1 /* $NetBSD: device-mapper.c,v 1.22 2010/03/26 15:46:04 jakllsch Exp $ */ 2 3 /* 4 * Copyright (c) 2010-2011 Alex Hornung <alex@alexhornung.com> 5 * Copyright (c) 2010 The NetBSD Foundation, Inc. 6 * All rights reserved. 7 * 8 * This code is derived from software contributed to The NetBSD Foundation 9 * by Adam Hamsik. 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 * I want to say thank you to all people who helped me with this project. 35 */ 36 37 #include <sys/types.h> 38 #include <sys/ctype.h> 39 40 #include <sys/buf.h> 41 #include <sys/conf.h> 42 #include <sys/device.h> 43 #include <sys/disk.h> 44 #include <sys/disklabel.h> 45 #include <sys/dtype.h> 46 #include <sys/malloc.h> 47 #include <sys/module.h> 48 #include <sys/sysctl.h> 49 #include <dev/disk/dm/dm.h> 50 51 #include "netbsd-dm.h" 52 53 static d_ioctl_t dmioctl; 54 static d_open_t dmopen; 55 static d_close_t dmclose; 56 static d_psize_t dmsize; 57 static d_strategy_t dmstrategy; 58 static d_dump_t dmdump; 59 60 /* New module handle and destroy routines */ 61 static int dm_modcmd(module_t mod, int cmd, void *unused); 62 static int dmdestroy(void); 63 64 static void dm_doinit(void); 65 66 static int dm_cmd_to_fun(prop_dictionary_t); 67 static int disk_ioctl_switch(cdev_t, u_long, void *); 68 static int dm_ioctl_switch(u_long); 69 #if 0 70 static void dmminphys(struct buf *); 71 #endif 72 73 /* ***Variable-definitions*** */ 74 struct dev_ops dm_ops = { 75 { "dm", 0, D_DISK | D_MPSAFE }, 76 .d_open = dmopen, 77 .d_close = dmclose, 78 .d_read = physread, 79 .d_write = physwrite, 80 .d_ioctl = dmioctl, 81 .d_strategy = dmstrategy, 82 .d_psize = dmsize, 83 .d_dump = dmdump, 84 /* D_DISK */ 85 }; 86 87 MALLOC_DEFINE(M_DM, "dm", "Device Mapper allocations"); 88 89 int dm_debug_level = 0; 90 91 extern uint64_t dm_dev_counter; 92 93 static cdev_t dmcdev; 94 95 static moduledata_t dm_mod = { 96 "dm", 97 dm_modcmd, 98 NULL 99 }; 100 DECLARE_MODULE(dm, dm_mod, SI_SUB_RAID, SI_ORDER_ANY); 101 MODULE_VERSION(dm, 1); 102 103 /* 104 * This array is used to translate cmd to function pointer. 105 * 106 * Interface between libdevmapper and lvm2tools uses different 107 * names for one IOCTL call because libdevmapper do another thing 108 * then. When I run "info" or "mknodes" libdevmapper will send same 109 * ioctl to kernel but will do another things in userspace. 110 * 111 */ 112 static struct cmd_function cmd_fn[] = { 113 { .cmd = "version", .fn = dm_get_version_ioctl}, 114 { .cmd = "targets", .fn = dm_list_versions_ioctl}, 115 { .cmd = "create", .fn = dm_dev_create_ioctl}, 116 { .cmd = "info", .fn = dm_dev_status_ioctl}, 117 { .cmd = "mknodes", .fn = dm_dev_status_ioctl}, 118 { .cmd = "names", .fn = dm_dev_list_ioctl}, 119 { .cmd = "suspend", .fn = dm_dev_suspend_ioctl}, 120 { .cmd = "remove", .fn = dm_dev_remove_ioctl}, 121 { .cmd = "remove_all", .fn = dm_dev_remove_all_ioctl}, 122 { .cmd = "rename", .fn = dm_dev_rename_ioctl}, 123 { .cmd = "resume", .fn = dm_dev_resume_ioctl}, 124 { .cmd = "clear", .fn = dm_table_clear_ioctl}, 125 { .cmd = "deps", .fn = dm_table_deps_ioctl}, 126 { .cmd = "reload", .fn = dm_table_load_ioctl}, 127 { .cmd = "status", .fn = dm_table_status_ioctl}, 128 { .cmd = "table", .fn = dm_table_status_ioctl}, 129 { .cmd = "message", .fn = dm_message_ioctl}, 130 {NULL, NULL} 131 }; 132 133 /* New module handle routine */ 134 static int 135 dm_modcmd(module_t mod, int cmd, void *unused) 136 { 137 int error; 138 139 error = 0; 140 141 switch (cmd) { 142 case MOD_LOAD: 143 dm_doinit(); 144 kprintf("Device Mapper version %d.%d.%d loaded\n", 145 DM_VERSION_MAJOR, DM_VERSION_MINOR, DM_VERSION_PATCHLEVEL); 146 break; 147 148 case MOD_UNLOAD: 149 /* 150 * Disable unloading of dm module if there are any devices 151 * defined in driver. This is probably too strong we need 152 * to disable auto-unload only if there is mounted dm device 153 * present. 154 */ 155 if (dm_dev_counter > 0) 156 return EBUSY; 157 158 error = dmdestroy(); 159 if (error) 160 break; 161 kprintf("Device Mapper unloaded\n"); 162 break; 163 164 default: 165 break; 166 } 167 168 return error; 169 } 170 171 static void 172 dm_doinit(void) 173 { 174 dm_target_init(); 175 dm_dev_init(); 176 dm_pdev_init(); 177 dmcdev = make_dev(&dm_ops, 0, UID_ROOT, GID_OPERATOR, 0640, "mapper/control"); 178 } 179 180 /* Destroy routine */ 181 static int 182 dmdestroy(void) 183 { 184 destroy_dev(dmcdev); 185 186 dm_dev_uninit(); 187 dm_pdev_uninit(); 188 dm_target_uninit(); 189 190 return 0; 191 } 192 193 static int 194 dmopen(struct dev_open_args *ap) 195 { 196 cdev_t dev = ap->a_head.a_dev; 197 dm_dev_t *dmv; 198 199 /* Shortcut for the control device */ 200 if (minor(dev) == 0) 201 return 0; 202 203 if ((dmv = dm_dev_lookup(NULL, NULL, minor(dev))) == NULL) 204 return ENXIO; 205 206 dmv->is_open = 1; 207 dm_dev_unbusy(dmv); 208 209 aprint_debug("dm open routine called %" PRIu32 "\n", 210 minor(ap->a_head.a_dev)); 211 return 0; 212 } 213 214 static int 215 dmclose(struct dev_close_args *ap) 216 { 217 cdev_t dev = ap->a_head.a_dev; 218 dm_dev_t *dmv; 219 220 /* Shortcut for the control device */ 221 if (minor(dev) == 0) 222 return 0; 223 224 if ((dmv = dm_dev_lookup(NULL, NULL, minor(dev))) == NULL) 225 return ENXIO; 226 227 dmv->is_open = 0; 228 dm_dev_unbusy(dmv); 229 230 aprint_debug("dm close routine called %" PRIu32 "\n", 231 minor(ap->a_head.a_dev)); 232 return 0; 233 } 234 235 236 static int 237 dmioctl(struct dev_ioctl_args *ap) 238 { 239 cdev_t dev = ap->a_head.a_dev; 240 u_long cmd = ap->a_cmd; 241 void *data = ap->a_data; 242 243 int r, err; 244 prop_dictionary_t dm_dict_in; 245 246 err = r = 0; 247 248 aprint_debug("dmioctl called\n"); 249 250 KKASSERT(data != NULL); 251 252 if ((r = disk_ioctl_switch(dev, cmd, data)) == ENOTTY) { 253 struct plistref *pref = (struct plistref *) data; 254 255 /* Check if we were called with NETBSD_DM_IOCTL ioctl 256 otherwise quit. */ 257 if ((r = dm_ioctl_switch(cmd)) != 0) 258 return r; 259 260 if ((r = prop_dictionary_copyin_ioctl(pref, cmd, &dm_dict_in)) != 0) 261 return r; 262 263 if ((r = dm_check_version(dm_dict_in)) != 0) 264 goto cleanup_exit; 265 266 /* run ioctl routine */ 267 if ((err = dm_cmd_to_fun(dm_dict_in)) != 0) 268 goto cleanup_exit; 269 270 cleanup_exit: 271 r = prop_dictionary_copyout_ioctl(pref, cmd, dm_dict_in); 272 prop_object_release(dm_dict_in); 273 } 274 275 /* 276 * Return the error of the actual command if one one has 277 * happened. Otherwise return 'r' which indicates errors 278 * that occurred during helper operations. 279 */ 280 return (err != 0)?err:r; 281 } 282 283 /* 284 * Translate command sent from libdevmapper to func. 285 */ 286 static int 287 dm_cmd_to_fun(prop_dictionary_t dm_dict) 288 { 289 int i, r; 290 prop_string_t command; 291 292 r = 0; 293 294 if ((command = prop_dictionary_get(dm_dict, DM_IOCTL_COMMAND)) == NULL) 295 return EINVAL; 296 297 for (i = 0; cmd_fn[i].cmd != NULL; i++) 298 if (prop_string_equals_cstring(command, cmd_fn[i].cmd)) 299 break; 300 301 if (cmd_fn[i].cmd == NULL) 302 return EINVAL; 303 304 aprint_debug("ioctl %s called\n", cmd_fn[i].cmd); 305 r = cmd_fn[i].fn(dm_dict); 306 307 return r; 308 } 309 310 /* Call apropriate ioctl handler function. */ 311 static int 312 dm_ioctl_switch(u_long cmd) 313 { 314 315 switch(cmd) { 316 317 case NETBSD_DM_IOCTL: 318 aprint_debug("dm NETBSD_DM_IOCTL called\n"); 319 break; 320 default: 321 aprint_debug("dm unknown ioctl called\n"); 322 return ENOTTY; 323 break; /* NOT REACHED */ 324 } 325 326 return 0; 327 } 328 329 /* 330 * Check for disk specific ioctls. 331 */ 332 333 static int 334 disk_ioctl_switch(cdev_t dev, u_long cmd, void *data) 335 { 336 dm_dev_t *dmv; 337 338 /* disk ioctls make sense only on block devices */ 339 if (minor(dev) == 0) 340 return ENOTTY; 341 342 switch(cmd) { 343 case DIOCGPART: 344 { 345 struct partinfo *dpart; 346 u_int64_t size; 347 dpart = data; 348 bzero(dpart, sizeof(*dpart)); 349 350 if ((dmv = dev->si_drv1) == NULL) 351 return ENODEV; 352 if (dmv->diskp->d_info.d_media_blksize == 0) { 353 return ENOTSUP; 354 } else { 355 size = dm_table_size(&dmv->table_head); 356 dpart->media_offset = 0; 357 dpart->media_size = size * DEV_BSIZE; 358 dpart->media_blocks = size; 359 dpart->media_blksize = DEV_BSIZE; 360 dpart->fstype = FS_BSDFFS; 361 } 362 break; 363 } 364 365 default: 366 aprint_debug("unknown disk_ioctl called\n"); 367 return ENOTTY; 368 break; /* NOT REACHED */ 369 } 370 371 return 0; 372 } 373 374 /* 375 * Do all IO operations on dm logical devices. 376 */ 377 static int 378 dmstrategy(struct dev_strategy_args *ap) 379 { 380 cdev_t dev = ap->a_head.a_dev; 381 struct bio *bio = ap->a_bio; 382 struct buf *bp = bio->bio_buf; 383 int bypass; 384 385 dm_dev_t *dmv; 386 dm_table_t *tbl; 387 dm_table_entry_t *table_en; 388 struct buf *nestbuf; 389 390 uint64_t buf_start, buf_len, issued_len; 391 uint64_t table_start, table_end; 392 uint64_t start, end; 393 394 buf_start = bio->bio_offset; 395 buf_len = bp->b_bcount; 396 397 tbl = NULL; 398 399 table_end = 0; 400 issued_len = 0; 401 402 dmv = dev->si_drv1; 403 404 switch(bp->b_cmd) { 405 case BUF_CMD_READ: 406 case BUF_CMD_WRITE: 407 case BUF_CMD_FREEBLKS: 408 bypass = 0; 409 break; 410 case BUF_CMD_FLUSH: 411 bypass = 1; 412 KKASSERT(buf_len == 0); 413 break; 414 default: 415 bp->b_error = EIO; 416 bp->b_resid = bp->b_bcount; 417 biodone(bio); 418 return 0; 419 } 420 421 if (bypass == 0 && 422 bounds_check_with_mediasize(bio, DEV_BSIZE, 423 dm_table_size(&dmv->table_head)) <= 0) { 424 bp->b_resid = bp->b_bcount; 425 biodone(bio); 426 return 0; 427 } 428 429 /* Select active table */ 430 tbl = dm_table_get_entry(&dmv->table_head, DM_TABLE_ACTIVE); 431 432 nestiobuf_init(bio); 433 devstat_start_transaction(&dmv->stats); 434 435 /* 436 * Find out what tables I want to select. 437 */ 438 SLIST_FOREACH(table_en, tbl, next) { 439 /* 440 * I need need number of bytes not blocks. 441 */ 442 table_start = table_en->start * DEV_BSIZE; 443 table_end = table_start + (table_en->length) * DEV_BSIZE; 444 445 /* 446 * Calculate the start and end 447 */ 448 start = MAX(table_start, buf_start); 449 end = MIN(table_end, buf_start + buf_len); 450 451 if (dm_debug_level) { 452 aprint_normal("----------------------------------------\n"); 453 aprint_normal("table_start %010" PRIu64", table_end %010" 454 PRIu64 "\n", table_start, table_end); 455 aprint_normal("buf_start %010" PRIu64", buf_len %010" 456 PRIu64"\n", buf_start, buf_len); 457 aprint_normal("start-buf_start %010"PRIu64", end %010" 458 PRIu64"\n", start - buf_start, end); 459 aprint_normal("start %010" PRIu64", end %010" 460 PRIu64"\n", start, end); 461 aprint_normal("----------------------------------------\n"); 462 } 463 464 if (bypass) { 465 nestbuf = getpbuf(NULL); 466 nestbuf->b_flags |= bio->bio_buf->b_flags & B_HASBOGUS; 467 468 nestiobuf_add(bio, nestbuf, 0, 0, &dmv->stats); 469 nestbuf->b_bio1.bio_offset = 0; 470 table_en->target->strategy(table_en, nestbuf); 471 } else if (start < end) { 472 nestbuf = getpbuf(NULL); 473 nestbuf->b_flags |= bio->bio_buf->b_flags & B_HASBOGUS; 474 475 nestiobuf_add(bio, nestbuf, 476 start - buf_start, (end - start), 477 &dmv->stats); 478 issued_len += end - start; 479 480 nestbuf->b_bio1.bio_offset = (start - table_start); 481 table_en->target->strategy(table_en, nestbuf); 482 } 483 } 484 485 if (issued_len < buf_len) 486 nestiobuf_error(bio, EINVAL); 487 nestiobuf_start(bio); 488 dm_table_release(&dmv->table_head, DM_TABLE_ACTIVE); 489 490 return 0; 491 } 492 493 static int 494 dmdump(struct dev_dump_args *ap) 495 { 496 cdev_t dev = ap->a_head.a_dev; 497 dm_dev_t *dmv; 498 dm_table_t *tbl; 499 dm_table_entry_t *table_en; 500 uint64_t buf_start, buf_len, issued_len; 501 uint64_t table_start, table_end; 502 uint64_t start, end, data_offset; 503 off_t offset; 504 size_t length; 505 int error = 0; 506 507 buf_start = ap->a_offset; 508 buf_len = ap->a_length; 509 510 tbl = NULL; 511 512 table_end = 0; 513 issued_len = 0; 514 515 dmv = dev->si_drv1; 516 517 /* Select active table */ 518 tbl = dm_table_get_entry(&dmv->table_head, DM_TABLE_ACTIVE); 519 520 521 /* 522 * Find out what tables I want to select. 523 */ 524 SLIST_FOREACH(table_en, tbl, next) { 525 /* 526 * I need need number of bytes not blocks. 527 */ 528 table_start = table_en->start * DEV_BSIZE; 529 table_end = table_start + (table_en->length) * DEV_BSIZE; 530 531 /* 532 * Calculate the start and end 533 */ 534 start = MAX(table_start, buf_start); 535 end = MIN(table_end, buf_start + buf_len); 536 537 if (ap->a_length == 0) { 538 if (table_en->target->dump == NULL) { 539 error = ENXIO; 540 goto out; 541 } 542 543 table_en->target->dump(table_en, NULL, 0, 0); 544 } else if (start < end) { 545 data_offset = start - buf_start; 546 offset = start - table_start; 547 length = end - start; 548 549 if (table_en->target->dump == NULL) { 550 error = ENXIO; 551 goto out; 552 } 553 554 table_en->target->dump(table_en, 555 (char *)ap->a_virtual + data_offset, 556 length, offset); 557 558 issued_len += end - start; 559 } 560 } 561 562 if (issued_len < buf_len) 563 error = EINVAL; 564 565 out: 566 dm_table_release(&dmv->table_head, DM_TABLE_ACTIVE); 567 568 return error; 569 } 570 571 static int 572 dmsize(struct dev_psize_args *ap) 573 { 574 cdev_t dev = ap->a_head.a_dev; 575 dm_dev_t *dmv; 576 uint64_t size; 577 578 size = 0; 579 580 if ((dmv = dev->si_drv1) == NULL) 581 return ENXIO; 582 583 size = dm_table_size(&dmv->table_head); 584 ap->a_result = (int64_t)size; 585 586 return 0; 587 } 588 589 #if 0 590 static void 591 dmminphys(struct buf *bp) 592 { 593 594 bp->b_bcount = MIN(bp->b_bcount, MAXPHYS); 595 } 596 #endif 597 598 void 599 dmsetdiskinfo(struct disk *disk, dm_table_head_t *head) 600 { 601 struct disk_info info; 602 uint64_t dmp_size; 603 604 dmp_size = dm_table_size(head); 605 606 bzero(&info, sizeof(struct disk_info)); 607 info.d_media_blksize = DEV_BSIZE; 608 info.d_media_blocks = dmp_size; 609 #if 0 610 /* this is set by disk_setdiskinfo */ 611 info.d_media_size = dmp_size * DEV_BSIZE; 612 #endif 613 info.d_dsflags = DSO_MBRQUIET | DSO_DEVICEMAPPER | DSO_RAWPSIZE; 614 615 info.d_secpertrack = 32; 616 info.d_nheads = 64; 617 info.d_secpercyl = info.d_secpertrack * info.d_nheads; 618 info.d_ncylinders = dmp_size / info.d_secpercyl; 619 620 disk_setdiskinfo(disk, &info); 621 } 622 623 /* 624 * Transform char s to uint64_t offset number. 625 */ 626 uint64_t 627 atoi64(const char *s) 628 { 629 uint64_t n; 630 n = 0; 631 632 while (*s != '\0') { 633 if (!isdigit(*s)) 634 break; 635 636 n = (10 * n) + (*s - '0'); 637 s++; 638 } 639 640 return n; 641 } 642 643 void 644 dm_builtin_init(void *arg) 645 { 646 modeventhand_t evh = (modeventhand_t)arg; 647 648 KKASSERT(evh != NULL); 649 evh(NULL, MOD_LOAD, NULL); 650 } 651 652 void 653 dm_builtin_uninit(void *arg) 654 { 655 modeventhand_t evh = (modeventhand_t)arg; 656 657 KKASSERT(evh != NULL); 658 evh(NULL, MOD_UNLOAD, NULL); 659 } 660 661 TUNABLE_INT("debug.dm_debug", &dm_debug_level); 662 SYSCTL_INT(_debug, OID_AUTO, dm_debug, CTLFLAG_RW, &dm_debug_level, 663 0, "Enable device mapper debugging"); 664 665