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