drivers/md/raid0.c

   1 /*
   2    raid0.c : Multiple Devices driver for Linux
   3              Copyright (C) 1994-96 Marc ZYNGIER
   4              <zyngier@ufr-info-p7.ibp.fr> or
   5              <maz@gloups.fdn.fr>
   6              Copyright (C) 1999, 2000 Ingo Molnar, Red Hat
   7
   8    RAID-0 management functions.
   9
  10    This program is free software; you can redistribute it and/or modify
  11    it under the terms of the GNU General Public License as published by
  12    the Free Software Foundation; either version 2, or (at your option)
  13    any later version.
  14
  15    You should have received a copy of the GNU General Public License
  16    (for example /usr/src/linux/COPYING); if not, write to the Free
  17    Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  18 */
  19
  20 #include <linux/blkdev.h>
  21 #include <linux/seq_file.h>
  22 #include <linux/module.h>
  23 #include <linux/slab.h>
  24 #include <trace/events/block.h>
  25 #include "md.h"
  26 #include "raid0.h"
  27 #include "raid5.h"
  28
  29 static int raid0_congested(struct mddev *mddev, int bits)
  30 {
  31         struct r0conf *conf = mddev->private;
  32         struct md_rdev **devlist = conf->devlist;
  33         int raid_disks = conf->strip_zone[0].nb_dev;
  34         int i, ret = 0;
  35
  36         for (i = 0; i < raid_disks && !ret ; i++) {
  37                 struct request_queue *q = bdev_get_queue(devlist[i]->bdev);
  38
  39                 ret |= bdi_congested(&q->backing_dev_info, bits);
  40         }
  41         return ret;
  42 }
  43
  44 /*
  45  * inform the user of the raid configuration
  46 */
  47 static void dump_zones(struct mddev *mddev)
  48 {
  49         int j, k;
  50         sector_t zone_size = 0;
  51         sector_t zone_start = 0;
  52         char b[BDEVNAME_SIZE];
  53         struct r0conf *conf = mddev->private;
  54         int raid_disks = conf->strip_zone[0].nb_dev;
  55         pr_debug("md: RAID0 configuration for %s - %d zone%s\n",
  56                  mdname(mddev),
  57                  conf->nr_strip_zones, conf->nr_strip_zones==1?"":"s");
  58         for (j = 0; j < conf->nr_strip_zones; j++) {
  59                 char line[200];
  60                 int len = 0;
  61
  62                 for (k = 0; k < conf->strip_zone[j].nb_dev; k++)
  63                         len += snprintf(line+len, 200-len, "%s%s", k?"/":"",
  64                                         bdevname(conf->devlist[j*raid_disks
  65                                                                + k]->bdev, b));
  66                 pr_debug("md: zone%d=[%s]\n", j, line);
  67
  68                 zone_size  = conf->strip_zone[j].zone_end - zone_start;
  69                 pr_debug("      zone-offset=%10lluKB, device-offset=%10lluKB, size=%10lluKB\n",
  70                         (unsigned long long)zone_start>>1,
  71                         (unsigned long long)conf->strip_zone[j].dev_start>>1,
  72                         (unsigned long long)zone_size>>1);
  73                 zone_start = conf->strip_zone[j].zone_end;
  74         }
  75 }
  76
  77 static int create_strip_zones(struct mddev *mddev, struct r0conf **private_conf)
  78 {
  79         int i, c, err;
  80         sector_t curr_zone_end, sectors;
  81         struct md_rdev *smallest, *rdev1, *rdev2, *rdev, **dev;
  82         struct strip_zone *zone;
  83         int cnt;
  84         char b[BDEVNAME_SIZE];
  85         char b2[BDEVNAME_SIZE];
  86         struct r0conf *conf = kzalloc(sizeof(*conf), GFP_KERNEL);
  87         unsigned short blksize = 512;
  88
  89         *private_conf = ERR_PTR(-ENOMEM);
  90         if (!conf)
  91                 return -ENOMEM;
  92         rdev_for_each(rdev1, mddev) {
  93                 pr_debug("md/raid0:%s: looking at %s\n",
  94                          mdname(mddev),
  95                          bdevname(rdev1->bdev, b));
  96                 c = 0;
  97
  98                 /* round size to chunk_size */
  99                 sectors = rdev1->sectors;
 100                 sector_div(sectors, mddev->chunk_sectors);
 101                 rdev1->sectors = sectors * mddev->chunk_sectors;
 102
 103                 blksize = max(blksize, queue_logical_block_size(
 104                                       rdev1->bdev->bd_disk->queue));
 105
 106                 rdev_for_each(rdev2, mddev) {
 107                         pr_debug("md/raid0:%s:   comparing %s(%llu)"
 108                                  " with %s(%llu)\n",
 109                                  mdname(mddev),
 110                                  bdevname(rdev1->bdev,b),
 111                                  (unsigned long long)rdev1->sectors,
 112                                  bdevname(rdev2->bdev,b2),
 113                                  (unsigned long long)rdev2->sectors);
 114                         if (rdev2 == rdev1) {
 115                                 pr_debug("md/raid0:%s:   END\n",
 116                                          mdname(mddev));
 117                                 break;
 118                         }
 119                         if (rdev2->sectors == rdev1->sectors) {
 120                                 /*
 121                                  * Not unique, don't count it as a new
 122                                  * group
 123                                  */
 124                                 pr_debug("md/raid0:%s:   EQUAL\n",
 125                                          mdname(mddev));
 126                                 c = 1;
 127                                 break;
 128                         }
 129                         pr_debug("md/raid0:%s:   NOT EQUAL\n",
 130                                  mdname(mddev));
 131                 }
 132                 if (!c) {
 133                         pr_debug("md/raid0:%s:   ==> UNIQUE\n",
 134                                  mdname(mddev));
 135                         conf->nr_strip_zones++;
 136                         pr_debug("md/raid0:%s: %d zones\n",
 137                                  mdname(mddev), conf->nr_strip_zones);
 138                 }
 139         }
 140         pr_debug("md/raid0:%s: FINAL %d zones\n",
 141                  mdname(mddev), conf->nr_strip_zones);
 142         /*
 143          * now since we have the hard sector sizes, we can make sure
 144          * chunk size is a multiple of that sector size
 145          */
 146         if ((mddev->chunk_sectors << 9) % blksize) {
 147                 pr_warn("md/raid0:%s: chunk_size of %d not multiple of block size %d\n",
 148                         mdname(mddev),
 149                         mddev->chunk_sectors << 9, blksize);
 150                 err = -EINVAL;
 151                 goto abort;
 152         }
 153
 154         err = -ENOMEM;
 155         conf->strip_zone = kzalloc(sizeof(struct strip_zone)*
 156                                 conf->nr_strip_zones, GFP_KERNEL);
 157         if (!conf->strip_zone)
 158                 goto abort;
 159         conf->devlist = kzalloc(sizeof(struct md_rdev*)*
 160                                 conf->nr_strip_zones*mddev->raid_disks,
 161                                 GFP_KERNEL);
 162         if (!conf->devlist)
 163                 goto abort;
 164
 165         /* The first zone must contain all devices, so here we check that
 166          * there is a proper alignment of slots to devices and find them all
 167          */
 168         zone = &conf->strip_zone[0];
 169         cnt = 0;
 170         smallest = NULL;
 171         dev = conf->devlist;
 172         err = -EINVAL;
 173         rdev_for_each(rdev1, mddev) {
 174                 int j = rdev1->raid_disk;
 175
 176                 if (mddev->level == 10) {
 177                         /* taking over a raid10-n2 array */
 178                         j /= 2;
 179                         rdev1->new_raid_disk = j;
 180                 }
 181
 182                 if (mddev->level == 1) {
 183                         /* taiking over a raid1 array-
 184                          * we have only one active disk
 185                          */
 186                         j = 0;
 187                         rdev1->new_raid_disk = j;
 188                 }
 189
 190                 if (j < 0) {
 191                         pr_warn("md/raid0:%s: remove inactive devices before converting to RAID0\n",
 192                                 mdname(mddev));
 193                         goto abort;
 194                 }
 195                 if (j >= mddev->raid_disks) {
 196                         pr_warn("md/raid0:%s: bad disk number %d - aborting!\n",
 197                                 mdname(mddev), j);
 198                         goto abort;
 199                 }
 200                 if (dev[j]) {
 201                         pr_warn("md/raid0:%s: multiple devices for %d - aborting!\n",
 202                                 mdname(mddev), j);
 203                         goto abort;
 204                 }
 205                 dev[j] = rdev1;
 206
 207                 if (!smallest || (rdev1->sectors < smallest->sectors))
 208                         smallest = rdev1;
 209                 cnt++;
 210         }
 211         if (cnt != mddev->raid_disks) {
 212                 pr_warn("md/raid0:%s: too few disks (%d of %d) - aborting!\n",
 213                         mdname(mddev), cnt, mddev->raid_disks);
 214                 goto abort;
 215         }
 216         zone->nb_dev = cnt;
 217         zone->zone_end = smallest->sectors * cnt;
 218
 219         curr_zone_end = zone->zone_end;
 220
 221         /* now do the other zones */
 222         for (i = 1; i < conf->nr_strip_zones; i++)
 223         {
 224                 int j;
 225
 226                 zone = conf->strip_zone + i;
 227                 dev = conf->devlist + i * mddev->raid_disks;
 228
 229                 pr_debug("md/raid0:%s: zone %d\n", mdname(mddev), i);
 230                 zone->dev_start = smallest->sectors;
 231                 smallest = NULL;
 232                 c = 0;
 233
 234                 for (j=0; j<cnt; j++) {
 235                         rdev = conf->devlist[j];
 236                         if (rdev->sectors <= zone->dev_start) {
 237                                 pr_debug("md/raid0:%s: checking %s ... nope\n",
 238                                          mdname(mddev),
 239                                          bdevname(rdev->bdev, b));
 240                                 continue;
 241                         }
 242                         pr_debug("md/raid0:%s: checking %s ..."
 243                                  " contained as device %d\n",
 244                                  mdname(mddev),
 245                                  bdevname(rdev->bdev, b), c);
 246                         dev[c] = rdev;
 247                         c++;
 248                         if (!smallest || rdev->sectors < smallest->sectors) {
 249                                 smallest = rdev;
 250                                 pr_debug("md/raid0:%s:  (%llu) is smallest!.\n",
 251                                          mdname(mddev),
 252                                          (unsigned long long)rdev->sectors);
 253                         }
 254                 }
 255
 256                 zone->nb_dev = c;
 257                 sectors = (smallest->sectors - zone->dev_start) * c;
 258                 pr_debug("md/raid0:%s: zone->nb_dev: %d, sectors: %llu\n",
 259                          mdname(mddev),
 260                          zone->nb_dev, (unsigned long long)sectors);
 261
 262                 curr_zone_end += sectors;
 263                 zone->zone_end = curr_zone_end;
 264
 265                 pr_debug("md/raid0:%s: current zone start: %llu\n",
 266                          mdname(mddev),
 267                          (unsigned long long)smallest->sectors);
 268         }
 269
 270         pr_debug("md/raid0:%s: done.\n", mdname(mddev));
 271         *private_conf = conf;
 272
 273         return 0;
 274 abort:
 275         kfree(conf->strip_zone);
 276         kfree(conf->devlist);
 277         kfree(conf);
 278         *private_conf = ERR_PTR(err);
 279         return err;
 280 }
 281
 282 /* Find the zone which holds a particular offset
 283  * Update *sectorp to be an offset in that zone
 284  */
 285 static struct strip_zone *find_zone(struct r0conf *conf,
 286                                     sector_t *sectorp)
 287 {
 288         int i;
 289         struct strip_zone *z = conf->strip_zone;
 290         sector_t sector = *sectorp;
 291
 292         for (i = 0; i < conf->nr_strip_zones; i++)
 293                 if (sector < z[i].zone_end) {
 294                         if (i)
 295                                 *sectorp = sector - z[i-1].zone_end;
 296                         return z + i;
 297                 }
 298         BUG();
 299 }
 300
 301 /*
 302  * remaps the bio to the target device. we separate two flows.
 303  * power 2 flow and a general flow for the sake of performance
 304 */
 305 static struct md_rdev *map_sector(struct mddev *mddev, struct strip_zone *zone,
 306                                 sector_t sector, sector_t *sector_offset)
 307 {
 308         unsigned int sect_in_chunk;
 309         sector_t chunk;
 310         struct r0conf *conf = mddev->private;
 311         int raid_disks = conf->strip_zone[0].nb_dev;
 312         unsigned int chunk_sects = mddev->chunk_sectors;
 313
 314         if (is_power_of_2(chunk_sects)) {
 315                 int chunksect_bits = ffz(~chunk_sects);
 316                 /* find the sector offset inside the chunk */
 317                 sect_in_chunk  = sector & (chunk_sects - 1);
 318                 sector >>= chunksect_bits;
 319                 /* chunk in zone */
 320                 chunk = *sector_offset;
 321                 /* quotient is the chunk in real device*/
 322                 sector_div(chunk, zone->nb_dev << chunksect_bits);
 323         } else{
 324                 sect_in_chunk = sector_div(sector, chunk_sects);
 325                 chunk = *sector_offset;
 326                 sector_div(chunk, chunk_sects * zone->nb_dev);
 327         }
 328         /*
 329         *  position the bio over the real device
 330         *  real sector = chunk in device + starting of zone
 331         *       + the position in the chunk
 332         */
 333         *sector_offset = (chunk * chunk_sects) + sect_in_chunk;
 334         return conf->devlist[(zone - conf->strip_zone)*raid_disks
 335                              + sector_div(sector, zone->nb_dev)];
 336 }
 337
 338 static sector_t raid0_size(struct mddev *mddev, sector_t sectors, int raid_disks)
 339 {
 340         sector_t array_sectors = 0;
 341         struct md_rdev *rdev;
 342
 343         WARN_ONCE(sectors || raid_disks,
 344                   "%s does not support generic reshape\n", __func__);
 345
 346         rdev_for_each(rdev, mddev)
 347                 array_sectors += (rdev->sectors &
 348                                   ~(sector_t)(mddev->chunk_sectors-1));
 349
 350         return array_sectors;
 351 }
 352
 353 static void raid0_free(struct mddev *mddev, void *priv);
 354
 355 static int raid0_run(struct mddev *mddev)
 356 {
 357         struct r0conf *conf;
 358         int ret;
 359
 360         if (mddev->chunk_sectors == 0) {
 361                 pr_warn("md/raid0:%s: chunk size must be set.\n", mdname(mddev));
 362                 return -EINVAL;
 363         }
 364         if (md_check_no_bitmap(mddev))
 365                 return -EINVAL;
 366
 367         /* if private is not null, we are here after takeover */
 368         if (mddev->private == NULL) {
 369                 ret = create_strip_zones(mddev, &conf);
 370                 if (ret < 0)
 371                         return ret;
 372                 mddev->private = conf;
 373         }
 374         conf = mddev->private;
 375         if (mddev->queue) {
 376                 struct md_rdev *rdev;
 377                 bool discard_supported = false;
 378
 379                 blk_queue_max_hw_sectors(mddev->queue, mddev->chunk_sectors);
 380                 blk_queue_max_write_same_sectors(mddev->queue, mddev->chunk_sectors);
 381                 blk_queue_max_discard_sectors(mddev->queue, mddev->chunk_sectors);
 382
 383                 blk_queue_io_min(mddev->queue, mddev->chunk_sectors << 9);
 384                 blk_queue_io_opt(mddev->queue,
 385                                  (mddev->chunk_sectors << 9) * mddev->raid_disks);
 386
 387                 rdev_for_each(rdev, mddev) {
 388                         disk_stack_limits(mddev->gendisk, rdev->bdev,
 389                                           rdev->data_offset << 9);
 390                         if (blk_queue_discard(bdev_get_queue(rdev->bdev)))
 391                                 discard_supported = true;
 392                 }
 393                 if (!discard_supported)
 394                         queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD, mddev->queue);
 395                 else
 396                         queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, mddev->queue);
 397         }
 398
 399         /* calculate array device size */
 400         md_set_array_sectors(mddev, raid0_size(mddev, 0, 0));
 401
 402         pr_debug("md/raid0:%s: md_size is %llu sectors.\n",
 403                  mdname(mddev),
 404                  (unsigned long long)mddev->array_sectors);
 405
 406         if (mddev->queue) {
 407                 /* calculate the max read-ahead size.
 408                  * For read-ahead of large files to be effective, we need to
 409                  * readahead at least twice a whole stripe. i.e. number of devices
 410                  * multiplied by chunk size times 2.
 411                  * If an individual device has an ra_pages greater than the
 412                  * chunk size, then we will not drive that device as hard as it
 413                  * wants.  We consider this a configuration error: a larger
 414                  * chunksize should be used in that case.
 415                  */
 416                 int stripe = mddev->raid_disks *
 417                         (mddev->chunk_sectors << 9) / PAGE_SIZE;
 418                 if (mddev->queue->backing_dev_info.ra_pages < 2* stripe)
 419                         mddev->queue->backing_dev_info.ra_pages = 2* stripe;
 420         }
 421
 422         dump_zones(mddev);
 423
 424         ret = md_integrity_register(mddev);
 425
 426         return ret;
 427 }
 428
 429 static void raid0_free(struct mddev *mddev, void *priv)
 430 {
 431         struct r0conf *conf = priv;
 432
 433         kfree(conf->strip_zone);
 434         kfree(conf->devlist);
 435         kfree(conf);
 436 }
 437
 438 /*
 439  * Is io distribute over 1 or more chunks ?
 440 */
 441 static inline int is_io_in_chunk_boundary(struct mddev *mddev,
 442                         unsigned int chunk_sects, struct bio *bio)
 443 {
 444         if (likely(is_power_of_2(chunk_sects))) {
 445                 return chunk_sects >=
 446                         ((bio->bi_iter.bi_sector & (chunk_sects-1))
 447                                         + bio_sectors(bio));
 448         } else{
 449                 sector_t sector = bio->bi_iter.bi_sector;
 450                 return chunk_sects >= (sector_div(sector, chunk_sects)
 451                                                 + bio_sectors(bio));
 452         }
 453 }
 454
 455 static void raid0_make_request(struct mddev *mddev, struct bio *bio)
 456 {
 457         struct strip_zone *zone;
 458         struct md_rdev *tmp_dev;
 459         struct bio *split;
 460
 461         if (unlikely(bio->bi_opf & REQ_PREFLUSH)) {
 462                 md_flush_request(mddev, bio);
 463                 return;
 464         }
 465
 466         do {
 467                 sector_t bio_sector = bio->bi_iter.bi_sector;
 468                 sector_t sector = bio_sector;
 469                 unsigned chunk_sects = mddev->chunk_sectors;
 470
 471                 unsigned sectors = chunk_sects -
 472                         (likely(is_power_of_2(chunk_sects))
 473                          ? (sector & (chunk_sects-1))
 474                          : sector_div(sector, chunk_sects));
 475
 476                 /* Restore due to sector_div */
 477                 sector = bio_sector;
 478
 479                 if (sectors < bio_sectors(bio)) {
 480                         split = bio_split(bio, sectors, GFP_NOIO, fs_bio_set);
 481                         bio_chain(split, bio);
 482                 } else {
 483                         split = bio;
 484                 }
 485
 486                 zone = find_zone(mddev->private, &sector);
 487                 tmp_dev = map_sector(mddev, zone, sector, &sector);
 488                 split->bi_bdev = tmp_dev->bdev;
 489                 split->bi_iter.bi_sector = sector + zone->dev_start +
 490                         tmp_dev->data_offset;
 491
 492                 if (unlikely((bio_op(split) == REQ_OP_DISCARD) &&
 493                          !blk_queue_discard(bdev_get_queue(split->bi_bdev)))) {
 494                         /* Just ignore it */
 495                         bio_endio(split);
 496                 } else {
 497                         if (mddev->gendisk)
 498                                 trace_block_bio_remap(bdev_get_queue(split->bi_bdev),
 499                                                       split, disk_devt(mddev->gendisk),
 500                                                       bio_sector);
 501                         generic_make_request(split);
 502                 }
 503         } while (split != bio);
 504 }
 505
 506 static void raid0_status(struct seq_file *seq, struct mddev *mddev)
 507 {
 508         seq_printf(seq, " %dk chunks", mddev->chunk_sectors / 2);
 509         return;
 510 }
 511
 512 static void *raid0_takeover_raid45(struct mddev *mddev)
 513 {
 514         struct md_rdev *rdev;
 515         struct r0conf *priv_conf;
 516
 517         if (mddev->degraded != 1) {
 518                 pr_warn("md/raid0:%s: raid5 must be degraded! Degraded disks: %d\n",
 519                         mdname(mddev),
 520                         mddev->degraded);
 521                 return ERR_PTR(-EINVAL);
 522         }
 523
 524         rdev_for_each(rdev, mddev) {
 525                 /* check slot number for a disk */
 526                 if (rdev->raid_disk == mddev->raid_disks-1) {
 527                         pr_warn("md/raid0:%s: raid5 must have missing parity disk!\n",
 528                                 mdname(mddev));
 529                         return ERR_PTR(-EINVAL);
 530                 }
 531                 rdev->sectors = mddev->dev_sectors;
 532         }
 533
 534         /* Set new parameters */
 535         mddev->new_level = 0;
 536         mddev->new_layout = 0;
 537         mddev->new_chunk_sectors = mddev->chunk_sectors;
 538         mddev->raid_disks--;
 539         mddev->delta_disks = -1;
 540         /* make sure it will be not marked as dirty */
 541         mddev->recovery_cp = MaxSector;
 542         clear_bit(MD_HAS_JOURNAL, &mddev->flags);
 543         clear_bit(MD_JOURNAL_CLEAN, &mddev->flags);
 544
 545         create_strip_zones(mddev, &priv_conf);
 546
 547         return priv_conf;
 548 }
 549
 550 static void *raid0_takeover_raid10(struct mddev *mddev)
 551 {
 552         struct r0conf *priv_conf;
 553
 554         /* Check layout:
 555          *  - far_copies must be 1
 556          *  - near_copies must be 2
 557          *  - disks number must be even
 558          *  - all mirrors must be already degraded
 559          */
 560         if (mddev->layout != ((1 << 8) + 2)) {
 561                 pr_warn("md/raid0:%s:: Raid0 cannot takeover layout: 0x%x\n",
 562                         mdname(mddev),
 563                         mddev->layout);
 564                 return ERR_PTR(-EINVAL);
 565         }
 566         if (mddev->raid_disks & 1) {
 567                 pr_warn("md/raid0:%s: Raid0 cannot takeover Raid10 with odd disk number.\n",
 568                         mdname(mddev));
 569                 return ERR_PTR(-EINVAL);
 570         }
 571         if (mddev->degraded != (mddev->raid_disks>>1)) {
 572                 pr_warn("md/raid0:%s: All mirrors must be already degraded!\n",
 573                         mdname(mddev));
 574                 return ERR_PTR(-EINVAL);
 575         }
 576
 577         /* Set new parameters */
 578         mddev->new_level = 0;
 579         mddev->new_layout = 0;
 580         mddev->new_chunk_sectors = mddev->chunk_sectors;
 581         mddev->delta_disks = - mddev->raid_disks / 2;
 582         mddev->raid_disks += mddev->delta_disks;
 583         mddev->degraded = 0;
 584         /* make sure it will be not marked as dirty */
 585         mddev->recovery_cp = MaxSector;
 586         clear_bit(MD_FAILFAST_SUPPORTED, &mddev->flags);
 587
 588         create_strip_zones(mddev, &priv_conf);
 589         return priv_conf;
 590 }
 591
 592 static void *raid0_takeover_raid1(struct mddev *mddev)
 593 {
 594         struct r0conf *priv_conf;
 595         int chunksect;
 596
 597         /* Check layout:
 598          *  - (N - 1) mirror drives must be already faulty
 599          */
 600         if ((mddev->raid_disks - 1) != mddev->degraded) {
 601                 pr_err("md/raid0:%s: (N - 1) mirrors drives must be already faulty!\n",
 602                        mdname(mddev));
 603                 return ERR_PTR(-EINVAL);
 604         }
 605
 606         /*
 607          * a raid1 doesn't have the notion of chunk size, so
 608          * figure out the largest suitable size we can use.
 609          */
 610         chunksect = 64 * 2; /* 64K by default */
 611
 612         /* The array must be an exact multiple of chunksize */
 613         while (chunksect && (mddev->array_sectors & (chunksect - 1)))
 614                 chunksect >>= 1;
 615
 616         if ((chunksect << 9) < PAGE_SIZE)
 617                 /* array size does not allow a suitable chunk size */
 618                 return ERR_PTR(-EINVAL);
 619
 620         /* Set new parameters */
 621         mddev->new_level = 0;
 622         mddev->new_layout = 0;
 623         mddev->new_chunk_sectors = chunksect;
 624         mddev->chunk_sectors = chunksect;
 625         mddev->delta_disks = 1 - mddev->raid_disks;
 626         mddev->raid_disks = 1;
 627         /* make sure it will be not marked as dirty */
 628         mddev->recovery_cp = MaxSector;
 629         clear_bit(MD_FAILFAST_SUPPORTED, &mddev->flags);
 630
 631         create_strip_zones(mddev, &priv_conf);
 632         return priv_conf;
 633 }
 634
 635 static void *raid0_takeover(struct mddev *mddev)
 636 {
 637         /* raid0 can take over:
 638          *  raid4 - if all data disks are active.
 639          *  raid5 - providing it is Raid4 layout and one disk is faulty
 640          *  raid10 - assuming we have all necessary active disks
 641          *  raid1 - with (N -1) mirror drives faulty
 642          */
 643
 644         if (mddev->bitmap) {
 645                 pr_warn("md/raid0: %s: cannot takeover array with bitmap\n",
 646                         mdname(mddev));
 647                 return ERR_PTR(-EBUSY);
 648         }
 649         if (mddev->level == 4)
 650                 return raid0_takeover_raid45(mddev);
 651
 652         if (mddev->level == 5) {
 653                 if (mddev->layout == ALGORITHM_PARITY_N)
 654                         return raid0_takeover_raid45(mddev);
 655
 656                 pr_warn("md/raid0:%s: Raid can only takeover Raid5 with layout: %d\n",
 657                         mdname(mddev), ALGORITHM_PARITY_N);
 658         }
 659
 660         if (mddev->level == 10)
 661                 return raid0_takeover_raid10(mddev);
 662
 663         if (mddev->level == 1)
 664                 return raid0_takeover_raid1(mddev);
 665
 666         pr_warn("Takeover from raid%i to raid0 not supported\n",
 667                 mddev->level);
 668
 669         return ERR_PTR(-EINVAL);
 670 }
 671
 672 static void raid0_quiesce(struct mddev *mddev, int state)
 673 {
 674 }
 675
 676 static struct md_personality raid0_personality=
 677 {
 678         .name           = "raid0",
 679         .level          = 0,
 680         .owner          = THIS_MODULE,
 681         .make_request   = raid0_make_request,
 682         .run            = raid0_run,
 683         .free           = raid0_free,
 684         .status         = raid0_status,
 685         .size           = raid0_size,
 686         .takeover       = raid0_takeover,
 687         .quiesce        = raid0_quiesce,
 688         .congested      = raid0_congested,
 689 };
 690
 691 static int __init raid0_init (void)
 692 {
 693         return register_md_personality (&raid0_personality);
 694 }
 695
 696 static void raid0_exit (void)
 697 {
 698         unregister_md_personality (&raid0_personality);
 699 }
 700
 701 module_init(raid0_init);
 702 module_exit(raid0_exit);
 703 MODULE_LICENSE("GPL");
 704 MODULE_DESCRIPTION("RAID0 (striping) personality for MD");
 705 MODULE_ALIAS("md-personality-2"); /* RAID0 */
 706 MODULE_ALIAS("md-raid0");
 707 MODULE_ALIAS("md-level-0");