| 1 | /*- |
| 2 | * Copyright (c) 1997, 1998 |
| 3 | * Nan Yang Computer Services Limited. All rights reserved. |
| 4 | * |
| 5 | * Written by Greg Lehey |
| 6 | * |
| 7 | * This software is distributed under the so-called ``Berkeley |
| 8 | * License'': |
| 9 | * |
| 10 | * Redistribution and use in source and binary forms, with or without |
| 11 | * modification, are permitted provided that the following conditions |
| 12 | * are met: |
| 13 | * 1. Redistributions of source code must retain the above copyright |
| 14 | * notice, this list of conditions and the following disclaimer. |
| 15 | * 2. Redistributions in binary form must reproduce the above copyright |
| 16 | * notice, this list of conditions and the following disclaimer in the |
| 17 | * documentation and/or other materials provided with the distribution. |
| 18 | * 3. All advertising materials mentioning features or use of this software |
| 19 | * must display the following acknowledgement: |
| 20 | * This product includes software developed by Nan Yang Computer |
| 21 | * Services Limited. |
| 22 | * 4. Neither the name of the Company nor the names of its contributors |
| 23 | * may be used to endorse or promote products derived from this software |
| 24 | * without specific prior written permission. |
| 25 | * |
| 26 | * This software is provided ``as is'', and any express or implied |
| 27 | * warranties, including, but not limited to, the implied warranties of |
| 28 | * merchantability and fitness for a particular purpose are disclaimed. |
| 29 | * In no event shall the company or contributors be liable for any |
| 30 | * direct, indirect, incidental, special, exemplary, or consequential |
| 31 | * damages (including, but not limited to, procurement of substitute |
| 32 | * goods or services; loss of use, data, or profits; or business |
| 33 | * interruption) however caused and on any theory of liability, whether |
| 34 | * in contract, strict liability, or tort (including negligence or |
| 35 | * otherwise) arising in any way out of the use of this software, even if |
| 36 | * advised of the possibility of such damage. |
| 37 | * |
| 38 | * $Id: vinum.c,v 1.33 2001/01/09 06:19:15 grog Exp grog $ |
| 39 | * $FreeBSD: src/sys/dev/vinum/vinum.c,v 1.38.2.3 2003/01/07 12:14:16 joerg Exp $ |
| 40 | * $DragonFly: src/sys/dev/raid/vinum/vinum.c,v 1.11 2004/05/19 22:52:48 dillon Exp $ |
| 41 | */ |
| 42 | |
| 43 | #define STATIC static /* nothing while we're testing XXX */ |
| 44 | |
| 45 | #include "vinumhdr.h" |
| 46 | #include <sys/sysproto.h> /* for sync(2) */ |
| 47 | #include <sys/devicestat.h> |
| 48 | #ifdef VINUMDEBUG |
| 49 | #include <sys/reboot.h> |
| 50 | int debug = 0; |
| 51 | extern int total_malloced; |
| 52 | extern int malloccount; |
| 53 | extern struct mc malloced[]; |
| 54 | #endif |
| 55 | #include "request.h" |
| 56 | |
| 57 | struct cdevsw vinum_cdevsw = |
| 58 | { |
| 59 | /* name */ "vinum", |
| 60 | /* cmaj */ VINUM_CDEV_MAJOR, |
| 61 | /* flags */ D_DISK, |
| 62 | /* port */ NULL, |
| 63 | /* clone */ NULL, |
| 64 | |
| 65 | vinumopen, vinumclose, physread, physwrite, |
| 66 | vinumioctl, seltrue, nommap, vinumstrategy, |
| 67 | vinumdump, vinumsize, |
| 68 | }; |
| 69 | |
| 70 | /* Called by main() during pseudo-device attachment. */ |
| 71 | STATIC void vinumattach(void *); |
| 72 | |
| 73 | STATIC int vinum_modevent(module_t mod, modeventtype_t type, void *unused); |
| 74 | |
| 75 | struct _vinum_conf vinum_conf; /* configuration information */ |
| 76 | |
| 77 | /* |
| 78 | * Called by main() during pseudo-device attachment. All we need |
| 79 | * to do is allocate enough space for devices to be configured later, and |
| 80 | * add devsw entries. |
| 81 | */ |
| 82 | void |
| 83 | vinumattach(void *dummy) |
| 84 | { |
| 85 | char *cp, *cp1, *cp2, **drives; |
| 86 | int i, rv; |
| 87 | struct volume *vol; |
| 88 | |
| 89 | /* modload should prevent multiple loads, so this is worth a panic */ |
| 90 | if ((vinum_conf.flags & VF_LOADED) != 0) |
| 91 | panic("vinum: already loaded"); |
| 92 | |
| 93 | log(LOG_INFO, "vinum: loaded\n"); |
| 94 | vinum_conf.flags |= VF_LOADED; /* we're loaded now */ |
| 95 | |
| 96 | daemonq = NULL; /* initialize daemon's work queue */ |
| 97 | dqend = NULL; |
| 98 | |
| 99 | cdevsw_add(&vinum_cdevsw, 0, 0); /* add the cdevsw entry */ |
| 100 | |
| 101 | /* allocate space: drives... */ |
| 102 | DRIVE = (struct drive *) Malloc(sizeof(struct drive) * INITIAL_DRIVES); |
| 103 | CHECKALLOC(DRIVE, "vinum: no memory\n"); |
| 104 | bzero(DRIVE, sizeof(struct drive) * INITIAL_DRIVES); |
| 105 | vinum_conf.drives_allocated = INITIAL_DRIVES; /* number of drive slots allocated */ |
| 106 | vinum_conf.drives_used = 0; /* and number in use */ |
| 107 | |
| 108 | /* volumes, ... */ |
| 109 | VOL = (struct volume *) Malloc(sizeof(struct volume) * INITIAL_VOLUMES); |
| 110 | CHECKALLOC(VOL, "vinum: no memory\n"); |
| 111 | bzero(VOL, sizeof(struct volume) * INITIAL_VOLUMES); |
| 112 | vinum_conf.volumes_allocated = INITIAL_VOLUMES; /* number of volume slots allocated */ |
| 113 | vinum_conf.volumes_used = 0; /* and number in use */ |
| 114 | |
| 115 | /* plexes, ... */ |
| 116 | PLEX = (struct plex *) Malloc(sizeof(struct plex) * INITIAL_PLEXES); |
| 117 | CHECKALLOC(PLEX, "vinum: no memory\n"); |
| 118 | bzero(PLEX, sizeof(struct plex) * INITIAL_PLEXES); |
| 119 | vinum_conf.plexes_allocated = INITIAL_PLEXES; /* number of plex slots allocated */ |
| 120 | vinum_conf.plexes_used = 0; /* and number in use */ |
| 121 | |
| 122 | /* and subdisks */ |
| 123 | SD = (struct sd *) Malloc(sizeof(struct sd) * INITIAL_SUBDISKS); |
| 124 | CHECKALLOC(SD, "vinum: no memory\n"); |
| 125 | bzero(SD, sizeof(struct sd) * INITIAL_SUBDISKS); |
| 126 | vinum_conf.subdisks_allocated = INITIAL_SUBDISKS; /* number of sd slots allocated */ |
| 127 | vinum_conf.subdisks_used = 0; /* and number in use */ |
| 128 | |
| 129 | /* |
| 130 | * See if the loader has passed us a disk to |
| 131 | * read the initial configuration from. |
| 132 | */ |
| 133 | if ((cp = getenv("vinum.drives")) != NULL) { |
| 134 | for (cp1 = cp, i = 0, drives = 0; *cp1 != '\0'; i++) { |
| 135 | cp2 = cp1; |
| 136 | while (*cp1 != '\0' && *cp1 != ',' && *cp1 != ' ') |
| 137 | cp1++; |
| 138 | if (*cp1 != '\0') |
| 139 | *cp1++ = '\0'; |
| 140 | drives = realloc(drives, (unsigned long)((i + 1) * sizeof(char *)), |
| 141 | M_TEMP, M_WAITOK); |
| 142 | drives[i] = cp2; |
| 143 | } |
| 144 | if (i == 0) |
| 145 | goto bailout; |
| 146 | rv = vinum_scandisk(drives, i); |
| 147 | if (rv) |
| 148 | log(LOG_NOTICE, "vinum_scandisk() returned %d", rv); |
| 149 | bailout: |
| 150 | free(drives, M_TEMP); |
| 151 | } |
| 152 | if ((cp = getenv("vinum.root")) != NULL) { |
| 153 | for (i = 0; i < vinum_conf.volumes_used; i++) { |
| 154 | vol = &vinum_conf.volume[i]; |
| 155 | if ((vol->state == volume_up) |
| 156 | && (strcmp (vol->name, cp) == 0) |
| 157 | ) { |
| 158 | rootdev = make_dev(&vinum_cdevsw, i, UID_ROOT, GID_OPERATOR, |
| 159 | 0640, "vinum"); |
| 160 | log(LOG_INFO, "vinum: using volume %s for root device\n", cp); |
| 161 | break; |
| 162 | } |
| 163 | } |
| 164 | } |
| 165 | } |
| 166 | |
| 167 | /* |
| 168 | * Check if we have anything open. If confopen is != 0, |
| 169 | * that goes for the super device as well, otherwise |
| 170 | * only for volumes. |
| 171 | * |
| 172 | * Return 0 if not inactive, 1 if inactive. |
| 173 | */ |
| 174 | int |
| 175 | vinum_inactive(int confopen) |
| 176 | { |
| 177 | int i; |
| 178 | int can_do = 1; /* assume we can do it */ |
| 179 | |
| 180 | if (confopen && (vinum_conf.flags & VF_OPEN)) /* open by vinum(8)? */ |
| 181 | return 0; /* can't do it while we're open */ |
| 182 | lock_config(); |
| 183 | for (i = 0; i < vinum_conf.volumes_allocated; i++) { |
| 184 | if ((VOL[i].state > volume_down) |
| 185 | && (VOL[i].flags & VF_OPEN)) { /* volume is open */ |
| 186 | can_do = 0; |
| 187 | break; |
| 188 | } |
| 189 | } |
| 190 | unlock_config(); |
| 191 | return can_do; |
| 192 | } |
| 193 | |
| 194 | /* |
| 195 | * Free all structures. |
| 196 | * If cleardrive is 0, save the configuration; otherwise |
| 197 | * remove the configuration from the drive. |
| 198 | * |
| 199 | * Before coming here, ensure that no volumes are open. |
| 200 | */ |
| 201 | void |
| 202 | free_vinum(int cleardrive) |
| 203 | { |
| 204 | int i; |
| 205 | int drives_allocated = vinum_conf.drives_allocated; |
| 206 | |
| 207 | if (DRIVE != NULL) { |
| 208 | if (cleardrive) { /* remove the vinum config */ |
| 209 | for (i = 0; i < drives_allocated; i++) |
| 210 | remove_drive(i); /* remove the drive */ |
| 211 | } else { /* keep the config */ |
| 212 | for (i = 0; i < drives_allocated; i++) |
| 213 | free_drive(&DRIVE[i]); /* close files and things */ |
| 214 | } |
| 215 | Free(DRIVE); |
| 216 | } |
| 217 | while ((vinum_conf.flags & (VF_STOPPING | VF_DAEMONOPEN)) |
| 218 | == (VF_STOPPING | VF_DAEMONOPEN)) { /* at least one daemon open, we're stopping */ |
| 219 | queue_daemon_request(daemonrq_return, (union daemoninfo) 0); /* stop the daemon */ |
| 220 | tsleep(&vinumclose, 0, "vstop", 1); /* and wait for it */ |
| 221 | } |
| 222 | if (SD != NULL) |
| 223 | Free(SD); |
| 224 | if (PLEX != NULL) { |
| 225 | for (i = 0; i < vinum_conf.plexes_allocated; i++) { |
| 226 | struct plex *plex = &vinum_conf.plex[i]; |
| 227 | |
| 228 | if (plex->state != plex_unallocated) { /* we have real data there */ |
| 229 | if (plex->sdnos) |
| 230 | Free(plex->sdnos); |
| 231 | } |
| 232 | } |
| 233 | Free(PLEX); |
| 234 | } |
| 235 | if (VOL != NULL) |
| 236 | Free(VOL); |
| 237 | bzero(&vinum_conf, sizeof(vinum_conf)); |
| 238 | } |
| 239 | |
| 240 | STATIC int |
| 241 | vinum_modevent(module_t mod, modeventtype_t type, void *unused) |
| 242 | { |
| 243 | switch (type) { |
| 244 | case MOD_LOAD: |
| 245 | vinumattach(NULL); |
| 246 | return 0; /* OK */ |
| 247 | case MOD_UNLOAD: |
| 248 | if (!vinum_inactive(1)) /* is anything open? */ |
| 249 | return EBUSY; /* yes, we can't do it */ |
| 250 | vinum_conf.flags |= VF_STOPPING; /* note that we want to stop */ |
| 251 | sync(NULL); /* write out buffers */ |
| 252 | free_vinum(0); /* clean up */ |
| 253 | #ifdef VINUMDEBUG |
| 254 | if (total_malloced) { |
| 255 | int i; |
| 256 | #ifdef INVARIANTS |
| 257 | int *poke; |
| 258 | #endif |
| 259 | |
| 260 | for (i = 0; i < malloccount; i++) { |
| 261 | if (debug & DEBUG_WARNINGS) /* want to hear about them */ |
| 262 | log(LOG_WARNING, |
| 263 | "vinum: exiting with %d bytes malloced from %s:%d\n", |
| 264 | malloced[i].size, |
| 265 | malloced[i].file, |
| 266 | malloced[i].line); |
| 267 | #ifdef INVARIANTS |
| 268 | poke = &((int *) malloced[i].address) |
| 269 | [malloced[i].size / (2 * sizeof(int))]; /* middle of the area */ |
| 270 | if (*poke == 0xdeadc0de) /* already freed */ |
| 271 | log(LOG_ERR, |
| 272 | "vinum: exiting with malloc table inconsistency at %p from %s:%d\n", |
| 273 | malloced[i].address, |
| 274 | malloced[i].file, |
| 275 | malloced[i].line); |
| 276 | #endif |
| 277 | Free(malloced[i].address); |
| 278 | } |
| 279 | } |
| 280 | #endif |
| 281 | cdevsw_remove(&vinum_cdevsw, 0, 0); |
| 282 | log(LOG_INFO, "vinum: unloaded\n"); /* tell the world */ |
| 283 | return 0; |
| 284 | default: |
| 285 | break; |
| 286 | } |
| 287 | return 0; |
| 288 | } |
| 289 | |
| 290 | moduledata_t vinum_mod = |
| 291 | { |
| 292 | "vinum", |
| 293 | (modeventhand_t) vinum_modevent, |
| 294 | 0 |
| 295 | }; |
| 296 | DECLARE_MODULE(vinum, vinum_mod, SI_SUB_RAID, SI_ORDER_MIDDLE); |
| 297 | |
| 298 | /* ARGSUSED */ |
| 299 | /* Open a vinum object */ |
| 300 | int |
| 301 | vinumopen(dev_t dev, int flags, int fmt, d_thread_t *td) |
| 302 | { |
| 303 | int error; |
| 304 | unsigned int index; |
| 305 | struct volume *vol; |
| 306 | struct plex *plex; |
| 307 | struct sd *sd; |
| 308 | int devminor; /* minor number */ |
| 309 | |
| 310 | devminor = minor(dev); |
| 311 | error = 0; |
| 312 | /* First, decide what we're looking at */ |
| 313 | switch (DEVTYPE(dev)) { |
| 314 | case VINUM_VOLUME_TYPE: |
| 315 | index = Volno(dev); |
| 316 | if (index >= vinum_conf.volumes_allocated) |
| 317 | return ENXIO; /* no such device */ |
| 318 | vol = &VOL[index]; |
| 319 | |
| 320 | switch (vol->state) { |
| 321 | case volume_unallocated: |
| 322 | case volume_uninit: |
| 323 | return ENXIO; |
| 324 | |
| 325 | case volume_up: |
| 326 | vol->flags |= VF_OPEN; /* note we're open */ |
| 327 | return 0; |
| 328 | |
| 329 | case volume_down: |
| 330 | return EIO; |
| 331 | |
| 332 | default: |
| 333 | return EINVAL; |
| 334 | } |
| 335 | |
| 336 | case VINUM_PLEX_TYPE: |
| 337 | if (Volno(dev) >= vinum_conf.volumes_allocated) |
| 338 | return ENXIO; |
| 339 | /* FALLTHROUGH */ |
| 340 | |
| 341 | case VINUM_RAWPLEX_TYPE: |
| 342 | index = Plexno(dev); /* get plex index in vinum_conf */ |
| 343 | if (index >= vinum_conf.plexes_allocated) |
| 344 | return ENXIO; /* no such device */ |
| 345 | plex = &PLEX[index]; |
| 346 | |
| 347 | switch (plex->state) { |
| 348 | case plex_referenced: |
| 349 | case plex_unallocated: |
| 350 | return EINVAL; |
| 351 | |
| 352 | default: |
| 353 | plex->flags |= VF_OPEN; /* note we're open */ |
| 354 | return 0; |
| 355 | } |
| 356 | |
| 357 | case VINUM_SD_TYPE: |
| 358 | if ((Volno(dev) >= vinum_conf.volumes_allocated) /* no such volume */ |
| 359 | ||(Plexno(dev) >= vinum_conf.plexes_allocated)) /* or no such plex */ |
| 360 | return ENXIO; /* no such device */ |
| 361 | |
| 362 | /* FALLTHROUGH */ |
| 363 | |
| 364 | case VINUM_RAWSD_TYPE: |
| 365 | index = Sdno(dev); /* get the subdisk number */ |
| 366 | if ((index >= vinum_conf.subdisks_allocated) /* not a valid SD entry */ |
| 367 | ||(SD[index].state < sd_init)) /* or SD is not real */ |
| 368 | return ENXIO; /* no such device */ |
| 369 | sd = &SD[index]; |
| 370 | |
| 371 | /* |
| 372 | * Opening a subdisk is always a special operation, so we |
| 373 | * ignore the state as long as it represents a real subdisk |
| 374 | */ |
| 375 | switch (sd->state) { |
| 376 | case sd_unallocated: |
| 377 | case sd_uninit: |
| 378 | return EINVAL; |
| 379 | |
| 380 | default: |
| 381 | sd->flags |= VF_OPEN; /* note we're open */ |
| 382 | return 0; |
| 383 | } |
| 384 | |
| 385 | case VINUM_SUPERDEV_TYPE: |
| 386 | error = suser(td); /* are we root? */ |
| 387 | if (error == 0) { /* yes, can do */ |
| 388 | if (devminor == VINUM_DAEMON_DEV) /* daemon device */ |
| 389 | vinum_conf.flags |= VF_DAEMONOPEN; /* we're open */ |
| 390 | else if (devminor == VINUM_SUPERDEV) |
| 391 | vinum_conf.flags |= VF_OPEN; /* we're open */ |
| 392 | else |
| 393 | error = ENODEV; /* nothing, maybe a debug mismatch */ |
| 394 | } |
| 395 | return error; |
| 396 | |
| 397 | /* Vinum drives are disks. We already have a disk |
| 398 | * driver, so don't handle them here */ |
| 399 | case VINUM_DRIVE_TYPE: |
| 400 | default: |
| 401 | return ENODEV; /* don't know what to do with these */ |
| 402 | } |
| 403 | } |
| 404 | |
| 405 | /* ARGSUSED */ |
| 406 | int |
| 407 | vinumclose(dev_t dev, int flags, int fmt, d_thread_t *td) |
| 408 | { |
| 409 | unsigned int index; |
| 410 | struct volume *vol; |
| 411 | int devminor; |
| 412 | |
| 413 | devminor = minor(dev); |
| 414 | index = Volno(dev); |
| 415 | /* First, decide what we're looking at */ |
| 416 | switch (DEVTYPE(dev)) { |
| 417 | case VINUM_VOLUME_TYPE: |
| 418 | if (index >= vinum_conf.volumes_allocated) |
| 419 | return ENXIO; /* no such device */ |
| 420 | vol = &VOL[index]; |
| 421 | |
| 422 | switch (vol->state) { |
| 423 | case volume_unallocated: |
| 424 | case volume_uninit: |
| 425 | return ENXIO; |
| 426 | |
| 427 | case volume_up: |
| 428 | vol->flags &= ~VF_OPEN; /* reset our flags */ |
| 429 | return 0; |
| 430 | |
| 431 | case volume_down: |
| 432 | return EIO; |
| 433 | |
| 434 | default: |
| 435 | return EINVAL; |
| 436 | } |
| 437 | |
| 438 | case VINUM_PLEX_TYPE: |
| 439 | if (Volno(dev) >= vinum_conf.volumes_allocated) |
| 440 | return ENXIO; |
| 441 | /* FALLTHROUGH */ |
| 442 | |
| 443 | case VINUM_RAWPLEX_TYPE: |
| 444 | index = Plexno(dev); /* get plex index in vinum_conf */ |
| 445 | if (index >= vinum_conf.plexes_allocated) |
| 446 | return ENXIO; /* no such device */ |
| 447 | PLEX[index].flags &= ~VF_OPEN; /* reset our flags */ |
| 448 | return 0; |
| 449 | |
| 450 | case VINUM_SD_TYPE: |
| 451 | if ((Volno(dev) >= vinum_conf.volumes_allocated) || /* no such volume */ |
| 452 | (Plexno(dev) >= vinum_conf.plexes_allocated)) /* or no such plex */ |
| 453 | return ENXIO; /* no such device */ |
| 454 | /* FALLTHROUGH */ |
| 455 | |
| 456 | case VINUM_RAWSD_TYPE: |
| 457 | index = Sdno(dev); /* get the subdisk number */ |
| 458 | if (index >= vinum_conf.subdisks_allocated) |
| 459 | return ENXIO; /* no such device */ |
| 460 | SD[index].flags &= ~VF_OPEN; /* reset our flags */ |
| 461 | return 0; |
| 462 | |
| 463 | case VINUM_SUPERDEV_TYPE: |
| 464 | /* |
| 465 | * don't worry about whether we're root: |
| 466 | * nobody else would get this far. |
| 467 | */ |
| 468 | if (devminor == VINUM_SUPERDEV) /* normal superdev */ |
| 469 | vinum_conf.flags &= ~VF_OPEN; /* no longer open */ |
| 470 | else if (devminor == VINUM_DAEMON_DEV) { /* the daemon device */ |
| 471 | vinum_conf.flags &= ~VF_DAEMONOPEN; /* no longer open */ |
| 472 | if (vinum_conf.flags & VF_STOPPING) /* we're stopping, */ |
| 473 | wakeup(&vinumclose); /* we can continue stopping now */ |
| 474 | } |
| 475 | return 0; |
| 476 | |
| 477 | case VINUM_DRIVE_TYPE: |
| 478 | default: |
| 479 | return ENODEV; /* don't know what to do with these */ |
| 480 | } |
| 481 | } |
| 482 | |
| 483 | /* size routine */ |
| 484 | int |
| 485 | vinumsize(dev_t dev) |
| 486 | { |
| 487 | struct volume *vol; |
| 488 | int size; |
| 489 | |
| 490 | vol = &VOL[Volno(dev)]; |
| 491 | |
| 492 | if (vol->state == volume_up) |
| 493 | size = vol->size; |
| 494 | else |
| 495 | return 0; /* err on the size of conservatism */ |
| 496 | |
| 497 | return size; |
| 498 | } |
| 499 | |
| 500 | int |
| 501 | vinumdump(dev_t dev, u_int count, u_int blkno, u_int secsize) |
| 502 | { |
| 503 | /* Not implemented. */ |
| 504 | return ENXIO; |
| 505 | } |
| 506 | |
| 507 | /* Local Variables: */ |
| 508 | /* fill-column: 50 */ |
| 509 | /* End: */ |