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