block/genhd.c

   1 /*
   2  *  gendisk handling
   3  */
   4
   5 #include <linux/module.h>
   6 #include <linux/fs.h>
   7 #include <linux/genhd.h>
   8 #include <linux/kdev_t.h>
   9 #include <linux/kernel.h>
  10 #include <linux/blkdev.h>
  11 #include <linux/init.h>
  12 #include <linux/spinlock.h>
  13 #include <linux/seq_file.h>
  14 #include <linux/slab.h>
  15 #include <linux/kmod.h>
  16 #include <linux/kobj_map.h>
  17 #include <linux/buffer_head.h>
  18 #include <linux/mutex.h>
  19 #include <linux/idr.h>
  20
  21 #include "blk.h"
  22
  23 static DEFINE_MUTEX(block_class_lock);
  24 #ifndef CONFIG_SYSFS_DEPRECATED
  25 struct kobject *block_depr;
  26 #endif
  27
  28 /* for extended dynamic devt allocation, currently only one major is used */
  29 #define MAX_EXT_DEVT            (1 << MINORBITS)
  30
  31 /* For extended devt allocation.  ext_devt_mutex prevents look up
  32  * results from going away underneath its user.
  33  */
  34 static DEFINE_MUTEX(ext_devt_mutex);
  35 static DEFINE_IDR(ext_devt_idr);
  36
  37 static struct device_type disk_type;
  38
  39 /**
  40  * disk_get_part - get partition
  41  * @disk: disk to look partition from
  42  * @partno: partition number
  43  *
  44  * Look for partition @partno from @disk.  If found, increment
  45  * reference count and return it.
  46  *
  47  * CONTEXT:
  48  * Don't care.
  49  *
  50  * RETURNS:
  51  * Pointer to the found partition on success, NULL if not found.
  52  */
  53 struct hd_struct *disk_get_part(struct gendisk *disk, int partno)
  54 {
  55         struct hd_struct *part = NULL;
  56         struct disk_part_tbl *ptbl;
  57
  58         if (unlikely(partno < 0))
  59                 return NULL;
  60
  61         rcu_read_lock();
  62
  63         ptbl = rcu_dereference(disk->part_tbl);
  64         if (likely(partno < ptbl->len)) {
  65                 part = rcu_dereference(ptbl->part[partno]);
  66                 if (part)
  67                         get_device(part_to_dev(part));
  68         }
  69
  70         rcu_read_unlock();
  71
  72         return part;
  73 }
  74 EXPORT_SYMBOL_GPL(disk_get_part);
  75
  76 /**
  77  * disk_part_iter_init - initialize partition iterator
  78  * @piter: iterator to initialize
  79  * @disk: disk to iterate over
  80  * @flags: DISK_PITER_* flags
  81  *
  82  * Initialize @piter so that it iterates over partitions of @disk.
  83  *
  84  * CONTEXT:
  85  * Don't care.
  86  */
  87 void disk_part_iter_init(struct disk_part_iter *piter, struct gendisk *disk,
  88                           unsigned int flags)
  89 {
  90         struct disk_part_tbl *ptbl;
  91
  92         rcu_read_lock();
  93         ptbl = rcu_dereference(disk->part_tbl);
  94
  95         piter->disk = disk;
  96         piter->part = NULL;
  97
  98         if (flags & DISK_PITER_REVERSE)
  99                 piter->idx = ptbl->len - 1;
 100         else if (flags & DISK_PITER_INCL_PART0)
 101                 piter->idx = 0;
 102         else
 103                 piter->idx = 1;
 104
 105         piter->flags = flags;
 106
 107         rcu_read_unlock();
 108 }
 109 EXPORT_SYMBOL_GPL(disk_part_iter_init);
 110
 111 /**
 112  * disk_part_iter_next - proceed iterator to the next partition and return it
 113  * @piter: iterator of interest
 114  *
 115  * Proceed @piter to the next partition and return it.
 116  *
 117  * CONTEXT:
 118  * Don't care.
 119  */
 120 struct hd_struct *disk_part_iter_next(struct disk_part_iter *piter)
 121 {
 122         struct disk_part_tbl *ptbl;
 123         int inc, end;
 124
 125         /* put the last partition */
 126         disk_put_part(piter->part);
 127         piter->part = NULL;
 128
 129         /* get part_tbl */
 130         rcu_read_lock();
 131         ptbl = rcu_dereference(piter->disk->part_tbl);
 132
 133         /* determine iteration parameters */
 134         if (piter->flags & DISK_PITER_REVERSE) {
 135                 inc = -1;
 136                 if (piter->flags & DISK_PITER_INCL_PART0)
 137                         end = -1;
 138                 else
 139                         end = 0;
 140         } else {
 141                 inc = 1;
 142                 end = ptbl->len;
 143         }
 144
 145         /* iterate to the next partition */
 146         for (; piter->idx != end; piter->idx += inc) {
 147                 struct hd_struct *part;
 148
 149                 part = rcu_dereference(ptbl->part[piter->idx]);
 150                 if (!part)
 151                         continue;
 152                 if (!(piter->flags & DISK_PITER_INCL_EMPTY) && !part->nr_sects)
 153                         continue;
 154
 155                 get_device(part_to_dev(part));
 156                 piter->part = part;
 157                 piter->idx += inc;
 158                 break;
 159         }
 160
 161         rcu_read_unlock();
 162
 163         return piter->part;
 164 }
 165 EXPORT_SYMBOL_GPL(disk_part_iter_next);
 166
 167 /**
 168  * disk_part_iter_exit - finish up partition iteration
 169  * @piter: iter of interest
 170  *
 171  * Called when iteration is over.  Cleans up @piter.
 172  *
 173  * CONTEXT:
 174  * Don't care.
 175  */
 176 void disk_part_iter_exit(struct disk_part_iter *piter)
 177 {
 178         disk_put_part(piter->part);
 179         piter->part = NULL;
 180 }
 181 EXPORT_SYMBOL_GPL(disk_part_iter_exit);
 182
 183 /**
 184  * disk_map_sector_rcu - map sector to partition
 185  * @disk: gendisk of interest
 186  * @sector: sector to map
 187  *
 188  * Find out which partition @sector maps to on @disk.  This is
 189  * primarily used for stats accounting.
 190  *
 191  * CONTEXT:
 192  * RCU read locked.  The returned partition pointer is valid only
 193  * while preemption is disabled.
 194  *
 195  * RETURNS:
 196  * Found partition on success, part0 is returned if no partition matches
 197  */
 198 struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector)
 199 {
 200         struct disk_part_tbl *ptbl;
 201         int i;
 202
 203         ptbl = rcu_dereference(disk->part_tbl);
 204
 205         for (i = 1; i < ptbl->len; i++) {
 206                 struct hd_struct *part = rcu_dereference(ptbl->part[i]);
 207
 208                 if (part && part->start_sect <= sector &&
 209                     sector < part->start_sect + part->nr_sects)
 210                         return part;
 211         }
 212         return &disk->part0;
 213 }
 214 EXPORT_SYMBOL_GPL(disk_map_sector_rcu);
 215
 216 /*
 217  * Can be deleted altogether. Later.
 218  *
 219  */
 220 static struct blk_major_name {
 221         struct blk_major_name *next;
 222         int major;
 223         char name[16];
 224 } *major_names[BLKDEV_MAJOR_HASH_SIZE];
 225
 226 /* index in the above - for now: assume no multimajor ranges */
 227 static inline int major_to_index(int major)
 228 {
 229         return major % BLKDEV_MAJOR_HASH_SIZE;
 230 }
 231
 232 #ifdef CONFIG_PROC_FS
 233 void blkdev_show(struct seq_file *seqf, off_t offset)
 234 {
 235         struct blk_major_name *dp;
 236
 237         if (offset < BLKDEV_MAJOR_HASH_SIZE) {
 238                 mutex_lock(&block_class_lock);
 239                 for (dp = major_names[offset]; dp; dp = dp->next)
 240                         seq_printf(seqf, "%3d %s\n", dp->major, dp->name);
 241                 mutex_unlock(&block_class_lock);
 242         }
 243 }
 244 #endif /* CONFIG_PROC_FS */
 245
 246 int register_blkdev(unsigned int major, const char *name)
 247 {
 248         struct blk_major_name **n, *p;
 249         int index, ret = 0;
 250
 251         mutex_lock(&block_class_lock);
 252
 253         /* temporary */
 254         if (major == 0) {
 255                 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
 256                         if (major_names[index] == NULL)
 257                                 break;
 258                 }
 259
 260                 if (index == 0) {
 261                         printk("register_blkdev: failed to get major for %s\n",
 262                                name);
 263                         ret = -EBUSY;
 264                         goto out;
 265                 }
 266                 major = index;
 267                 ret = major;
 268         }
 269
 270         p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
 271         if (p == NULL) {
 272                 ret = -ENOMEM;
 273                 goto out;
 274         }
 275
 276         p->major = major;
 277         strlcpy(p->name, name, sizeof(p->name));
 278         p->next = NULL;
 279         index = major_to_index(major);
 280
 281         for (n = &major_names[index]; *n; n = &(*n)->next) {
 282                 if ((*n)->major == major)
 283                         break;
 284         }
 285         if (!*n)
 286                 *n = p;
 287         else
 288                 ret = -EBUSY;
 289
 290         if (ret < 0) {
 291                 printk("register_blkdev: cannot get major %d for %s\n",
 292                        major, name);
 293                 kfree(p);
 294         }
 295 out:
 296         mutex_unlock(&block_class_lock);
 297         return ret;
 298 }
 299
 300 EXPORT_SYMBOL(register_blkdev);
 301
 302 void unregister_blkdev(unsigned int major, const char *name)
 303 {
 304         struct blk_major_name **n;
 305         struct blk_major_name *p = NULL;
 306         int index = major_to_index(major);
 307
 308         mutex_lock(&block_class_lock);
 309         for (n = &major_names[index]; *n; n = &(*n)->next)
 310                 if ((*n)->major == major)
 311                         break;
 312         if (!*n || strcmp((*n)->name, name)) {
 313                 WARN_ON(1);
 314         } else {
 315                 p = *n;
 316                 *n = p->next;
 317         }
 318         mutex_unlock(&block_class_lock);
 319         kfree(p);
 320 }
 321
 322 EXPORT_SYMBOL(unregister_blkdev);
 323
 324 static struct kobj_map *bdev_map;
 325
 326 /**
 327  * blk_mangle_minor - scatter minor numbers apart
 328  * @minor: minor number to mangle
 329  *
 330  * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
 331  * is enabled.  Mangling twice gives the original value.
 332  *
 333  * RETURNS:
 334  * Mangled value.
 335  *
 336  * CONTEXT:
 337  * Don't care.
 338  */
 339 static int blk_mangle_minor(int minor)
 340 {
 341 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
 342         int i;
 343
 344         for (i = 0; i < MINORBITS / 2; i++) {
 345                 int low = minor & (1 << i);
 346                 int high = minor & (1 << (MINORBITS - 1 - i));
 347                 int distance = MINORBITS - 1 - 2 * i;
 348
 349                 minor ^= low | high;    /* clear both bits */
 350                 low <<= distance;       /* swap the positions */
 351                 high >>= distance;
 352                 minor |= low | high;    /* and set */
 353         }
 354 #endif
 355         return minor;
 356 }
 357
 358 /**
 359  * blk_alloc_devt - allocate a dev_t for a partition
 360  * @part: partition to allocate dev_t for
 361  * @devt: out parameter for resulting dev_t
 362  *
 363  * Allocate a dev_t for block device.
 364  *
 365  * RETURNS:
 366  * 0 on success, allocated dev_t is returned in *@devt.  -errno on
 367  * failure.
 368  *
 369  * CONTEXT:
 370  * Might sleep.
 371  */
 372 int blk_alloc_devt(struct hd_struct *part, dev_t *devt)
 373 {
 374         struct gendisk *disk = part_to_disk(part);
 375         int idx, rc;
 376
 377         /* in consecutive minor range? */
 378         if (part->partno < disk->minors) {
 379                 *devt = MKDEV(disk->major, disk->first_minor + part->partno);
 380                 return 0;
 381         }
 382
 383         /* allocate ext devt */
 384         do {
 385                 if (!idr_pre_get(&ext_devt_idr, GFP_KERNEL))
 386                         return -ENOMEM;
 387                 rc = idr_get_new(&ext_devt_idr, part, &idx);
 388         } while (rc == -EAGAIN);
 389
 390         if (rc)
 391                 return rc;
 392
 393         if (idx > MAX_EXT_DEVT) {
 394                 idr_remove(&ext_devt_idr, idx);
 395                 return -EBUSY;
 396         }
 397
 398         *devt = MKDEV(BLOCK_EXT_MAJOR, blk_mangle_minor(idx));
 399         return 0;
 400 }
 401
 402 /**
 403  * blk_free_devt - free a dev_t
 404  * @devt: dev_t to free
 405  *
 406  * Free @devt which was allocated using blk_alloc_devt().
 407  *
 408  * CONTEXT:
 409  * Might sleep.
 410  */
 411 void blk_free_devt(dev_t devt)
 412 {
 413         might_sleep();
 414
 415         if (devt == MKDEV(0, 0))
 416                 return;
 417
 418         if (MAJOR(devt) == BLOCK_EXT_MAJOR) {
 419                 mutex_lock(&ext_devt_mutex);
 420                 idr_remove(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
 421                 mutex_unlock(&ext_devt_mutex);
 422         }
 423 }
 424
 425 static char *bdevt_str(dev_t devt, char *buf)
 426 {
 427         if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
 428                 char tbuf[BDEVT_SIZE];
 429                 snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
 430                 snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
 431         } else
 432                 snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
 433
 434         return buf;
 435 }
 436
 437 /*
 438  * Register device numbers dev..(dev+range-1)
 439  * range must be nonzero
 440  * The hash chain is sorted on range, so that subranges can override.
 441  */
 442 void blk_register_region(dev_t devt, unsigned long range, struct module *module,
 443                          struct kobject *(*probe)(dev_t, int *, void *),
 444                          int (*lock)(dev_t, void *), void *data)
 445 {
 446         kobj_map(bdev_map, devt, range, module, probe, lock, data);
 447 }
 448
 449 EXPORT_SYMBOL(blk_register_region);
 450
 451 void blk_unregister_region(dev_t devt, unsigned long range)
 452 {
 453         kobj_unmap(bdev_map, devt, range);
 454 }
 455
 456 EXPORT_SYMBOL(blk_unregister_region);
 457
 458 static struct kobject *exact_match(dev_t devt, int *partno, void *data)
 459 {
 460         struct gendisk *p = data;
 461
 462         return &disk_to_dev(p)->kobj;
 463 }
 464
 465 static int exact_lock(dev_t devt, void *data)
 466 {
 467         struct gendisk *p = data;
 468
 469         if (!get_disk(p))
 470                 return -1;
 471         return 0;
 472 }
 473
 474 /**
 475  * add_disk - add partitioning information to kernel list
 476  * @disk: per-device partitioning information
 477  *
 478  * This function registers the partitioning information in @disk
 479  * with the kernel.
 480  *
 481  * FIXME: error handling
 482  */
 483 void add_disk(struct gendisk *disk)
 484 {
 485         struct backing_dev_info *bdi;
 486         dev_t devt;
 487         int retval;
 488
 489         /* minors == 0 indicates to use ext devt from part0 and should
 490          * be accompanied with EXT_DEVT flag.  Make sure all
 491          * parameters make sense.
 492          */
 493         WARN_ON(disk->minors && !(disk->major || disk->first_minor));
 494         WARN_ON(!disk->minors && !(disk->flags & GENHD_FL_EXT_DEVT));
 495
 496         disk->flags |= GENHD_FL_UP;
 497
 498         retval = blk_alloc_devt(&disk->part0, &devt);
 499         if (retval) {
 500                 WARN_ON(1);
 501                 return;
 502         }
 503         disk_to_dev(disk)->devt = devt;
 504
 505         /* ->major and ->first_minor aren't supposed to be
 506          * dereferenced from here on, but set them just in case.
 507          */
 508         disk->major = MAJOR(devt);
 509         disk->first_minor = MINOR(devt);
 510
 511         blk_register_region(disk_devt(disk), disk->minors, NULL,
 512                             exact_match, exact_lock, disk);
 513         register_disk(disk);
 514         blk_register_queue(disk);
 515
 516         bdi = &disk->queue->backing_dev_info;
 517         bdi_register_dev(bdi, disk_devt(disk));
 518         retval = sysfs_create_link(&disk_to_dev(disk)->kobj, &bdi->dev->kobj,
 519                                    "bdi");
 520         WARN_ON(retval);
 521 }
 522
 523 EXPORT_SYMBOL(add_disk);
 524 EXPORT_SYMBOL(del_gendisk);     /* in partitions/check.c */
 525
 526 void unlink_gendisk(struct gendisk *disk)
 527 {
 528         sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
 529         bdi_unregister(&disk->queue->backing_dev_info);
 530         blk_unregister_queue(disk);
 531         blk_unregister_region(disk_devt(disk), disk->minors);
 532 }
 533
 534 /**
 535  * get_gendisk - get partitioning information for a given device
 536  * @devt: device to get partitioning information for
 537  * @partno: returned partition index
 538  *
 539  * This function gets the structure containing partitioning
 540  * information for the given device @devt.
 541  */
 542 struct gendisk *get_gendisk(dev_t devt, int *partno)
 543 {
 544         struct gendisk *disk = NULL;
 545
 546         if (MAJOR(devt) != BLOCK_EXT_MAJOR) {
 547                 struct kobject *kobj;
 548
 549                 kobj = kobj_lookup(bdev_map, devt, partno);
 550                 if (kobj)
 551                         disk = dev_to_disk(kobj_to_dev(kobj));
 552         } else {
 553                 struct hd_struct *part;
 554
 555                 mutex_lock(&ext_devt_mutex);
 556                 part = idr_find(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
 557                 if (part && get_disk(part_to_disk(part))) {
 558                         *partno = part->partno;
 559                         disk = part_to_disk(part);
 560                 }
 561                 mutex_unlock(&ext_devt_mutex);
 562         }
 563
 564         return disk;
 565 }
 566
 567 /**
 568  * bdget_disk - do bdget() by gendisk and partition number
 569  * @disk: gendisk of interest
 570  * @partno: partition number
 571  *
 572  * Find partition @partno from @disk, do bdget() on it.
 573  *
 574  * CONTEXT:
 575  * Don't care.
 576  *
 577  * RETURNS:
 578  * Resulting block_device on success, NULL on failure.
 579  */
 580 struct block_device *bdget_disk(struct gendisk *disk, int partno)
 581 {
 582         struct hd_struct *part;
 583         struct block_device *bdev = NULL;
 584
 585         part = disk_get_part(disk, partno);
 586         if (part)
 587                 bdev = bdget(part_devt(part));
 588         disk_put_part(part);
 589
 590         return bdev;
 591 }
 592 EXPORT_SYMBOL(bdget_disk);
 593
 594 /*
 595  * print a full list of all partitions - intended for places where the root
 596  * filesystem can't be mounted and thus to give the victim some idea of what
 597  * went wrong
 598  */
 599 void __init printk_all_partitions(void)
 600 {
 601         struct class_dev_iter iter;
 602         struct device *dev;
 603
 604         class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
 605         while ((dev = class_dev_iter_next(&iter))) {
 606                 struct gendisk *disk = dev_to_disk(dev);
 607                 struct disk_part_iter piter;
 608                 struct hd_struct *part;
 609                 char name_buf[BDEVNAME_SIZE];
 610                 char devt_buf[BDEVT_SIZE];
 611
 612                 /*
 613                  * Don't show empty devices or things that have been
 614                  * surpressed
 615                  */
 616                 if (get_capacity(disk) == 0 ||
 617                     (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
 618                         continue;
 619
 620                 /*
 621                  * Note, unlike /proc/partitions, I am showing the
 622                  * numbers in hex - the same format as the root=
 623                  * option takes.
 624                  */
 625                 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0);
 626                 while ((part = disk_part_iter_next(&piter))) {
 627                         bool is_part0 = part == &disk->part0;
 628
 629                         printk("%s%s %10llu %s", is_part0 ? "" : "  ",
 630                                bdevt_str(part_devt(part), devt_buf),
 631                                (unsigned long long)part->nr_sects >> 1,
 632                                disk_name(disk, part->partno, name_buf));
 633                         if (is_part0) {
 634                                 if (disk->driverfs_dev != NULL &&
 635                                     disk->driverfs_dev->driver != NULL)
 636                                         printk(" driver: %s\n",
 637                                               disk->driverfs_dev->driver->name);
 638                                 else
 639                                         printk(" (driver?)\n");
 640                         } else
 641                                 printk("\n");
 642                 }
 643                 disk_part_iter_exit(&piter);
 644         }
 645         class_dev_iter_exit(&iter);
 646 }
 647
 648 #ifdef CONFIG_PROC_FS
 649 /* iterator */
 650 static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
 651 {
 652         loff_t skip = *pos;
 653         struct class_dev_iter *iter;
 654         struct device *dev;
 655
 656         iter = kmalloc(sizeof(*iter), GFP_KERNEL);
 657         if (!iter)
 658                 return ERR_PTR(-ENOMEM);
 659
 660         seqf->private = iter;
 661         class_dev_iter_init(iter, &block_class, NULL, &disk_type);
 662         do {
 663                 dev = class_dev_iter_next(iter);
 664                 if (!dev)
 665                         return NULL;
 666         } while (skip--);
 667
 668         return dev_to_disk(dev);
 669 }
 670
 671 static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
 672 {
 673         struct device *dev;
 674
 675         (*pos)++;
 676         dev = class_dev_iter_next(seqf->private);
 677         if (dev)
 678                 return dev_to_disk(dev);
 679
 680         return NULL;
 681 }
 682
 683 static void disk_seqf_stop(struct seq_file *seqf, void *v)
 684 {
 685         struct class_dev_iter *iter = seqf->private;
 686
 687         /* stop is called even after start failed :-( */
 688         if (iter) {
 689                 class_dev_iter_exit(iter);
 690                 kfree(iter);
 691         }
 692 }
 693
 694 static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
 695 {
 696         static void *p;
 697
 698         p = disk_seqf_start(seqf, pos);
 699         if (!IS_ERR(p) && p && !*pos)
 700                 seq_puts(seqf, "major minor  #blocks  name\n\n");
 701         return p;
 702 }
 703
 704 static int show_partition(struct seq_file *seqf, void *v)
 705 {
 706         struct gendisk *sgp = v;
 707         struct disk_part_iter piter;
 708         struct hd_struct *part;
 709         char buf[BDEVNAME_SIZE];
 710
 711         /* Don't show non-partitionable removeable devices or empty devices */
 712         if (!get_capacity(sgp) || (!disk_partitionable(sgp) &&
 713                                    (sgp->flags & GENHD_FL_REMOVABLE)))
 714                 return 0;
 715         if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
 716                 return 0;
 717
 718         /* show the full disk and all non-0 size partitions of it */
 719         disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0);
 720         while ((part = disk_part_iter_next(&piter)))
 721                 seq_printf(seqf, "%4d  %7d %10llu %s\n",
 722                            MAJOR(part_devt(part)), MINOR(part_devt(part)),
 723                            (unsigned long long)part->nr_sects >> 1,
 724                            disk_name(sgp, part->partno, buf));
 725         disk_part_iter_exit(&piter);
 726
 727         return 0;
 728 }
 729
 730 const struct seq_operations partitions_op = {
 731         .start  = show_partition_start,
 732         .next   = disk_seqf_next,
 733         .stop   = disk_seqf_stop,
 734         .show   = show_partition
 735 };
 736 #endif
 737
 738
 739 static struct kobject *base_probe(dev_t devt, int *partno, void *data)
 740 {
 741         if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
 742                 /* Make old-style 2.4 aliases work */
 743                 request_module("block-major-%d", MAJOR(devt));
 744         return NULL;
 745 }
 746
 747 static int __init genhd_device_init(void)
 748 {
 749         int error;
 750
 751         block_class.dev_kobj = sysfs_dev_block_kobj;
 752         error = class_register(&block_class);
 753         if (unlikely(error))
 754                 return error;
 755         bdev_map = kobj_map_init(base_probe, &block_class_lock);
 756         blk_dev_init();
 757
 758 #ifndef CONFIG_SYSFS_DEPRECATED
 759         /* create top-level block dir */
 760         block_depr = kobject_create_and_add("block", NULL);
 761 #endif
 762         return 0;
 763 }
 764
 765 subsys_initcall(genhd_device_init);
 766
 767 static ssize_t disk_range_show(struct device *dev,
 768                                struct device_attribute *attr, char *buf)
 769 {
 770         struct gendisk *disk = dev_to_disk(dev);
 771
 772         return sprintf(buf, "%d\n", disk->minors);
 773 }
 774
 775 static ssize_t disk_ext_range_show(struct device *dev,
 776                                    struct device_attribute *attr, char *buf)
 777 {
 778         struct gendisk *disk = dev_to_disk(dev);
 779
 780         return sprintf(buf, "%d\n", disk_max_parts(disk));
 781 }
 782
 783 static ssize_t disk_removable_show(struct device *dev,
 784                                    struct device_attribute *attr, char *buf)
 785 {
 786         struct gendisk *disk = dev_to_disk(dev);
 787
 788         return sprintf(buf, "%d\n",
 789                        (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
 790 }
 791
 792 static ssize_t disk_ro_show(struct device *dev,
 793                                    struct device_attribute *attr, char *buf)
 794 {
 795         struct gendisk *disk = dev_to_disk(dev);
 796
 797         return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
 798 }
 799
 800 static ssize_t disk_capability_show(struct device *dev,
 801                                     struct device_attribute *attr, char *buf)
 802 {
 803         struct gendisk *disk = dev_to_disk(dev);
 804
 805         return sprintf(buf, "%x\n", disk->flags);
 806 }
 807
 808 static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL);
 809 static DEVICE_ATTR(ext_range, S_IRUGO, disk_ext_range_show, NULL);
 810 static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL);
 811 static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL);
 812 static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL);
 813 static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL);
 814 static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL);
 815 #ifdef CONFIG_FAIL_MAKE_REQUEST
 816 static struct device_attribute dev_attr_fail =
 817         __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store);
 818 #endif
 819 #ifdef CONFIG_FAIL_IO_TIMEOUT
 820 static struct device_attribute dev_attr_fail_timeout =
 821         __ATTR(io-timeout-fail,  S_IRUGO|S_IWUSR, part_timeout_show,
 822                 part_timeout_store);
 823 #endif
 824
 825 static struct attribute *disk_attrs[] = {
 826         &dev_attr_range.attr,
 827         &dev_attr_ext_range.attr,
 828         &dev_attr_removable.attr,
 829         &dev_attr_ro.attr,
 830         &dev_attr_size.attr,
 831         &dev_attr_capability.attr,
 832         &dev_attr_stat.attr,
 833 #ifdef CONFIG_FAIL_MAKE_REQUEST
 834         &dev_attr_fail.attr,
 835 #endif
 836 #ifdef CONFIG_FAIL_IO_TIMEOUT
 837         &dev_attr_fail_timeout.attr,
 838 #endif
 839         NULL
 840 };
 841
 842 static struct attribute_group disk_attr_group = {
 843         .attrs = disk_attrs,
 844 };
 845
 846 static struct attribute_group *disk_attr_groups[] = {
 847         &disk_attr_group,
 848         NULL
 849 };
 850
 851 static void disk_free_ptbl_rcu_cb(struct rcu_head *head)
 852 {
 853         struct disk_part_tbl *ptbl =
 854                 container_of(head, struct disk_part_tbl, rcu_head);
 855
 856         kfree(ptbl);
 857 }
 858
 859 /**
 860  * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way
 861  * @disk: disk to replace part_tbl for
 862  * @new_ptbl: new part_tbl to install
 863  *
 864  * Replace disk->part_tbl with @new_ptbl in RCU-safe way.  The
 865  * original ptbl is freed using RCU callback.
 866  *
 867  * LOCKING:
 868  * Matching bd_mutx locked.
 869  */
 870 static void disk_replace_part_tbl(struct gendisk *disk,
 871                                   struct disk_part_tbl *new_ptbl)
 872 {
 873         struct disk_part_tbl *old_ptbl = disk->part_tbl;
 874
 875         rcu_assign_pointer(disk->part_tbl, new_ptbl);
 876         if (old_ptbl)
 877                 call_rcu(&old_ptbl->rcu_head, disk_free_ptbl_rcu_cb);
 878 }
 879
 880 /**
 881  * disk_expand_part_tbl - expand disk->part_tbl
 882  * @disk: disk to expand part_tbl for
 883  * @partno: expand such that this partno can fit in
 884  *
 885  * Expand disk->part_tbl such that @partno can fit in.  disk->part_tbl
 886  * uses RCU to allow unlocked dereferencing for stats and other stuff.
 887  *
 888  * LOCKING:
 889  * Matching bd_mutex locked, might sleep.
 890  *
 891  * RETURNS:
 892  * 0 on success, -errno on failure.
 893  */
 894 int disk_expand_part_tbl(struct gendisk *disk, int partno)
 895 {
 896         struct disk_part_tbl *old_ptbl = disk->part_tbl;
 897         struct disk_part_tbl *new_ptbl;
 898         int len = old_ptbl ? old_ptbl->len : 0;
 899         int target = partno + 1;
 900         size_t size;
 901         int i;
 902
 903         /* disk_max_parts() is zero during initialization, ignore if so */
 904         if (disk_max_parts(disk) && target > disk_max_parts(disk))
 905                 return -EINVAL;
 906
 907         if (target <= len)
 908                 return 0;
 909
 910         size = sizeof(*new_ptbl) + target * sizeof(new_ptbl->part[0]);
 911         new_ptbl = kzalloc_node(size, GFP_KERNEL, disk->node_id);
 912         if (!new_ptbl)
 913                 return -ENOMEM;
 914
 915         INIT_RCU_HEAD(&new_ptbl->rcu_head);
 916         new_ptbl->len = target;
 917
 918         for (i = 0; i < len; i++)
 919                 rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]);
 920
 921         disk_replace_part_tbl(disk, new_ptbl);
 922         return 0;
 923 }
 924
 925 static void disk_release(struct device *dev)
 926 {
 927         struct gendisk *disk = dev_to_disk(dev);
 928
 929         kfree(disk->random);
 930         disk_replace_part_tbl(disk, NULL);
 931         free_part_stats(&disk->part0);
 932         kfree(disk);
 933 }
 934 struct class block_class = {
 935         .name           = "block",
 936 };
 937
 938 static struct device_type disk_type = {
 939         .name           = "disk",
 940         .groups         = disk_attr_groups,
 941         .release        = disk_release,
 942 };
 943
 944 #ifdef CONFIG_PROC_FS
 945 /*
 946  * aggregate disk stat collector.  Uses the same stats that the sysfs
 947  * entries do, above, but makes them available through one seq_file.
 948  *
 949  * The output looks suspiciously like /proc/partitions with a bunch of
 950  * extra fields.
 951  */
 952 static int diskstats_show(struct seq_file *seqf, void *v)
 953 {
 954         struct gendisk *gp = v;
 955         struct disk_part_iter piter;
 956         struct hd_struct *hd;
 957         char buf[BDEVNAME_SIZE];
 958         int cpu;
 959
 960         /*
 961         if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
 962                 seq_puts(seqf,  "major minor name"
 963                                 "     rio rmerge rsect ruse wio wmerge "
 964                                 "wsect wuse running use aveq"
 965                                 "\n\n");
 966         */
 967
 968         disk_part_iter_init(&piter, gp, DISK_PITER_INCL_PART0);
 969         while ((hd = disk_part_iter_next(&piter))) {
 970                 cpu = part_stat_lock();
 971                 part_round_stats(cpu, hd);
 972                 part_stat_unlock();
 973                 seq_printf(seqf, "%4d %7d %s %lu %lu %llu "
 974                            "%u %lu %lu %llu %u %u %u %u\n",
 975                            MAJOR(part_devt(hd)), MINOR(part_devt(hd)),
 976                            disk_name(gp, hd->partno, buf),
 977                            part_stat_read(hd, ios[0]),
 978                            part_stat_read(hd, merges[0]),
 979                            (unsigned long long)part_stat_read(hd, sectors[0]),
 980                            jiffies_to_msecs(part_stat_read(hd, ticks[0])),
 981                            part_stat_read(hd, ios[1]),
 982                            part_stat_read(hd, merges[1]),
 983                            (unsigned long long)part_stat_read(hd, sectors[1]),
 984                            jiffies_to_msecs(part_stat_read(hd, ticks[1])),
 985                            hd->in_flight,
 986                            jiffies_to_msecs(part_stat_read(hd, io_ticks)),
 987                            jiffies_to_msecs(part_stat_read(hd, time_in_queue))
 988                         );
 989         }
 990         disk_part_iter_exit(&piter);
 991
 992         return 0;
 993 }
 994
 995 const struct seq_operations diskstats_op = {
 996         .start  = disk_seqf_start,
 997         .next   = disk_seqf_next,
 998         .stop   = disk_seqf_stop,
 999         .show   = diskstats_show
1000 };
1001 #endif /* CONFIG_PROC_FS */
1002
1003 static void media_change_notify_thread(struct work_struct *work)
1004 {
1005         struct gendisk *gd = container_of(work, struct gendisk, async_notify);
1006         char event[] = "MEDIA_CHANGE=1";
1007         char *envp[] = { event, NULL };
1008
1009         /*
1010          * set enviroment vars to indicate which event this is for
1011          * so that user space will know to go check the media status.
1012          */
1013         kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1014         put_device(gd->driverfs_dev);
1015 }
1016
1017 #if 0
1018 void genhd_media_change_notify(struct gendisk *disk)
1019 {
1020         get_device(disk->driverfs_dev);
1021         schedule_work(&disk->async_notify);
1022 }
1023 EXPORT_SYMBOL_GPL(genhd_media_change_notify);
1024 #endif  /*  0  */
1025
1026 dev_t blk_lookup_devt(const char *name, int partno)
1027 {
1028         dev_t devt = MKDEV(0, 0);
1029         struct class_dev_iter iter;
1030         struct device *dev;
1031
1032         class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1033         while ((dev = class_dev_iter_next(&iter))) {
1034                 struct gendisk *disk = dev_to_disk(dev);
1035                 struct hd_struct *part;
1036
1037                 if (strcmp(dev->bus_id, name))
1038                         continue;
1039
1040                 part = disk_get_part(disk, partno);
1041                 if (part) {
1042                         devt = part_devt(part);
1043                         disk_put_part(part);
1044                         break;
1045                 }
1046                 disk_put_part(part);
1047         }
1048         class_dev_iter_exit(&iter);
1049         return devt;
1050 }
1051 EXPORT_SYMBOL(blk_lookup_devt);
1052
1053 struct gendisk *alloc_disk(int minors)
1054 {
1055         return alloc_disk_node(minors, -1);
1056 }
1057 EXPORT_SYMBOL(alloc_disk);
1058
1059 struct gendisk *alloc_disk_node(int minors, int node_id)
1060 {
1061         struct gendisk *disk;
1062
1063         disk = kmalloc_node(sizeof(struct gendisk),
1064                                 GFP_KERNEL | __GFP_ZERO, node_id);
1065         if (disk) {
1066                 if (!init_part_stats(&disk->part0)) {
1067                         kfree(disk);
1068                         return NULL;
1069                 }
1070                 if (disk_expand_part_tbl(disk, 0)) {
1071                         free_part_stats(&disk->part0);
1072                         kfree(disk);
1073                         return NULL;
1074                 }
1075                 disk->part_tbl->part[0] = &disk->part0;
1076
1077                 disk->minors = minors;
1078                 rand_initialize_disk(disk);
1079                 disk_to_dev(disk)->class = &block_class;
1080                 disk_to_dev(disk)->type = &disk_type;
1081                 device_initialize(disk_to_dev(disk));
1082                 INIT_WORK(&disk->async_notify,
1083                         media_change_notify_thread);
1084                 disk->node_id = node_id;
1085         }
1086         return disk;
1087 }
1088 EXPORT_SYMBOL(alloc_disk_node);
1089
1090 struct kobject *get_disk(struct gendisk *disk)
1091 {
1092         struct module *owner;
1093         struct kobject *kobj;
1094
1095         if (!disk->fops)
1096                 return NULL;
1097         owner = disk->fops->owner;
1098         if (owner && !try_module_get(owner))
1099                 return NULL;
1100         kobj = kobject_get(&disk_to_dev(disk)->kobj);
1101         if (kobj == NULL) {
1102                 module_put(owner);
1103                 return NULL;
1104         }
1105         return kobj;
1106
1107 }
1108
1109 EXPORT_SYMBOL(get_disk);
1110
1111 void put_disk(struct gendisk *disk)
1112 {
1113         if (disk)
1114                 kobject_put(&disk_to_dev(disk)->kobj);
1115 }
1116
1117 EXPORT_SYMBOL(put_disk);
1118
1119 void set_device_ro(struct block_device *bdev, int flag)
1120 {
1121         bdev->bd_part->policy = flag;
1122 }
1123
1124 EXPORT_SYMBOL(set_device_ro);
1125
1126 void set_disk_ro(struct gendisk *disk, int flag)
1127 {
1128         struct disk_part_iter piter;
1129         struct hd_struct *part;
1130
1131         disk_part_iter_init(&piter, disk,
1132                             DISK_PITER_INCL_EMPTY | DISK_PITER_INCL_PART0);
1133         while ((part = disk_part_iter_next(&piter)))
1134                 part->policy = flag;
1135         disk_part_iter_exit(&piter);
1136 }
1137
1138 EXPORT_SYMBOL(set_disk_ro);
1139
1140 int bdev_read_only(struct block_device *bdev)
1141 {
1142         if (!bdev)
1143                 return 0;
1144         return bdev->bd_part->policy;
1145 }
1146
1147 EXPORT_SYMBOL(bdev_read_only);
1148
1149 int invalidate_partition(struct gendisk *disk, int partno)
1150 {
1151         int res = 0;
1152         struct block_device *bdev = bdget_disk(disk, partno);
1153         if (bdev) {
1154                 fsync_bdev(bdev);
1155                 res = __invalidate_device(bdev);
1156                 bdput(bdev);
1157         }
1158         return res;
1159 }
1160
1161 EXPORT_SYMBOL(invalidate_partition);