blksize; 424 uint32_t ressize; 425 uint64_t blkmask; /* 64 bits so we can ~ */ 426 size_t lpcrcsize; 427 428 /* 429 * Setup the initial label. Use of a block size of at least 4KB 430 * for calculating the initial reserved areas to allow some degree 431 * of portability between media with different sector sizes. 432 * 433 * Note that the modified blksize is stored in d_align as a hint 434 * to the disklabeling program. 435 */ 436 bzero(lp, sizeof(*lp)); 437 if ((blksize = info->d_media_blksize) < 4096) 438 blksize = 4096; 439 blkmask = blksize - 1; 440 441 if (sp) 442 lp->d_total_size = (uint64_t)sp->ds_size * ssp->dss_secsize; 443 else 444 lp->d_total_size = info->d_media_blocks * info->d_media_blksize; 445 446 lp->d_magic = DISKMAGIC64; 447 lp->d_align = blksize; 448 lp->d_npartitions = MAXPARTITIONS64; 449 kern_uuidgen(&lp->d_stor_uuid, 1); 450 451 ressize = offsetof(struct disklabel64, d_partitions[RESPARTITIONS64]); 452 ressize = (ressize + (uint32_t)blkmask) & ~blkmask; 453 454 /* 455 * NOTE: When calculating pbase take into account the slice offset 456 * so the partitions are at least 32K-aligned relative to the 457 * start of the physical disk. This will accomodate efficient 458 * access to 4096 byte physical sector drives. 459 */ 460 lp->d_bbase = ressize; 461 lp->d_pbase = lp->d_bbase + ((32768 + blkmask) & ~blkmask); 462 lp->d_pbase = (lp->d_pbase + PALIGN_MASK) & ~(uint64_t)PALIGN_MASK; 463 464 /* adjust for slice offset so we are physically aligned */ 465 lp->d_pbase += 32768 - (sp->ds_offset * info->d_media_blksize) % 32768; 466 467 lp->d_pstop = (lp->d_total_size - lp->d_bbase) & ~blkmask; 468 lp->d_abase = lp->d_pstop; 469 470 /* 471 * All partitions are left empty unless DSO_COMPATPARTA is set 472 */ 473 474 if (info->d_dsflags & DSO_COMPATPARTA) { 475 pp = &lp->d_partitions[0]; 476 pp->p_boffset = lp->d_pbase; 477 pp->p_bsize = lp->d_pstop - lp->d_pbase; 478 /* XXX SET FS TYPE */ 479 } 480 481 lpcrcsize = offsetof(struct disklabel64, 482 d_partitions[lp->d_npartitions]) - 483 offsetof(struct disklabel64, d_magic); 484 lp->d_crc = crc32(&lp->d_magic, lpcrcsize); 485 } 486 487 /* 488 * Set the number of blocks at the beginning of the slice which have 489 * been reserved for label operations. This area will be write-protected 490 * when accessed via the slice. 491 * 492 * For now just protect the label area proper. Do not protect the 493 * boot area. Note partitions in 64 bit disklabels do not overlap 494 * the disklabel or boot area. 495 */ 496 static void 497 l64_adjust_label_reserved(struct diskslices *ssp, int slice, 498 struct diskslice *sp) 499 { 500 struct disklabel64 *lp = sp->ds_label.lab64; 501 502 sp->ds_reserved = lp->d_bbase / ssp->dss_secsize; 503 } 504 505 struct disklabel_ops disklabel64_ops = { 506 .labelsize = sizeof(struct disklabel64), 507 .op_readdisklabel = l64_readdisklabel, 508 .op_setdisklabel = l64_setdisklabel, 509 .op_writedisklabel = l64_writedisklabel, 510 .op_clone_label = l64_clone_label, 511 .op_adjust_label_reserved = l64_adjust_label_reserved, 512 .op_getpartbounds = l64_getpartbounds, 513 .op_loadpartinfo = l64_loadpartinfo, 514 .op_getnumparts = l64_getnumparts, 515 .op_makevirginlabel = l64_makevirginlabel, 516 .op_freedisklabel = l64_freedisklabel 517 }; 518 519