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/proc_fs.h>
  14 #include <linux/seq_file.h>
  15 #include <linux/slab.h>
  16 #include <linux/kmod.h>
  17 #include <linux/kobj_map.h>
  18 #include <linux/buffer_head.h>
  19 #include <linux/mutex.h>
  20 #include <linux/idr.h>
  21
  22 #include "blk.h"
  23
  24 static DEFINE_MUTEX(block_class_lock);
  25 #ifndef CONFIG_SYSFS_DEPRECATED
  26 struct kobject *block_depr;
  27 #endif
  28
  29 /* for extended dynamic devt allocation, currently only one major is used */
  30 #define MAX_EXT_DEVT            (1 << MINORBITS)
  31
  32 /* For extended devt allocation.  ext_devt_mutex prevents look up
  33  * results from going away underneath its user.
  34  */
  35 static DEFINE_MUTEX(ext_devt_mutex);
  36 static DEFINE_IDR(ext_devt_idr);
  37
  38 static struct device_type disk_type;
  39
  40 /**
  41  * disk_get_part - get partition
  42  * @disk: disk to look partition from
  43  * @partno: partition number
  44  *
  45  * Look for partition @partno from @disk.  If found, increment
  46  * reference count and return it.
  47  *
  48  * CONTEXT:
  49  * Don't care.
  50  *
  51  * RETURNS:
  52  * Pointer to the found partition on success, NULL if not found.
  53  */
  54 struct hd_struct *disk_get_part(struct gendisk *disk, int partno)
  55 {
  56         struct hd_struct *part = NULL;
  57         struct disk_part_tbl *ptbl;
  58
  59         if (unlikely(partno < 0))
  60                 return NULL;
  61
  62         rcu_read_lock();
  63
  64         ptbl = rcu_dereference(disk->part_tbl);
  65         if (likely(partno < ptbl->len)) {
  66                 part = rcu_dereference(ptbl->part[partno]);
  67                 if (part)
  68                         get_device(part_to_dev(part));
  69         }
  70
  71         rcu_read_unlock();
  72
  73         return part;
  74 }
  75 EXPORT_SYMBOL_GPL(disk_get_part);
  76
  77 /**
  78  * disk_part_iter_init - initialize partition iterator
  79  * @piter: iterator to initialize
  80  * @disk: disk to iterate over
  81  * @flags: DISK_PITER_* flags
  82  *
  83  * Initialize @piter so that it iterates over partitions of @disk.
  84  *
  85  * CONTEXT:
  86  * Don't care.
  87  */
  88 void disk_part_iter_init(struct disk_part_iter *piter, struct gendisk *disk,
  89                           unsigned int flags)
  90 {
  91         struct disk_part_tbl *ptbl;
  92
  93         rcu_read_lock();
  94         ptbl = rcu_dereference(disk->part_tbl);
  95
  96         piter->disk = disk;
  97         piter->part = NULL;
  98
  99         if (flags & DISK_PITER_REVERSE)
 100                 piter->idx = ptbl->len - 1;
 101         else if (flags & (DISK_PITER_INCL_PART0 | DISK_PITER_INCL_EMPTY_PART0))
 102                 piter->idx = 0;
 103         else
 104                 piter->idx = 1;
 105
 106         piter->flags = flags;
 107
 108         rcu_read_unlock();
 109 }
 110 EXPORT_SYMBOL_GPL(disk_part_iter_init);
 111
 112 /**
 113  * disk_part_iter_next - proceed iterator to the next partition and return it
 114  * @piter: iterator of interest
 115  *
 116  * Proceed @piter to the next partition and return it.
 117  *
 118  * CONTEXT:
 119  * Don't care.
 120  */
 121 struct hd_struct *disk_part_iter_next(struct disk_part_iter *piter)
 122 {
 123         struct disk_part_tbl *ptbl;
 124         int inc, end;
 125
 126         /* put the last partition */
 127         disk_put_part(piter->part);
 128         piter->part = NULL;
 129
 130         /* get part_tbl */
 131         rcu_read_lock();
 132         ptbl = rcu_dereference(piter->disk->part_tbl);
 133
 134         /* determine iteration parameters */
 135         if (piter->flags & DISK_PITER_REVERSE) {
 136                 inc = -1;
 137                 if (piter->flags & (DISK_PITER_INCL_PART0 |
 138                                     DISK_PITER_INCL_EMPTY_PART0))
 139                         end = -1;
 140                 else
 141                         end = 0;
 142         } else {
 143                 inc = 1;
 144                 end = ptbl->len;
 145         }
 146
 147         /* iterate to the next partition */
 148         for (; piter->idx != end; piter->idx += inc) {
 149                 struct hd_struct *part;
 150
 151                 part = rcu_dereference(ptbl->part[piter->idx]);
 152                 if (!part)
 153                         continue;
 154                 if (!part->nr_sects &&
 155                     !(piter->flags & DISK_PITER_INCL_EMPTY) &&
 156                     !(piter->flags & DISK_PITER_INCL_EMPTY_PART0 &&
 157                       piter->idx == 0))
 158                         continue;
 159
 160                 get_device(part_to_dev(part));
 161                 piter->part = part;
 162                 piter->idx += inc;
 163                 break;
 164         }
 165
 166         rcu_read_unlock();
 167
 168         return piter->part;
 169 }
 170 EXPORT_SYMBOL_GPL(disk_part_iter_next);
 171
 172 /**
 173  * disk_part_iter_exit - finish up partition iteration
 174  * @piter: iter of interest
 175  *
 176  * Called when iteration is over.  Cleans up @piter.
 177  *
 178  * CONTEXT:
 179  * Don't care.
 180  */
 181 void disk_part_iter_exit(struct disk_part_iter *piter)
 182 {
 183         disk_put_part(piter->part);
 184         piter->part = NULL;
 185 }
 186 EXPORT_SYMBOL_GPL(disk_part_iter_exit);
 187
 188 /**
 189  * disk_map_sector_rcu - map sector to partition
 190  * @disk: gendisk of interest
 191  * @sector: sector to map
 192  *
 193  * Find out which partition @sector maps to on @disk.  This is
 194  * primarily used for stats accounting.
 195  *
 196  * CONTEXT:
 197  * RCU read locked.  The returned partition pointer is valid only
 198  * while preemption is disabled.
 199  *
 200  * RETURNS:
 201  * Found partition on success, part0 is returned if no partition matches
 202  */
 203 struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector)
 204 {
 205         struct disk_part_tbl *ptbl;
 206         int i;
 207
 208         ptbl = rcu_dereference(disk->part_tbl);
 209
 210         for (i = 1; i < ptbl->len; i++) {
 211                 struct hd_struct *part = rcu_dereference(ptbl->part[i]);
 212
 213                 if (part && part->start_sect <= sector &&
 214                     sector < part->start_sect + part->nr_sects)
 215                         return part;
 216         }
 217         return &disk->part0;
 218 }
 219 EXPORT_SYMBOL_GPL(disk_map_sector_rcu);
 220
 221 /*
 222  * Can be deleted altogether. Later.
 223  *
 224  */
 225 static struct blk_major_name {
 226         struct blk_major_name *next;
 227         int major;
 228         char name[16];
 229 } *major_names[BLKDEV_MAJOR_HASH_SIZE];
 230
 231 /* index in the above - for now: assume no multimajor ranges */
 232 static inline int major_to_index(int major)
 233 {
 234         return major % BLKDEV_MAJOR_HASH_SIZE;
 235 }
 236
 237 #ifdef CONFIG_PROC_FS
 238 void blkdev_show(struct seq_file *seqf, off_t offset)
 239 {
 240         struct blk_major_name *dp;
 241
 242         if (offset < BLKDEV_MAJOR_HASH_SIZE) {
 243                 mutex_lock(&block_class_lock);
 244                 for (dp = major_names[offset]; dp; dp = dp->next)
 245                         seq_printf(seqf, "%3d %s\n", dp->major, dp->name);
 246                 mutex_unlock(&block_class_lock);
 247         }
 248 }
 249 #endif /* CONFIG_PROC_FS */
 250
 251 int register_blkdev(unsigned int major, const char *name)
 252 {
 253         struct blk_major_name **n, *p;
 254         int index, ret = 0;
 255
 256         mutex_lock(&block_class_lock);
 257
 258         /* temporary */
 259         if (major == 0) {
 260                 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
 261                         if (major_names[index] == NULL)
 262                                 break;
 263                 }
 264
 265                 if (index == 0) {
 266                         printk("register_blkdev: failed to get major for %s\n",
 267                                name);
 268                         ret = -EBUSY;
 269                         goto out;
 270                 }
 271                 major = index;
 272                 ret = major;
 273         }
 274
 275         p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
 276         if (p == NULL) {
 277                 ret = -ENOMEM;
 278                 goto out;
 279         }
 280
 281         p->major = major;
 282         strlcpy(p->name, name, sizeof(p->name));
 283         p->next = NULL;
 284         index = major_to_index(major);
 285
 286         for (n = &major_names[index]; *n; n = &(*n)->next) {
 287                 if ((*n)->major == major)
 288                         break;
 289         }
 290         if (!*n)
 291                 *n = p;
 292         else
 293                 ret = -EBUSY;
 294
 295         if (ret < 0) {
 296                 printk("register_blkdev: cannot get major %d for %s\n",
 297                        major, name);
 298                 kfree(p);
 299         }
 300 out:
 301         mutex_unlock(&block_class_lock);
 302         return ret;
 303 }
 304
 305 EXPORT_SYMBOL(register_blkdev);
 306
 307 void unregister_blkdev(unsigned int major, const char *name)
 308 {
 309         struct blk_major_name **n;
 310         struct blk_major_name *p = NULL;
 311         int index = major_to_index(major);
 312
 313         mutex_lock(&block_class_lock);
 314         for (n = &major_names[index]; *n; n = &(*n)->next)
 315                 if ((*n)->major == major)
 316                         break;
 317         if (!*n || strcmp((*n)->name, name)) {
 318                 WARN_ON(1);
 319         } else {
 320                 p = *n;
 321                 *n = p->next;
 322         }
 323         mutex_unlock(&block_class_lock);
 324         kfree(p);
 325 }
 326
 327 EXPORT_SYMBOL(unregister_blkdev);
 328
 329 static struct kobj_map *bdev_map;
 330
 331 /**
 332  * blk_mangle_minor - scatter minor numbers apart
 333  * @minor: minor number to mangle
 334  *
 335  * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
 336  * is enabled.  Mangling twice gives the original value.
 337  *
 338  * RETURNS:
 339  * Mangled value.
 340  *
 341  * CONTEXT:
 342  * Don't care.
 343  */
 344 static int blk_mangle_minor(int minor)
 345 {
 346 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
 347         int i;
 348
 349         for (i = 0; i < MINORBITS / 2; i++) {
 350                 int low = minor & (1 << i);
 351                 int high = minor & (1 << (MINORBITS - 1 - i));
 352                 int distance = MINORBITS - 1 - 2 * i;
 353
 354                 minor ^= low | high;    /* clear both bits */
 355                 low <<= distance;       /* swap the positions */
 356                 high >>= distance;
 357                 minor |= low | high;    /* and set */
 358         }
 359 #endif
 360         return minor;
 361 }
 362
 363 /**
 364  * blk_alloc_devt - allocate a dev_t for a partition
 365  * @part: partition to allocate dev_t for
 366  * @devt: out parameter for resulting dev_t
 367  *
 368  * Allocate a dev_t for block device.
 369  *
 370  * RETURNS:
 371  * 0 on success, allocated dev_t is returned in *@devt.  -errno on
 372  * failure.
 373  *
 374  * CONTEXT:
 375  * Might sleep.
 376  */
 377 int blk_alloc_devt(struct hd_struct *part, dev_t *devt)
 378 {
 379         struct gendisk *disk = part_to_disk(part);
 380         int idx, rc;
 381
 382         /* in consecutive minor range? */
 383         if (part->partno < disk->minors) {
 384                 *devt = MKDEV(disk->major, disk->first_minor + part->partno);
 385                 return 0;
 386         }
 387
 388         /* allocate ext devt */
 389         do {
 390                 if (!idr_pre_get(&ext_devt_idr, GFP_KERNEL))
 391                         return -ENOMEM;
 392                 rc = idr_get_new(&ext_devt_idr, part, &idx);
 393         } while (rc == -EAGAIN);
 394
 395         if (rc)
 396                 return rc;
 397
 398         if (idx > MAX_EXT_DEVT) {
 399                 idr_remove(&ext_devt_idr, idx);
 400                 return -EBUSY;
 401         }
 402
 403         *devt = MKDEV(BLOCK_EXT_MAJOR, blk_mangle_minor(idx));
 404         return 0;
 405 }
 406
 407 /**
 408  * blk_free_devt - free a dev_t
 409  * @devt: dev_t to free
 410  *
 411  * Free @devt which was allocated using blk_alloc_devt().
 412  *
 413  * CONTEXT:
 414  * Might sleep.
 415  */
 416 void blk_free_devt(dev_t devt)
 417 {
 418         might_sleep();
 419
 420         if (devt == MKDEV(0, 0))
 421                 return;
 422
 423         if (MAJOR(devt) == BLOCK_EXT_MAJOR) {
 424                 mutex_lock(&ext_devt_mutex);
 425                 idr_remove(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
 426                 mutex_unlock(&ext_devt_mutex);
 427         }
 428 }
 429
 430 static char *bdevt_str(dev_t devt, char *buf)
 431 {
 432         if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
 433                 char tbuf[BDEVT_SIZE];
 434                 snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
 435                 snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
 436         } else
 437                 snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
 438
 439         return buf;
 440 }
 441
 442 /*
 443  * Register device numbers dev..(dev+range-1)
 444  * range must be nonzero
 445  * The hash chain is sorted on range, so that subranges can override.
 446  */
 447 void blk_register_region(dev_t devt, unsigned long range, struct module *module,
 448                          struct kobject *(*probe)(dev_t, int *, void *),
 449                          int (*lock)(dev_t, void *), void *data)
 450 {
 451         kobj_map(bdev_map, devt, range, module, probe, lock, data);
 452 }
 453
 454 EXPORT_SYMBOL(blk_register_region);
 455
 456 void blk_unregister_region(dev_t devt, unsigned long range)
 457 {
 458         kobj_unmap(bdev_map, devt, range);
 459 }
 460
 461 EXPORT_SYMBOL(blk_unregister_region);
 462
 463 static struct kobject *exact_match(dev_t devt, int *partno, void *data)
 464 {
 465         struct gendisk *p = data;
 466
 467         return &disk_to_dev(p)->kobj;
 468 }
 469
 470 static int exact_lock(dev_t devt, void *data)
 471 {
 472         struct gendisk *p = data;
 473
 474         if (!get_disk(p))
 475                 return -1;
 476         return 0;
 477 }
 478
 479 /**
 480  * add_disk - add partitioning information to kernel list
 481  * @disk: per-device partitioning information
 482  *
 483  * This function registers the partitioning information in @disk
 484  * with the kernel.
 485  *
 486  * FIXME: error handling
 487  */
 488 void add_disk(struct gendisk *disk)
 489 {
 490         struct backing_dev_info *bdi;
 491         dev_t devt;
 492         int retval;
 493
 494         /* minors == 0 indicates to use ext devt from part0 and should
 495          * be accompanied with EXT_DEVT flag.  Make sure all
 496          * parameters make sense.
 497          */
 498         WARN_ON(disk->minors && !(disk->major || disk->first_minor));
 499         WARN_ON(!disk->minors && !(disk->flags & GENHD_FL_EXT_DEVT));
 500
 501         disk->flags |= GENHD_FL_UP;
 502
 503         retval = blk_alloc_devt(&disk->part0, &devt);
 504         if (retval) {
 505                 WARN_ON(1);
 506                 return;
 507         }
 508         disk_to_dev(disk)->devt = devt;
 509
 510         /* ->major and ->first_minor aren't supposed to be
 511          * dereferenced from here on, but set them just in case.
 512          */
 513         disk->major = MAJOR(devt);
 514         disk->first_minor = MINOR(devt);
 515
 516         blk_register_region(disk_devt(disk), disk->minors, NULL,
 517                             exact_match, exact_lock, disk);
 518         register_disk(disk);
 519         blk_register_queue(disk);
 520
 521         bdi = &disk->queue->backing_dev_info;
 522         bdi_register_dev(bdi, disk_devt(disk));
 523         retval = sysfs_create_link(&disk_to_dev(disk)->kobj, &bdi->dev->kobj,
 524                                    "bdi");
 525         WARN_ON(retval);
 526 }
 527
 528 EXPORT_SYMBOL(add_disk);
 529 EXPORT_SYMBOL(del_gendisk);     /* in partitions/check.c */
 530
 531 void unlink_gendisk(struct gendisk *disk)
 532 {
 533         sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
 534         bdi_unregister(&disk->queue->backing_dev_info);
 535         blk_unregister_queue(disk);
 536         blk_unregister_region(disk_devt(disk), disk->minors);
 537 }
 538
 539 /**
 540  * get_gendisk - get partitioning information for a given device
 541  * @devt: device to get partitioning information for
 542  * @partno: returned partition index
 543  *
 544  * This function gets the structure containing partitioning
 545  * information for the given device @devt.
 546  */
 547 struct gendisk *get_gendisk(dev_t devt, int *partno)
 548 {
 549         struct gendisk *disk = NULL;
 550
 551         if (MAJOR(devt) != BLOCK_EXT_MAJOR) {
 552                 struct kobject *kobj;
 553
 554                 kobj = kobj_lookup(bdev_map, devt, partno);
 555                 if (kobj)
 556                         disk = dev_to_disk(kobj_to_dev(kobj));
 557         } else {
 558                 struct hd_struct *part;
 559
 560                 mutex_lock(&ext_devt_mutex);
 561                 part = idr_find(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
 562                 if (part && get_disk(part_to_disk(part))) {
 563                         *partno = part->partno;
 564                         disk = part_to_disk(part);
 565                 }
 566                 mutex_unlock(&ext_devt_mutex);
 567         }
 568
 569         return disk;
 570 }
 571
 572 /**
 573  * bdget_disk - do bdget() by gendisk and partition number
 574  * @disk: gendisk of interest
 575  * @partno: partition number
 576  *
 577  * Find partition @partno from @disk, do bdget() on it.
 578  *
 579  * CONTEXT:
 580  * Don't care.
 581  *
 582  * RETURNS:
 583  * Resulting block_device on success, NULL on failure.
 584  */
 585 struct block_device *bdget_disk(struct gendisk *disk, int partno)
 586 {
 587         struct hd_struct *part;
 588         struct block_device *bdev = NULL;
 589
 590         part = disk_get_part(disk, partno);
 591         if (part)
 592                 bdev = bdget(part_devt(part));
 593         disk_put_part(part);
 594
 595         return bdev;
 596 }
 597 EXPORT_SYMBOL(bdget_disk);
 598
 599 /*
 600  * print a full list of all partitions - intended for places where the root
 601  * filesystem can't be mounted and thus to give the victim some idea of what
 602  * went wrong
 603  */
 604 void __init printk_all_partitions(void)
 605 {
 606         struct class_dev_iter iter;
 607         struct device *dev;
 608
 609         class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
 610         while ((dev = class_dev_iter_next(&iter))) {
 611                 struct gendisk *disk = dev_to_disk(dev);
 612                 struct disk_part_iter piter;
 613                 struct hd_struct *part;
 614                 char name_buf[BDEVNAME_SIZE];
 615                 char devt_buf[BDEVT_SIZE];
 616
 617                 /*
 618                  * Don't show empty devices or things that have been
 619                  * surpressed
 620                  */
 621                 if (get_capacity(disk) == 0 ||
 622                     (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
 623                         continue;
 624
 625                 /*
 626                  * Note, unlike /proc/partitions, I am showing the
 627                  * numbers in hex - the same format as the root=
 628                  * option takes.
 629                  */
 630                 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0);
 631                 while ((part = disk_part_iter_next(&piter))) {
 632                         bool is_part0 = part == &disk->part0;
 633
 634                         printk("%s%s %10llu %s", is_part0 ? "" : "  ",
 635                                bdevt_str(part_devt(part), devt_buf),
 636                                (unsigned long long)part->nr_sects >> 1,
 637                                disk_name(disk, part->partno, name_buf));
 638                         if (is_part0) {
 639                                 if (disk->driverfs_dev != NULL &&
 640                                     disk->driverfs_dev->driver != NULL)
 641                                         printk(" driver: %s\n",
 642                                               disk->driverfs_dev->driver->name);
 643                                 else
 644                                         printk(" (driver?)\n");
 645                         } else
 646                                 printk("\n");
 647                 }
 648                 disk_part_iter_exit(&piter);
 649         }
 650         class_dev_iter_exit(&iter);
 651 }
 652
 653 #ifdef CONFIG_PROC_FS
 654 /* iterator */
 655 static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
 656 {
 657         loff_t skip = *pos;
 658         struct class_dev_iter *iter;
 659         struct device *dev;
 660
 661         iter = kmalloc(sizeof(*iter), GFP_KERNEL);
 662         if (!iter)
 663                 return ERR_PTR(-ENOMEM);
 664
 665         seqf->private = iter;
 666         class_dev_iter_init(iter, &block_class, NULL, &disk_type);
 667         do {
 668                 dev = class_dev_iter_next(iter);
 669                 if (!dev)
 670                         return NULL;
 671         } while (skip--);
 672
 673         return dev_to_disk(dev);
 674 }
 675
 676 static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
 677 {
 678         struct device *dev;
 679
 680         (*pos)++;
 681         dev = class_dev_iter_next(seqf->private);
 682         if (dev)
 683                 return dev_to_disk(dev);
 684
 685         return NULL;
 686 }
 687
 688 static void disk_seqf_stop(struct seq_file *seqf, void *v)
 689 {
 690         struct class_dev_iter *iter = seqf->private;
 691
 692         /* stop is called even after start failed :-( */
 693         if (iter) {
 694                 class_dev_iter_exit(iter);
 695                 kfree(iter);
 696         }
 697 }
 698
 699 static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
 700 {
 701         static void *p;
 702
 703         p = disk_seqf_start(seqf, pos);
 704         if (!IS_ERR(p) && p && !*pos)
 705                 seq_puts(seqf, "major minor  #blocks  name\n\n");
 706         return p;
 707 }
 708
 709 static int show_partition(struct seq_file *seqf, void *v)
 710 {
 711         struct gendisk *sgp = v;
 712         struct disk_part_iter piter;
 713         struct hd_struct *part;
 714         char buf[BDEVNAME_SIZE];
 715
 716         /* Don't show non-partitionable removeable devices or empty devices */
 717         if (!get_capacity(sgp) || (!disk_partitionable(sgp) &&
 718                                    (sgp->flags & GENHD_FL_REMOVABLE)))
 719                 return 0;
 720         if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
 721                 return 0;
 722
 723         /* show the full disk and all non-0 size partitions of it */
 724         disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0);
 725         while ((part = disk_part_iter_next(&piter)))
 726                 seq_printf(seqf, "%4d  %7d %10llu %s\n",
 727                            MAJOR(part_devt(part)), MINOR(part_devt(part)),
 728                            (unsigned long long)part->nr_sects >> 1,
 729                            disk_name(sgp, part->partno, buf));
 730         disk_part_iter_exit(&piter);
 731
 732         return 0;
 733 }
 734
 735 static const struct seq_operations partitions_op = {
 736         .start  = show_partition_start,
 737         .next   = disk_seqf_next,
 738         .stop   = disk_seqf_stop,
 739         .show   = show_partition
 740 };
 741
 742 static int partitions_open(struct inode *inode, struct file *file)
 743 {
 744         return seq_open(file, &partitions_op);
 745 }
 746
 747 static const struct file_operations proc_partitions_operations = {
 748         .open           = partitions_open,
 749         .read           = seq_read,
 750         .llseek         = seq_lseek,
 751         .release        = seq_release,
 752 };
 753 #endif
 754
 755
 756 static struct kobject *base_probe(dev_t devt, int *partno, void *data)
 757 {
 758         if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
 759                 /* Make old-style 2.4 aliases work */
 760                 request_module("block-major-%d", MAJOR(devt));
 761         return NULL;
 762 }
 763
 764 static int __init genhd_device_init(void)
 765 {
 766         int error;
 767
 768         block_class.dev_kobj = sysfs_dev_block_kobj;
 769         error = class_register(&block_class);
 770         if (unlikely(error))
 771                 return error;
 772         bdev_map = kobj_map_init(base_probe, &block_class_lock);
 773         blk_dev_init();
 774
 775         register_blkdev(BLOCK_EXT_MAJOR, "blkext");
 776
 777 #ifndef CONFIG_SYSFS_DEPRECATED
 778         /* create top-level block dir */
 779         block_depr = kobject_create_and_add("block", NULL);
 780 #endif
 781         return 0;
 782 }
 783
 784 subsys_initcall(genhd_device_init);
 785
 786 static ssize_t disk_range_show(struct device *dev,
 787                                struct device_attribute *attr, char *buf)
 788 {
 789         struct gendisk *disk = dev_to_disk(dev);
 790
 791         return sprintf(buf, "%d\n", disk->minors);
 792 }
 793
 794 static ssize_t disk_ext_range_show(struct device *dev,
 795                                    struct device_attribute *attr, char *buf)
 796 {
 797         struct gendisk *disk = dev_to_disk(dev);
 798
 799         return sprintf(buf, "%d\n", disk_max_parts(disk));
 800 }
 801
 802 static ssize_t disk_removable_show(struct device *dev,
 803                                    struct device_attribute *attr, char *buf)
 804 {
 805         struct gendisk *disk = dev_to_disk(dev);
 806
 807         return sprintf(buf, "%d\n",
 808                        (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
 809 }
 810
 811 static ssize_t disk_ro_show(struct device *dev,
 812                                    struct device_attribute *attr, char *buf)
 813 {
 814         struct gendisk *disk = dev_to_disk(dev);
 815
 816         return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
 817 }
 818
 819 static ssize_t disk_capability_show(struct device *dev,
 820                                     struct device_attribute *attr, char *buf)
 821 {
 822         struct gendisk *disk = dev_to_disk(dev);
 823
 824         return sprintf(buf, "%x\n", disk->flags);
 825 }
 826
 827 static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL);
 828 static DEVICE_ATTR(ext_range, S_IRUGO, disk_ext_range_show, NULL);
 829 static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL);
 830 static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL);
 831 static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL);
 832 static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL);
 833 static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL);
 834 #ifdef CONFIG_FAIL_MAKE_REQUEST
 835 static struct device_attribute dev_attr_fail =
 836         __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store);
 837 #endif
 838 #ifdef CONFIG_FAIL_IO_TIMEOUT
 839 static struct device_attribute dev_attr_fail_timeout =
 840         __ATTR(io-timeout-fail,  S_IRUGO|S_IWUSR, part_timeout_show,
 841                 part_timeout_store);
 842 #endif
 843
 844 static struct attribute *disk_attrs[] = {
 845         &dev_attr_range.attr,
 846         &dev_attr_ext_range.attr,
 847         &dev_attr_removable.attr,
 848         &dev_attr_ro.attr,
 849         &dev_attr_size.attr,
 850         &dev_attr_capability.attr,
 851         &dev_attr_stat.attr,
 852 #ifdef CONFIG_FAIL_MAKE_REQUEST
 853         &dev_attr_fail.attr,
 854 #endif
 855 #ifdef CONFIG_FAIL_IO_TIMEOUT
 856         &dev_attr_fail_timeout.attr,
 857 #endif
 858         NULL
 859 };
 860
 861 static struct attribute_group disk_attr_group = {
 862         .attrs = disk_attrs,
 863 };
 864
 865 static struct attribute_group *disk_attr_groups[] = {
 866         &disk_attr_group,
 867         NULL
 868 };
 869
 870 static void disk_free_ptbl_rcu_cb(struct rcu_head *head)
 871 {
 872         struct disk_part_tbl *ptbl =
 873                 container_of(head, struct disk_part_tbl, rcu_head);
 874
 875         kfree(ptbl);
 876 }
 877
 878 /**
 879  * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way
 880  * @disk: disk to replace part_tbl for
 881  * @new_ptbl: new part_tbl to install
 882  *
 883  * Replace disk->part_tbl with @new_ptbl in RCU-safe way.  The
 884  * original ptbl is freed using RCU callback.
 885  *
 886  * LOCKING:
 887  * Matching bd_mutx locked.
 888  */
 889 static void disk_replace_part_tbl(struct gendisk *disk,
 890                                   struct disk_part_tbl *new_ptbl)
 891 {
 892         struct disk_part_tbl *old_ptbl = disk->part_tbl;
 893
 894         rcu_assign_pointer(disk->part_tbl, new_ptbl);
 895         if (old_ptbl)
 896                 call_rcu(&old_ptbl->rcu_head, disk_free_ptbl_rcu_cb);
 897 }
 898
 899 /**
 900  * disk_expand_part_tbl - expand disk->part_tbl
 901  * @disk: disk to expand part_tbl for
 902  * @partno: expand such that this partno can fit in
 903  *
 904  * Expand disk->part_tbl such that @partno can fit in.  disk->part_tbl
 905  * uses RCU to allow unlocked dereferencing for stats and other stuff.
 906  *
 907  * LOCKING:
 908  * Matching bd_mutex locked, might sleep.
 909  *
 910  * RETURNS:
 911  * 0 on success, -errno on failure.
 912  */
 913 int disk_expand_part_tbl(struct gendisk *disk, int partno)
 914 {
 915         struct disk_part_tbl *old_ptbl = disk->part_tbl;
 916         struct disk_part_tbl *new_ptbl;
 917         int len = old_ptbl ? old_ptbl->len : 0;
 918         int target = partno + 1;
 919         size_t size;
 920         int i;
 921
 922         /* disk_max_parts() is zero during initialization, ignore if so */
 923         if (disk_max_parts(disk) && target > disk_max_parts(disk))
 924                 return -EINVAL;
 925
 926         if (target <= len)
 927                 return 0;
 928
 929         size = sizeof(*new_ptbl) + target * sizeof(new_ptbl->part[0]);
 930         new_ptbl = kzalloc_node(size, GFP_KERNEL, disk->node_id);
 931         if (!new_ptbl)
 932                 return -ENOMEM;
 933
 934         INIT_RCU_HEAD(&new_ptbl->rcu_head);
 935         new_ptbl->len = target;
 936
 937         for (i = 0; i < len; i++)
 938                 rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]);
 939
 940         disk_replace_part_tbl(disk, new_ptbl);
 941         return 0;
 942 }
 943
 944 static void disk_release(struct device *dev)
 945 {
 946         struct gendisk *disk = dev_to_disk(dev);
 947
 948         kfree(disk->random);
 949         disk_replace_part_tbl(disk, NULL);
 950         free_part_stats(&disk->part0);
 951         kfree(disk);
 952 }
 953 struct class block_class = {
 954         .name           = "block",
 955 };
 956
 957 static struct device_type disk_type = {
 958         .name           = "disk",
 959         .groups         = disk_attr_groups,
 960         .release        = disk_release,
 961 };
 962
 963 #ifdef CONFIG_PROC_FS
 964 /*
 965  * aggregate disk stat collector.  Uses the same stats that the sysfs
 966  * entries do, above, but makes them available through one seq_file.
 967  *
 968  * The output looks suspiciously like /proc/partitions with a bunch of
 969  * extra fields.
 970  */
 971 static int diskstats_show(struct seq_file *seqf, void *v)
 972 {
 973         struct gendisk *gp = v;
 974         struct disk_part_iter piter;
 975         struct hd_struct *hd;
 976         char buf[BDEVNAME_SIZE];
 977         int cpu;
 978
 979         /*
 980         if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
 981                 seq_puts(seqf,  "major minor name"
 982                                 "     rio rmerge rsect ruse wio wmerge "
 983                                 "wsect wuse running use aveq"
 984                                 "\n\n");
 985         */
 986
 987         disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0);
 988         while ((hd = disk_part_iter_next(&piter))) {
 989                 cpu = part_stat_lock();
 990                 part_round_stats(cpu, hd);
 991                 part_stat_unlock();
 992                 seq_printf(seqf, "%4d %7d %s %lu %lu %llu "
 993                            "%u %lu %lu %llu %u %u %u %u\n",
 994                            MAJOR(part_devt(hd)), MINOR(part_devt(hd)),
 995                            disk_name(gp, hd->partno, buf),
 996                            part_stat_read(hd, ios[0]),
 997                            part_stat_read(hd, merges[0]),
 998                            (unsigned long long)part_stat_read(hd, sectors[0]),
 999                            jiffies_to_msecs(part_stat_read(hd, ticks[0])),
1000                            part_stat_read(hd, ios[1]),
1001                            part_stat_read(hd, merges[1]),
1002                            (unsigned long long)part_stat_read(hd, sectors[1]),
1003                            jiffies_to_msecs(part_stat_read(hd, ticks[1])),
1004                            hd->in_flight,
1005                            jiffies_to_msecs(part_stat_read(hd, io_ticks)),
1006                            jiffies_to_msecs(part_stat_read(hd, time_in_queue))
1007                         );
1008         }
1009         disk_part_iter_exit(&piter);
1010
1011         return 0;
1012 }
1013
1014 static const struct seq_operations diskstats_op = {
1015         .start  = disk_seqf_start,
1016         .next   = disk_seqf_next,
1017         .stop   = disk_seqf_stop,
1018         .show   = diskstats_show
1019 };
1020
1021 static int diskstats_open(struct inode *inode, struct file *file)
1022 {
1023         return seq_open(file, &diskstats_op);
1024 }
1025
1026 static const struct file_operations proc_diskstats_operations = {
1027         .open           = diskstats_open,
1028         .read           = seq_read,
1029         .llseek         = seq_lseek,
1030         .release        = seq_release,
1031 };
1032
1033 static int __init proc_genhd_init(void)
1034 {
1035         proc_create("diskstats", 0, NULL, &proc_diskstats_operations);
1036         proc_create("partitions", 0, NULL, &proc_partitions_operations);
1037         return 0;
1038 }
1039 module_init(proc_genhd_init);
1040 #endif /* CONFIG_PROC_FS */
1041
1042 static void media_change_notify_thread(struct work_struct *work)
1043 {
1044         struct gendisk *gd = container_of(work, struct gendisk, async_notify);
1045         char event[] = "MEDIA_CHANGE=1";
1046         char *envp[] = { event, NULL };
1047
1048         /*
1049          * set enviroment vars to indicate which event this is for
1050          * so that user space will know to go check the media status.
1051          */
1052         kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1053         put_device(gd->driverfs_dev);
1054 }
1055
1056 #if 0
1057 void genhd_media_change_notify(struct gendisk *disk)
1058 {
1059         get_device(disk->driverfs_dev);
1060         schedule_work(&disk->async_notify);
1061 }
1062 EXPORT_SYMBOL_GPL(genhd_media_change_notify);
1063 #endif  /*  0  */
1064
1065 dev_t blk_lookup_devt(const char *name, int partno)
1066 {
1067         dev_t devt = MKDEV(0, 0);
1068         struct class_dev_iter iter;
1069         struct device *dev;
1070
1071         class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1072         while ((dev = class_dev_iter_next(&iter))) {
1073                 struct gendisk *disk = dev_to_disk(dev);
1074                 struct hd_struct *part;
1075
1076                 if (strcmp(dev->bus_id, name))
1077                         continue;
1078
1079                 part = disk_get_part(disk, partno);
1080                 if (part) {
1081                         devt = part_devt(part);
1082                         disk_put_part(part);
1083                         break;
1084                 }
1085                 disk_put_part(part);
1086         }
1087         class_dev_iter_exit(&iter);
1088         return devt;
1089 }
1090 EXPORT_SYMBOL(blk_lookup_devt);
1091
1092 struct gendisk *alloc_disk(int minors)
1093 {
1094         return alloc_disk_node(minors, -1);
1095 }
1096 EXPORT_SYMBOL(alloc_disk);
1097
1098 struct gendisk *alloc_disk_node(int minors, int node_id)
1099 {
1100         struct gendisk *disk;
1101
1102         disk = kmalloc_node(sizeof(struct gendisk),
1103                                 GFP_KERNEL | __GFP_ZERO, node_id);
1104         if (disk) {
1105                 if (!init_part_stats(&disk->part0)) {
1106                         kfree(disk);
1107                         return NULL;
1108                 }
1109                 disk->node_id = node_id;
1110                 if (disk_expand_part_tbl(disk, 0)) {
1111                         free_part_stats(&disk->part0);
1112                         kfree(disk);
1113                         return NULL;
1114                 }
1115                 disk->part_tbl->part[0] = &disk->part0;
1116
1117                 disk->minors = minors;
1118                 rand_initialize_disk(disk);
1119                 disk_to_dev(disk)->class = &block_class;
1120                 disk_to_dev(disk)->type = &disk_type;
1121                 device_initialize(disk_to_dev(disk));
1122                 INIT_WORK(&disk->async_notify,
1123                         media_change_notify_thread);
1124         }
1125         return disk;
1126 }
1127 EXPORT_SYMBOL(alloc_disk_node);
1128
1129 struct kobject *get_disk(struct gendisk *disk)
1130 {
1131         struct module *owner;
1132         struct kobject *kobj;
1133
1134         if (!disk->fops)
1135                 return NULL;
1136         owner = disk->fops->owner;
1137         if (owner && !try_module_get(owner))
1138                 return NULL;
1139         kobj = kobject_get(&disk_to_dev(disk)->kobj);
1140         if (kobj == NULL) {
1141                 module_put(owner);
1142                 return NULL;
1143         }
1144         return kobj;
1145
1146 }
1147
1148 EXPORT_SYMBOL(get_disk);
1149
1150 void put_disk(struct gendisk *disk)
1151 {
1152         if (disk)
1153                 kobject_put(&disk_to_dev(disk)->kobj);
1154 }
1155
1156 EXPORT_SYMBOL(put_disk);
1157
1158 void set_device_ro(struct block_device *bdev, int flag)
1159 {
1160         bdev->bd_part->policy = flag;
1161 }
1162
1163 EXPORT_SYMBOL(set_device_ro);
1164
1165 void set_disk_ro(struct gendisk *disk, int flag)
1166 {
1167         struct disk_part_iter piter;
1168         struct hd_struct *part;
1169
1170         disk_part_iter_init(&piter, disk,
1171                             DISK_PITER_INCL_EMPTY | DISK_PITER_INCL_PART0);
1172         while ((part = disk_part_iter_next(&piter)))
1173                 part->policy = flag;
1174         disk_part_iter_exit(&piter);
1175 }
1176
1177 EXPORT_SYMBOL(set_disk_ro);
1178
1179 int bdev_read_only(struct block_device *bdev)
1180 {
1181         if (!bdev)
1182                 return 0;
1183         return bdev->bd_part->policy;
1184 }
1185
1186 EXPORT_SYMBOL(bdev_read_only);
1187
1188 int invalidate_partition(struct gendisk *disk, int partno)
1189 {
1190         int res = 0;
1191         struct block_device *bdev = bdget_disk(disk, partno);
1192         if (bdev) {
1193                 fsync_bdev(bdev);
1194                 res = __invalidate_device(bdev);
1195                 bdput(bdev);
1196         }
1197         return res;
1198 }
1199
1200 EXPORT_SYMBOL(invalidate_partition);