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 #if 0 66 static void dmminphys(struct buf *); 67 #endif 68 69 struct dev_ops dm_ops = { 70 { "dm", 0, D_DISK | D_MPSAFE }, 71 .d_open = dmopen, 72 .d_close = dmclose, 73 .d_read = physread, 74 .d_write = physwrite, 75 .d_ioctl = dmioctl, 76 .d_strategy = dmstrategy, 77 .d_psize = dmsize, 78 .d_dump = dmdump, 79 }; 80 81 MALLOC_DEFINE(M_DM, "dm", "Device Mapper allocations"); 82 83 int dm_debug_level = 0; 84 85 extern uint64_t dm_dev_counter; 86 87 static cdev_t dmcdev; 88 89 static moduledata_t dm_mod = { 90 "dm", 91 dm_modcmd, 92 NULL 93 }; 94 DECLARE_MODULE(dm, dm_mod, SI_SUB_RAID, SI_ORDER_ANY); 95 MODULE_VERSION(dm, 1); 96 97 /* 98 * This structure is used to translate command sent to kernel driver in 99 * <key>command</key> 100 * <value></value> 101 * to function 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 static struct cmd_function { 112 const char *cmd; 113 int (*fn)(prop_dictionary_t); 114 } cmd_fn[] = { 115 {.cmd = "version", .fn = NULL}, 116 {.cmd = "targets", .fn = dm_list_versions_ioctl}, 117 {.cmd = "create", .fn = dm_dev_create_ioctl}, 118 {.cmd = "info", .fn = dm_dev_status_ioctl}, 119 {.cmd = "mknodes", .fn = dm_dev_status_ioctl}, 120 {.cmd = "names", .fn = dm_dev_list_ioctl}, 121 {.cmd = "suspend", .fn = dm_dev_suspend_ioctl}, 122 {.cmd = "remove", .fn = dm_dev_remove_ioctl}, 123 {.cmd = "remove_all", .fn = dm_dev_remove_all_ioctl}, 124 {.cmd = "rename", .fn = dm_dev_rename_ioctl}, 125 {.cmd = "resume", .fn = dm_dev_resume_ioctl}, 126 {.cmd = "clear", .fn = dm_table_clear_ioctl}, 127 {.cmd = "deps", .fn = dm_table_deps_ioctl}, 128 {.cmd = "reload", .fn = dm_table_load_ioctl}, 129 {.cmd = "status", .fn = dm_table_status_ioctl}, 130 {.cmd = "table", .fn = dm_table_status_ioctl}, 131 {.cmd = "message", .fn = dm_message_ioctl}, 132 }; 133 134 /* 135 * New module handle routine 136 */ 137 static int 138 dm_modcmd(module_t mod, int cmd, void *unused) 139 { 140 int error; 141 142 error = 0; 143 144 switch (cmd) { 145 case MOD_LOAD: 146 dm_doinit(); 147 kprintf("Device Mapper version %d.%d.%d loaded\n", 148 DM_VERSION_MAJOR, DM_VERSION_MINOR, DM_VERSION_PATCHLEVEL); 149 break; 150 151 case MOD_UNLOAD: 152 /* 153 * Disable unloading of dm module if there are any devices 154 * defined in driver. This is probably too strong we need 155 * to disable auto-unload only if there is mounted dm device 156 * present. 157 */ 158 if (dm_dev_counter > 0) 159 return EBUSY; 160 /* race window here */ 161 162 error = dmdestroy(); 163 if (error) 164 break; 165 kprintf("Device Mapper unloaded\n"); 166 break; 167 168 default: 169 break; 170 } 171 172 return error; 173 } 174 175 static void 176 dm_doinit(void) 177 { 178 dm_target_init(); 179 dm_dev_init(); 180 dm_pdev_init(); 181 dmcdev = make_dev(&dm_ops, 0, UID_ROOT, GID_OPERATOR, 0640, "mapper/control"); 182 } 183 184 /* 185 * Destroy routine 186 */ 187 static int 188 dmdestroy(void) 189 { 190 destroy_dev(dmcdev); 191 192 dm_dev_uninit(); 193 dm_pdev_uninit(); 194 dm_target_uninit(); 195 196 return 0; 197 } 198 199 static int 200 dmopen(struct dev_open_args *ap) 201 { 202 cdev_t dev = ap->a_head.a_dev; 203 dm_dev_t *dmv; 204 205 /* Shortcut for the control device */ 206 if (minor(dev) == 0) 207 return 0; 208 209 if ((dmv = dm_dev_lookup(NULL, NULL, minor(dev))) == NULL) 210 return ENXIO; 211 212 dmv->is_open = 1; 213 dm_dev_unbusy(dmv); 214 215 dmdebug("dm open routine called %" PRIu32 "\n", 216 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("dm close routine called %" PRIu32 "\n", 237 minor(ap->a_head.a_dev)); 238 return 0; 239 } 240 241 242 static int 243 dmioctl(struct dev_ioctl_args *ap) 244 { 245 cdev_t dev = ap->a_head.a_dev; 246 u_long cmd = ap->a_cmd; 247 void *data = ap->a_data; 248 struct plistref *pref; 249 250 int r, err; 251 prop_dictionary_t dm_dict_in; 252 253 err = r = 0; 254 255 dmdebug("dmioctl called\n"); 256 KKASSERT(data != NULL); 257 258 if ((r = disk_ioctl_switch(dev, cmd, data)) != ENOTTY) 259 return r; /* Handled disk ioctl */ 260 261 if ((r = dm_ioctl_switch(cmd)) != 0) 262 return r; /* Not NETBSD_DM_IOCTL */ 263 264 pref = (struct plistref *)data; /* data is for libprop */ 265 if ((r = prop_dictionary_copyin_ioctl(pref, cmd, &dm_dict_in)) != 0) 266 return r; 267 268 if ((r = dm_check_version(dm_dict_in)) == 0) 269 err = dm_cmd_to_fun(dm_dict_in); 270 271 r = prop_dictionary_copyout_ioctl(pref, cmd, dm_dict_in); 272 prop_object_release(dm_dict_in); 273 274 /* Return the dm ioctl error if any. */ 275 if (err) 276 return err; 277 return r; 278 } 279 280 /* 281 * Translate command sent from libdevmapper to func. 282 */ 283 static int 284 dm_cmd_to_fun(prop_dictionary_t dm_dict) 285 { 286 int i, size; 287 prop_string_t command; 288 struct cmd_function *p = NULL; 289 290 size = sizeof(cmd_fn) / sizeof(cmd_fn[0]); 291 292 if ((command = prop_dictionary_get(dm_dict, DM_IOCTL_COMMAND)) == NULL) 293 return EINVAL; 294 295 for (i = 0; i < size; i++) { 296 p = &cmd_fn[i]; 297 if (prop_string_equals_cstring(command, p->cmd)) 298 break; 299 } 300 301 KKASSERT(p); 302 if (i == size) { 303 dmdebug("Unknown ioctl\n"); 304 return EINVAL; 305 } 306 307 dmdebug("ioctl %s called %p\n", p->cmd, p->fn); 308 if (p->fn == NULL) 309 return 0; /* No handler required */ 310 311 return p->fn(dm_dict); 312 } 313 314 /* 315 * Call apropriate ioctl handler function. 316 */ 317 static int 318 dm_ioctl_switch(u_long cmd) 319 { 320 321 switch(cmd) { 322 case NETBSD_DM_IOCTL: 323 dmdebug("dm NETBSD_DM_IOCTL called\n"); 324 break; 325 default: 326 dmdebug("dm unknown ioctl called\n"); 327 return ENOTTY; 328 break; /* NOT REACHED */ 329 } 330 331 return 0; 332 } 333 334 /* 335 * Check for disk specific ioctls. 336 */ 337 static int 338 disk_ioctl_switch(cdev_t dev, u_long cmd, void *data) 339 { 340 dm_dev_t *dmv; 341 342 /* disk ioctls make sense only on block devices */ 343 if (minor(dev) == 0) 344 return ENOTTY; 345 346 switch(cmd) { 347 case DIOCGPART: 348 if ((dmv = dev->si_drv1) == NULL) 349 return ENODEV; 350 if (dmv->diskp->d_info.d_media_blksize == 0) { 351 return ENOTSUP; 352 } else { 353 u_int64_t size; 354 struct partinfo *dpart = data; 355 bzero(dpart, sizeof(*dpart)); 356 357 size = dm_table_size(&dmv->table_head); 358 dpart->media_offset = 0; 359 dpart->media_size = size * DEV_BSIZE; 360 dpart->media_blocks = size; 361 dpart->media_blksize = DEV_BSIZE; 362 dpart->fstype = FS_BSDFFS; 363 } 364 break; 365 366 default: 367 dmdebug("unknown disk_ioctl called\n"); 368 return ENOTTY; 369 break; /* NOT REACHED */ 370 } 371 372 return 0; 373 } 374 375 /* 376 * Do all IO operations on dm logical devices. 377 */ 378 static int 379 dmstrategy(struct dev_strategy_args *ap) 380 { 381 cdev_t dev = ap->a_head.a_dev; 382 struct bio *bio = ap->a_bio; 383 struct buf *bp = bio->bio_buf; 384 int bypass; 385 386 dm_dev_t *dmv; 387 dm_table_t *tbl; 388 dm_table_entry_t *table_en; 389 struct buf *nestbuf; 390 391 uint64_t buf_start, buf_len, issued_len; 392 uint64_t table_start, table_end; 393 uint64_t start, end; 394 395 buf_start = bio->bio_offset; 396 buf_len = bp->b_bcount; 397 398 tbl = NULL; 399 400 table_end = 0; 401 issued_len = 0; 402 403 dmv = dev->si_drv1; 404 405 switch(bp->b_cmd) { 406 case BUF_CMD_READ: 407 case BUF_CMD_WRITE: 408 case BUF_CMD_FREEBLKS: 409 bypass = 0; 410 break; 411 case BUF_CMD_FLUSH: 412 bypass = 1; 413 KKASSERT(buf_len == 0); 414 break; 415 default: 416 bp->b_error = EIO; 417 bp->b_resid = bp->b_bcount; 418 biodone(bio); 419 return 0; 420 } 421 422 if (bypass == 0 && 423 bounds_check_with_mediasize(bio, DEV_BSIZE, 424 dm_table_size(&dmv->table_head)) <= 0) { 425 bp->b_resid = bp->b_bcount; 426 biodone(bio); 427 return 0; 428 } 429 430 /* Select active table */ 431 tbl = dm_table_get_entry(&dmv->table_head, DM_TABLE_ACTIVE); 432 433 nestiobuf_init(bio); 434 devstat_start_transaction(&dmv->stats); 435 436 /* 437 * Find out what tables I want to select. 438 */ 439 TAILQ_FOREACH(table_en, tbl, next) { 440 /* 441 * I need need number of bytes not blocks. 442 */ 443 table_start = table_en->start * DEV_BSIZE; 444 table_end = table_start + table_en->length * DEV_BSIZE; 445 446 /* 447 * Calculate the start and end 448 */ 449 start = MAX(table_start, buf_start); 450 end = MIN(table_end, buf_start + buf_len); 451 452 if (dm_debug_level) { 453 kprintf("----------------------------------------\n"); 454 kprintf("table_start %010" PRIu64", table_end %010" 455 PRIu64 "\n", table_start, table_end); 456 kprintf("buf_start %010" PRIu64", buf_len %010" 457 PRIu64"\n", buf_start, buf_len); 458 kprintf("start-buf_start %010"PRIu64", end %010" 459 PRIu64"\n", start - buf_start, end); 460 kprintf("start %010" PRIu64", end %010" 461 PRIu64"\n", start, end); 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 * Find out what tables I want to select. 522 */ 523 TAILQ_FOREACH(table_en, tbl, next) { 524 /* 525 * I need need number of bytes not blocks. 526 */ 527 table_start = table_en->start * DEV_BSIZE; 528 table_end = table_start + table_en->length * DEV_BSIZE; 529 530 /* 531 * Calculate the start and end 532 */ 533 start = MAX(table_start, buf_start); 534 end = MIN(table_end, buf_start + buf_len); 535 536 if (ap->a_length == 0) { 537 if (table_en->target->dump == NULL) { 538 error = ENXIO; 539 goto out; 540 } 541 542 table_en->target->dump(table_en, NULL, 0, 0); 543 } else if (start < end) { 544 data_offset = start - buf_start; 545 offset = start - table_start; 546 length = end - start; 547 548 if (table_en->target->dump == NULL) { 549 error = ENXIO; 550 goto out; 551 } 552 553 table_en->target->dump(table_en, 554 (char *)ap->a_virtual + data_offset, 555 length, offset); 556 557 issued_len += end - start; 558 } 559 } 560 561 if (issued_len < buf_len) 562 error = EINVAL; 563 564 out: 565 dm_table_release(&dmv->table_head, DM_TABLE_ACTIVE); 566 567 return error; 568 } 569 570 static int 571 dmsize(struct dev_psize_args *ap) 572 { 573 cdev_t dev = ap->a_head.a_dev; 574 dm_dev_t *dmv; 575 uint64_t size; 576 577 size = 0; 578 579 if ((dmv = dev->si_drv1) == NULL) 580 return ENXIO; 581 582 size = dm_table_size(&dmv->table_head); 583 ap->a_result = (int64_t)size; 584 585 return 0; 586 } 587 588 #if 0 589 static void 590 dmminphys(struct buf *bp) 591 { 592 593 bp->b_bcount = MIN(bp->b_bcount, MAXPHYS); 594 } 595 #endif 596 597 void 598 dmsetdiskinfo(struct disk *disk, dm_table_head_t *head) 599 { 600 struct disk_info info; 601 uint64_t dmp_size; 602 603 dmp_size = dm_table_size(head); 604 605 bzero(&info, sizeof(struct disk_info)); 606 info.d_media_blksize = DEV_BSIZE; 607 info.d_media_blocks = dmp_size; 608 #if 0 609 /* this is set by disk_setdiskinfo */ 610 info.d_media_size = dmp_size * DEV_BSIZE; 611 #endif 612 info.d_dsflags = DSO_MBRQUIET | DSO_DEVICEMAPPER | DSO_RAWPSIZE; 613 614 info.d_secpertrack = 32; 615 info.d_nheads = 64; 616 info.d_secpercyl = info.d_secpertrack * info.d_nheads; 617 info.d_ncylinders = dmp_size / info.d_secpercyl; 618 619 disk_setdiskinfo(disk, &info); 620 } 621 622 /* 623 * Transform char s to uint64_t offset number. 624 */ 625 uint64_t 626 atoi64(const char *s) 627 { 628 uint64_t n; 629 n = 0; 630 631 while (*s != '\0') { 632 if (!isdigit(*s)) 633 break; 634 635 n = (10 * n) + (*s - '0'); 636 s++; 637 } 638 639 return n; 640 } 641 642 char * 643 dm_alloc_string(int len) 644 { 645 if (len <= 0) 646 len = DM_MAX_PARAMS_SIZE; 647 return kmalloc(len, M_DM, M_WAITOK | M_ZERO); 648 } 649 650 void 651 dm_builtin_init(void *arg) 652 { 653 modeventhand_t evh = (modeventhand_t)arg; 654 655 KKASSERT(evh != NULL); 656 evh(NULL, MOD_LOAD, NULL); 657 } 658 659 void 660 dm_builtin_uninit(void *arg) 661 { 662 modeventhand_t evh = (modeventhand_t)arg; 663 664 KKASSERT(evh != NULL); 665 evh(NULL, MOD_UNLOAD, NULL); 666 } 667 668 TUNABLE_INT("debug.dm_debug", &dm_debug_level); 669 SYSCTL_INT(_debug, OID_AUTO, dm_debug, CTLFLAG_RW, &dm_debug_level, 670 0, "Enable device mapper debugging"); 671