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)
 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                         end = -1;
 139                 else
 140                         end = 0;
 141         } else {
 142                 inc = 1;
 143                 end = ptbl->len;
 144         }
 145
 146         /* iterate to the next partition */
 147         for (; piter->idx != end; piter->idx += inc) {
 148                 struct hd_struct *part;
 149
 150                 part = rcu_dereference(ptbl->part[piter->idx]);
 151                 if (!part)
 152                         continue;
 153                 if (!(piter->flags & DISK_PITER_INCL_EMPTY) && !part->nr_sects)
 154                         continue;
 155
 156                 get_device(part_to_dev(part));
 157                 piter->part = part;
 158                 piter->idx += inc;
 159                 break;
 160         }
 161
 162         rcu_read_unlock();
 163
 164         return piter->part;
 165 }
 166 EXPORT_SYMBOL_GPL(disk_part_iter_next);
 167
 168 /**
 169  * disk_part_iter_exit - finish up partition iteration
 170  * @piter: iter of interest
 171  *
 172  * Called when iteration is over.  Cleans up @piter.
 173  *
 174  * CONTEXT:
 175  * Don't care.
 176  */
 177 void disk_part_iter_exit(struct disk_part_iter *piter)
 178 {
 179         disk_put_part(piter->part);
 180         piter->part = NULL;
 181 }
 182 EXPORT_SYMBOL_GPL(disk_part_iter_exit);
 183
 184 static inline int sector_in_part(struct hd_struct *part, sector_t sector)
 185 {
 186         return part->start_sect <= sector &&
 187                 sector < part->start_sect + part->nr_sects;
 188 }
 189
 190 /**
 191  * disk_map_sector_rcu - map sector to partition
 192  * @disk: gendisk of interest
 193  * @sector: sector to map
 194  *
 195  * Find out which partition @sector maps to on @disk.  This is
 196  * primarily used for stats accounting.
 197  *
 198  * CONTEXT:
 199  * RCU read locked.  The returned partition pointer is valid only
 200  * while preemption is disabled.
 201  *
 202  * RETURNS:
 203  * Found partition on success, part0 is returned if no partition matches
 204  */
 205 struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector)
 206 {
 207         struct disk_part_tbl *ptbl;
 208         struct hd_struct *part;
 209         int i;
 210
 211         ptbl = rcu_dereference(disk->part_tbl);
 212
 213         part = rcu_dereference(ptbl->last_lookup);
 214         if (part && sector_in_part(part, sector))
 215                 return part;
 216
 217         for (i = 1; i < ptbl->len; i++) {
 218                 part = rcu_dereference(ptbl->part[i]);
 219
 220                 if (part && sector_in_part(part, sector)) {
 221                         rcu_assign_pointer(ptbl->last_lookup, part);
 222                         return part;
 223                 }
 224         }
 225         return &disk->part0;
 226 }
 227 EXPORT_SYMBOL_GPL(disk_map_sector_rcu);
 228
 229 /*
 230  * Can be deleted altogether. Later.
 231  *
 232  */
 233 static struct blk_major_name {
 234         struct blk_major_name *next;
 235         int major;
 236         char name[16];
 237 } *major_names[BLKDEV_MAJOR_HASH_SIZE];
 238
 239 /* index in the above - for now: assume no multimajor ranges */
 240 static inline int major_to_index(int major)
 241 {
 242         return major % BLKDEV_MAJOR_HASH_SIZE;
 243 }
 244
 245 #ifdef CONFIG_PROC_FS
 246 void blkdev_show(struct seq_file *seqf, off_t offset)
 247 {
 248         struct blk_major_name *dp;
 249
 250         if (offset < BLKDEV_MAJOR_HASH_SIZE) {
 251                 mutex_lock(&block_class_lock);
 252                 for (dp = major_names[offset]; dp; dp = dp->next)
 253                         seq_printf(seqf, "%3d %s\n", dp->major, dp->name);
 254                 mutex_unlock(&block_class_lock);
 255         }
 256 }
 257 #endif /* CONFIG_PROC_FS */
 258
 259 int register_blkdev(unsigned int major, const char *name)
 260 {
 261         struct blk_major_name **n, *p;
 262         int index, ret = 0;
 263
 264         mutex_lock(&block_class_lock);
 265
 266         /* temporary */
 267         if (major == 0) {
 268                 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
 269                         if (major_names[index] == NULL)
 270                                 break;
 271                 }
 272
 273                 if (index == 0) {
 274                         printk("register_blkdev: failed to get major for %s\n",
 275                                name);
 276                         ret = -EBUSY;
 277                         goto out;
 278                 }
 279                 major = index;
 280                 ret = major;
 281         }
 282
 283         p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
 284         if (p == NULL) {
 285                 ret = -ENOMEM;
 286                 goto out;
 287         }
 288
 289         p->major = major;
 290         strlcpy(p->name, name, sizeof(p->name));
 291         p->next = NULL;
 292         index = major_to_index(major);
 293
 294         for (n = &major_names[index]; *n; n = &(*n)->next) {
 295                 if ((*n)->major == major)
 296                         break;
 297         }
 298         if (!*n)
 299                 *n = p;
 300         else
 301                 ret = -EBUSY;
 302
 303         if (ret < 0) {
 304                 printk("register_blkdev: cannot get major %d for %s\n",
 305                        major, name);
 306                 kfree(p);
 307         }
 308 out:
 309         mutex_unlock(&block_class_lock);
 310         return ret;
 311 }
 312
 313 EXPORT_SYMBOL(register_blkdev);
 314
 315 void unregister_blkdev(unsigned int major, const char *name)
 316 {
 317         struct blk_major_name **n;
 318         struct blk_major_name *p = NULL;
 319         int index = major_to_index(major);
 320
 321         mutex_lock(&block_class_lock);
 322         for (n = &major_names[index]; *n; n = &(*n)->next)
 323                 if ((*n)->major == major)
 324                         break;
 325         if (!*n || strcmp((*n)->name, name)) {
 326                 WARN_ON(1);
 327         } else {
 328                 p = *n;
 329                 *n = p->next;
 330         }
 331         mutex_unlock(&block_class_lock);
 332         kfree(p);
 333 }
 334
 335 EXPORT_SYMBOL(unregister_blkdev);
 336
 337 static struct kobj_map *bdev_map;
 338
 339 /**
 340  * blk_mangle_minor - scatter minor numbers apart
 341  * @minor: minor number to mangle
 342  *
 343  * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
 344  * is enabled.  Mangling twice gives the original value.
 345  *
 346  * RETURNS:
 347  * Mangled value.
 348  *
 349  * CONTEXT:
 350  * Don't care.
 351  */
 352 static int blk_mangle_minor(int minor)
 353 {
 354 #ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
 355         int i;
 356
 357         for (i = 0; i < MINORBITS / 2; i++) {
 358                 int low = minor & (1 << i);
 359                 int high = minor & (1 << (MINORBITS - 1 - i));
 360                 int distance = MINORBITS - 1 - 2 * i;
 361
 362                 minor ^= low | high;    /* clear both bits */
 363                 low <<= distance;       /* swap the positions */
 364                 high >>= distance;
 365                 minor |= low | high;    /* and set */
 366         }
 367 #endif
 368         return minor;
 369 }
 370
 371 /**
 372  * blk_alloc_devt - allocate a dev_t for a partition
 373  * @part: partition to allocate dev_t for
 374  * @devt: out parameter for resulting dev_t
 375  *
 376  * Allocate a dev_t for block device.
 377  *
 378  * RETURNS:
 379  * 0 on success, allocated dev_t is returned in *@devt.  -errno on
 380  * failure.
 381  *
 382  * CONTEXT:
 383  * Might sleep.
 384  */
 385 int blk_alloc_devt(struct hd_struct *part, dev_t *devt)
 386 {
 387         struct gendisk *disk = part_to_disk(part);
 388         int idx, rc;
 389
 390         /* in consecutive minor range? */
 391         if (part->partno < disk->minors) {
 392                 *devt = MKDEV(disk->major, disk->first_minor + part->partno);
 393                 return 0;
 394         }
 395
 396         /* allocate ext devt */
 397         do {
 398                 if (!idr_pre_get(&ext_devt_idr, GFP_KERNEL))
 399                         return -ENOMEM;
 400                 rc = idr_get_new(&ext_devt_idr, part, &idx);
 401         } while (rc == -EAGAIN);
 402
 403         if (rc)
 404                 return rc;
 405
 406         if (idx > MAX_EXT_DEVT) {
 407                 idr_remove(&ext_devt_idr, idx);
 408                 return -EBUSY;
 409         }
 410
 411         *devt = MKDEV(BLOCK_EXT_MAJOR, blk_mangle_minor(idx));
 412         return 0;
 413 }
 414
 415 /**
 416  * blk_free_devt - free a dev_t
 417  * @devt: dev_t to free
 418  *
 419  * Free @devt which was allocated using blk_alloc_devt().
 420  *
 421  * CONTEXT:
 422  * Might sleep.
 423  */
 424 void blk_free_devt(dev_t devt)
 425 {
 426         might_sleep();
 427
 428         if (devt == MKDEV(0, 0))
 429                 return;
 430
 431         if (MAJOR(devt) == BLOCK_EXT_MAJOR) {
 432                 mutex_lock(&ext_devt_mutex);
 433                 idr_remove(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
 434                 mutex_unlock(&ext_devt_mutex);
 435         }
 436 }
 437
 438 static char *bdevt_str(dev_t devt, char *buf)
 439 {
 440         if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
 441                 char tbuf[BDEVT_SIZE];
 442                 snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
 443                 snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
 444         } else
 445                 snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
 446
 447         return buf;
 448 }
 449
 450 /*
 451  * Register device numbers dev..(dev+range-1)
 452  * range must be nonzero
 453  * The hash chain is sorted on range, so that subranges can override.
 454  */
 455 void blk_register_region(dev_t devt, unsigned long range, struct module *module,
 456                          struct kobject *(*probe)(dev_t, int *, void *),
 457                          int (*lock)(dev_t, void *), void *data)
 458 {
 459         kobj_map(bdev_map, devt, range, module, probe, lock, data);
 460 }
 461
 462 EXPORT_SYMBOL(blk_register_region);
 463
 464 void blk_unregister_region(dev_t devt, unsigned long range)
 465 {
 466         kobj_unmap(bdev_map, devt, range);
 467 }
 468
 469 EXPORT_SYMBOL(blk_unregister_region);
 470
 471 static struct kobject *exact_match(dev_t devt, int *partno, void *data)
 472 {
 473         struct gendisk *p = data;
 474
 475         return &disk_to_dev(p)->kobj;
 476 }
 477
 478 static int exact_lock(dev_t devt, void *data)
 479 {
 480         struct gendisk *p = data;
 481
 482         if (!get_disk(p))
 483                 return -1;
 484         return 0;
 485 }
 486
 487 /**
 488  * add_disk - add partitioning information to kernel list
 489  * @disk: per-device partitioning information
 490  *
 491  * This function registers the partitioning information in @disk
 492  * with the kernel.
 493  *
 494  * FIXME: error handling
 495  */
 496 void add_disk(struct gendisk *disk)
 497 {
 498         struct backing_dev_info *bdi;
 499         dev_t devt;
 500         int retval;
 501
 502         /* minors == 0 indicates to use ext devt from part0 and should
 503          * be accompanied with EXT_DEVT flag.  Make sure all
 504          * parameters make sense.
 505          */
 506         WARN_ON(disk->minors && !(disk->major || disk->first_minor));
 507         WARN_ON(!disk->minors && !(disk->flags & GENHD_FL_EXT_DEVT));
 508
 509         disk->flags |= GENHD_FL_UP;
 510
 511         retval = blk_alloc_devt(&disk->part0, &devt);
 512         if (retval) {
 513                 WARN_ON(1);
 514                 return;
 515         }
 516         disk_to_dev(disk)->devt = devt;
 517
 518         /* ->major and ->first_minor aren't supposed to be
 519          * dereferenced from here on, but set them just in case.
 520          */
 521         disk->major = MAJOR(devt);
 522         disk->first_minor = MINOR(devt);
 523
 524         blk_register_region(disk_devt(disk), disk->minors, NULL,
 525                             exact_match, exact_lock, disk);
 526         register_disk(disk);
 527         blk_register_queue(disk);
 528
 529         bdi = &disk->queue->backing_dev_info;
 530         bdi_register_dev(bdi, disk_devt(disk));
 531         retval = sysfs_create_link(&disk_to_dev(disk)->kobj, &bdi->dev->kobj,
 532                                    "bdi");
 533         WARN_ON(retval);
 534 }
 535
 536 EXPORT_SYMBOL(add_disk);
 537 EXPORT_SYMBOL(del_gendisk);     /* in partitions/check.c */
 538
 539 void unlink_gendisk(struct gendisk *disk)
 540 {
 541         sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
 542         bdi_unregister(&disk->queue->backing_dev_info);
 543         blk_unregister_queue(disk);
 544         blk_unregister_region(disk_devt(disk), disk->minors);
 545 }
 546
 547 /**
 548  * get_gendisk - get partitioning information for a given device
 549  * @devt: device to get partitioning information for
 550  * @partno: returned partition index
 551  *
 552  * This function gets the structure containing partitioning
 553  * information for the given device @devt.
 554  */
 555 struct gendisk *get_gendisk(dev_t devt, int *partno)
 556 {
 557         struct gendisk *disk = NULL;
 558
 559         if (MAJOR(devt) != BLOCK_EXT_MAJOR) {
 560                 struct kobject *kobj;
 561
 562                 kobj = kobj_lookup(bdev_map, devt, partno);
 563                 if (kobj)
 564                         disk = dev_to_disk(kobj_to_dev(kobj));
 565         } else {
 566                 struct hd_struct *part;
 567
 568                 mutex_lock(&ext_devt_mutex);
 569                 part = idr_find(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
 570                 if (part && get_disk(part_to_disk(part))) {
 571                         *partno = part->partno;
 572                         disk = part_to_disk(part);
 573                 }
 574                 mutex_unlock(&ext_devt_mutex);
 575         }
 576
 577         return disk;
 578 }
 579
 580 /**
 581  * bdget_disk - do bdget() by gendisk and partition number
 582  * @disk: gendisk of interest
 583  * @partno: partition number
 584  *
 585  * Find partition @partno from @disk, do bdget() on it.
 586  *
 587  * CONTEXT:
 588  * Don't care.
 589  *
 590  * RETURNS:
 591  * Resulting block_device on success, NULL on failure.
 592  */
 593 struct block_device *bdget_disk(struct gendisk *disk, int partno)
 594 {
 595         struct hd_struct *part;
 596         struct block_device *bdev = NULL;
 597
 598         part = disk_get_part(disk, partno);
 599         if (part)
 600                 bdev = bdget(part_devt(part));
 601         disk_put_part(part);
 602
 603         return bdev;
 604 }
 605 EXPORT_SYMBOL(bdget_disk);
 606
 607 /*
 608  * print a full list of all partitions - intended for places where the root
 609  * filesystem can't be mounted and thus to give the victim some idea of what
 610  * went wrong
 611  */
 612 void __init printk_all_partitions(void)
 613 {
 614         struct class_dev_iter iter;
 615         struct device *dev;
 616
 617         class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
 618         while ((dev = class_dev_iter_next(&iter))) {
 619                 struct gendisk *disk = dev_to_disk(dev);
 620                 struct disk_part_iter piter;
 621                 struct hd_struct *part;
 622                 char name_buf[BDEVNAME_SIZE];
 623                 char devt_buf[BDEVT_SIZE];
 624
 625                 /*
 626                  * Don't show empty devices or things that have been
 627                  * surpressed
 628                  */
 629                 if (get_capacity(disk) == 0 ||
 630                     (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
 631                         continue;
 632
 633                 /*
 634                  * Note, unlike /proc/partitions, I am showing the
 635                  * numbers in hex - the same format as the root=
 636                  * option takes.
 637                  */
 638                 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0);
 639                 while ((part = disk_part_iter_next(&piter))) {
 640                         bool is_part0 = part == &disk->part0;
 641
 642                         printk("%s%s %10llu %s", is_part0 ? "" : "  ",
 643                                bdevt_str(part_devt(part), devt_buf),
 644                                (unsigned long long)part->nr_sects >> 1,
 645                                disk_name(disk, part->partno, name_buf));
 646                         if (is_part0) {
 647                                 if (disk->driverfs_dev != NULL &&
 648                                     disk->driverfs_dev->driver != NULL)
 649                                         printk(" driver: %s\n",
 650                                               disk->driverfs_dev->driver->name);
 651                                 else
 652                                         printk(" (driver?)\n");
 653                         } else
 654                                 printk("\n");
 655                 }
 656                 disk_part_iter_exit(&piter);
 657         }
 658         class_dev_iter_exit(&iter);
 659 }
 660
 661 #ifdef CONFIG_PROC_FS
 662 /* iterator */
 663 static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
 664 {
 665         loff_t skip = *pos;
 666         struct class_dev_iter *iter;
 667         struct device *dev;
 668
 669         iter = kmalloc(sizeof(*iter), GFP_KERNEL);
 670         if (!iter)
 671                 return ERR_PTR(-ENOMEM);
 672
 673         seqf->private = iter;
 674         class_dev_iter_init(iter, &block_class, NULL, &disk_type);
 675         do {
 676                 dev = class_dev_iter_next(iter);
 677                 if (!dev)
 678                         return NULL;
 679         } while (skip--);
 680
 681         return dev_to_disk(dev);
 682 }
 683
 684 static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
 685 {
 686         struct device *dev;
 687
 688         (*pos)++;
 689         dev = class_dev_iter_next(seqf->private);
 690         if (dev)
 691                 return dev_to_disk(dev);
 692
 693         return NULL;
 694 }
 695
 696 static void disk_seqf_stop(struct seq_file *seqf, void *v)
 697 {
 698         struct class_dev_iter *iter = seqf->private;
 699
 700         /* stop is called even after start failed :-( */
 701         if (iter) {
 702                 class_dev_iter_exit(iter);
 703                 kfree(iter);
 704         }
 705 }
 706
 707 static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
 708 {
 709         static void *p;
 710
 711         p = disk_seqf_start(seqf, pos);
 712         if (!IS_ERR(p) && p && !*pos)
 713                 seq_puts(seqf, "major minor  #blocks  name\n\n");
 714         return p;
 715 }
 716
 717 static int show_partition(struct seq_file *seqf, void *v)
 718 {
 719         struct gendisk *sgp = v;
 720         struct disk_part_iter piter;
 721         struct hd_struct *part;
 722         char buf[BDEVNAME_SIZE];
 723
 724         /* Don't show non-partitionable removeable devices or empty devices */
 725         if (!get_capacity(sgp) || (!disk_partitionable(sgp) &&
 726                                    (sgp->flags & GENHD_FL_REMOVABLE)))
 727                 return 0;
 728         if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
 729                 return 0;
 730
 731         /* show the full disk and all non-0 size partitions of it */
 732         disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0);
 733         while ((part = disk_part_iter_next(&piter)))
 734                 seq_printf(seqf, "%4d  %7d %10llu %s\n",
 735                            MAJOR(part_devt(part)), MINOR(part_devt(part)),
 736                            (unsigned long long)part->nr_sects >> 1,
 737                            disk_name(sgp, part->partno, buf));
 738         disk_part_iter_exit(&piter);
 739
 740         return 0;
 741 }
 742
 743 static const struct seq_operations partitions_op = {
 744         .start  = show_partition_start,
 745         .next   = disk_seqf_next,
 746         .stop   = disk_seqf_stop,
 747         .show   = show_partition
 748 };
 749
 750 static int partitions_open(struct inode *inode, struct file *file)
 751 {
 752         return seq_open(file, &partitions_op);
 753 }
 754
 755 static const struct file_operations proc_partitions_operations = {
 756         .open           = partitions_open,
 757         .read           = seq_read,
 758         .llseek         = seq_lseek,
 759         .release        = seq_release,
 760 };
 761 #endif
 762
 763
 764 static struct kobject *base_probe(dev_t devt, int *partno, void *data)
 765 {
 766         if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
 767                 /* Make old-style 2.4 aliases work */
 768                 request_module("block-major-%d", MAJOR(devt));
 769         return NULL;
 770 }
 771
 772 static int __init genhd_device_init(void)
 773 {
 774         int error;
 775
 776         block_class.dev_kobj = sysfs_dev_block_kobj;
 777         error = class_register(&block_class);
 778         if (unlikely(error))
 779                 return error;
 780         bdev_map = kobj_map_init(base_probe, &block_class_lock);
 781         blk_dev_init();
 782
 783         register_blkdev(BLOCK_EXT_MAJOR, "blkext");
 784
 785 #ifndef CONFIG_SYSFS_DEPRECATED
 786         /* create top-level block dir */
 787         block_depr = kobject_create_and_add("block", NULL);
 788 #endif
 789         return 0;
 790 }
 791
 792 subsys_initcall(genhd_device_init);
 793
 794 static ssize_t disk_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->minors);
 800 }
 801
 802 static ssize_t disk_ext_range_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", disk_max_parts(disk));
 808 }
 809
 810 static ssize_t disk_removable_show(struct device *dev,
 811                                    struct device_attribute *attr, char *buf)
 812 {
 813         struct gendisk *disk = dev_to_disk(dev);
 814
 815         return sprintf(buf, "%d\n",
 816                        (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
 817 }
 818
 819 static ssize_t disk_ro_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, "%d\n", get_disk_ro(disk) ? 1 : 0);
 825 }
 826
 827 static ssize_t disk_capability_show(struct device *dev,
 828                                     struct device_attribute *attr, char *buf)
 829 {
 830         struct gendisk *disk = dev_to_disk(dev);
 831
 832         return sprintf(buf, "%x\n", disk->flags);
 833 }
 834
 835 static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL);
 836 static DEVICE_ATTR(ext_range, S_IRUGO, disk_ext_range_show, NULL);
 837 static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL);
 838 static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL);
 839 static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL);
 840 static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL);
 841 static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL);
 842 #ifdef CONFIG_FAIL_MAKE_REQUEST
 843 static struct device_attribute dev_attr_fail =
 844         __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store);
 845 #endif
 846 #ifdef CONFIG_FAIL_IO_TIMEOUT
 847 static struct device_attribute dev_attr_fail_timeout =
 848         __ATTR(io-timeout-fail,  S_IRUGO|S_IWUSR, part_timeout_show,
 849                 part_timeout_store);
 850 #endif
 851
 852 static struct attribute *disk_attrs[] = {
 853         &dev_attr_range.attr,
 854         &dev_attr_ext_range.attr,
 855         &dev_attr_removable.attr,
 856         &dev_attr_ro.attr,
 857         &dev_attr_size.attr,
 858         &dev_attr_capability.attr,
 859         &dev_attr_stat.attr,
 860 #ifdef CONFIG_FAIL_MAKE_REQUEST
 861         &dev_attr_fail.attr,
 862 #endif
 863 #ifdef CONFIG_FAIL_IO_TIMEOUT
 864         &dev_attr_fail_timeout.attr,
 865 #endif
 866         NULL
 867 };
 868
 869 static struct attribute_group disk_attr_group = {
 870         .attrs = disk_attrs,
 871 };
 872
 873 static struct attribute_group *disk_attr_groups[] = {
 874         &disk_attr_group,
 875         NULL
 876 };
 877
 878 static void disk_free_ptbl_rcu_cb(struct rcu_head *head)
 879 {
 880         struct disk_part_tbl *ptbl =
 881                 container_of(head, struct disk_part_tbl, rcu_head);
 882
 883         kfree(ptbl);
 884 }
 885
 886 /**
 887  * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way
 888  * @disk: disk to replace part_tbl for
 889  * @new_ptbl: new part_tbl to install
 890  *
 891  * Replace disk->part_tbl with @new_ptbl in RCU-safe way.  The
 892  * original ptbl is freed using RCU callback.
 893  *
 894  * LOCKING:
 895  * Matching bd_mutx locked.
 896  */
 897 static void disk_replace_part_tbl(struct gendisk *disk,
 898                                   struct disk_part_tbl *new_ptbl)
 899 {
 900         struct disk_part_tbl *old_ptbl = disk->part_tbl;
 901
 902         rcu_assign_pointer(disk->part_tbl, new_ptbl);
 903
 904         if (old_ptbl) {
 905                 rcu_assign_pointer(old_ptbl->last_lookup, NULL);
 906                 call_rcu(&old_ptbl->rcu_head, disk_free_ptbl_rcu_cb);
 907         }
 908 }
 909
 910 /**
 911  * disk_expand_part_tbl - expand disk->part_tbl
 912  * @disk: disk to expand part_tbl for
 913  * @partno: expand such that this partno can fit in
 914  *
 915  * Expand disk->part_tbl such that @partno can fit in.  disk->part_tbl
 916  * uses RCU to allow unlocked dereferencing for stats and other stuff.
 917  *
 918  * LOCKING:
 919  * Matching bd_mutex locked, might sleep.
 920  *
 921  * RETURNS:
 922  * 0 on success, -errno on failure.
 923  */
 924 int disk_expand_part_tbl(struct gendisk *disk, int partno)
 925 {
 926         struct disk_part_tbl *old_ptbl = disk->part_tbl;
 927         struct disk_part_tbl *new_ptbl;
 928         int len = old_ptbl ? old_ptbl->len : 0;
 929         int target = partno + 1;
 930         size_t size;
 931         int i;
 932
 933         /* disk_max_parts() is zero during initialization, ignore if so */
 934         if (disk_max_parts(disk) && target > disk_max_parts(disk))
 935                 return -EINVAL;
 936
 937         if (target <= len)
 938                 return 0;
 939
 940         size = sizeof(*new_ptbl) + target * sizeof(new_ptbl->part[0]);
 941         new_ptbl = kzalloc_node(size, GFP_KERNEL, disk->node_id);
 942         if (!new_ptbl)
 943                 return -ENOMEM;
 944
 945         INIT_RCU_HEAD(&new_ptbl->rcu_head);
 946         new_ptbl->len = target;
 947
 948         for (i = 0; i < len; i++)
 949                 rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]);
 950
 951         disk_replace_part_tbl(disk, new_ptbl);
 952         return 0;
 953 }
 954
 955 static void disk_release(struct device *dev)
 956 {
 957         struct gendisk *disk = dev_to_disk(dev);
 958
 959         kfree(disk->random);
 960         disk_replace_part_tbl(disk, NULL);
 961         free_part_stats(&disk->part0);
 962         kfree(disk);
 963 }
 964 struct class block_class = {
 965         .name           = "block",
 966 };
 967
 968 static struct device_type disk_type = {
 969         .name           = "disk",
 970         .groups         = disk_attr_groups,
 971         .release        = disk_release,
 972 };
 973
 974 #ifdef CONFIG_PROC_FS
 975 /*
 976  * aggregate disk stat collector.  Uses the same stats that the sysfs
 977  * entries do, above, but makes them available through one seq_file.
 978  *
 979  * The output looks suspiciously like /proc/partitions with a bunch of
 980  * extra fields.
 981  */
 982 static int diskstats_show(struct seq_file *seqf, void *v)
 983 {
 984         struct gendisk *gp = v;
 985         struct disk_part_iter piter;
 986         struct hd_struct *hd;
 987         char buf[BDEVNAME_SIZE];
 988         int cpu;
 989
 990         /*
 991         if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
 992                 seq_puts(seqf,  "major minor name"
 993                                 "     rio rmerge rsect ruse wio wmerge "
 994                                 "wsect wuse running use aveq"
 995                                 "\n\n");
 996         */
 997
 998         disk_part_iter_init(&piter, gp, DISK_PITER_INCL_PART0);
 999         while ((hd = disk_part_iter_next(&piter))) {
1000                 cpu = part_stat_lock();
1001                 part_round_stats(cpu, hd);
1002                 part_stat_unlock();
1003                 seq_printf(seqf, "%4d %7d %s %lu %lu %llu "
1004                            "%u %lu %lu %llu %u %u %u %u\n",
1005                            MAJOR(part_devt(hd)), MINOR(part_devt(hd)),
1006                            disk_name(gp, hd->partno, buf),
1007                            part_stat_read(hd, ios[0]),
1008                            part_stat_read(hd, merges[0]),
1009                            (unsigned long long)part_stat_read(hd, sectors[0]),
1010                            jiffies_to_msecs(part_stat_read(hd, ticks[0])),
1011                            part_stat_read(hd, ios[1]),
1012                            part_stat_read(hd, merges[1]),
1013                            (unsigned long long)part_stat_read(hd, sectors[1]),
1014                            jiffies_to_msecs(part_stat_read(hd, ticks[1])),
1015                            hd->in_flight,
1016                            jiffies_to_msecs(part_stat_read(hd, io_ticks)),
1017                            jiffies_to_msecs(part_stat_read(hd, time_in_queue))
1018                         );
1019         }
1020         disk_part_iter_exit(&piter);
1021
1022         return 0;
1023 }
1024
1025 static const struct seq_operations diskstats_op = {
1026         .start  = disk_seqf_start,
1027         .next   = disk_seqf_next,
1028         .stop   = disk_seqf_stop,
1029         .show   = diskstats_show
1030 };
1031
1032 static int diskstats_open(struct inode *inode, struct file *file)
1033 {
1034         return seq_open(file, &diskstats_op);
1035 }
1036
1037 static const struct file_operations proc_diskstats_operations = {
1038         .open           = diskstats_open,
1039         .read           = seq_read,
1040         .llseek         = seq_lseek,
1041         .release        = seq_release,
1042 };
1043
1044 static int __init proc_genhd_init(void)
1045 {
1046         proc_create("diskstats", 0, NULL, &proc_diskstats_operations);
1047         proc_create("partitions", 0, NULL, &proc_partitions_operations);
1048         return 0;
1049 }
1050 module_init(proc_genhd_init);
1051 #endif /* CONFIG_PROC_FS */
1052
1053 static void media_change_notify_thread(struct work_struct *work)
1054 {
1055         struct gendisk *gd = container_of(work, struct gendisk, async_notify);
1056         char event[] = "MEDIA_CHANGE=1";
1057         char *envp[] = { event, NULL };
1058
1059         /*
1060          * set enviroment vars to indicate which event this is for
1061          * so that user space will know to go check the media status.
1062          */
1063         kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1064         put_device(gd->driverfs_dev);
1065 }
1066
1067 #if 0
1068 void genhd_media_change_notify(struct gendisk *disk)
1069 {
1070         get_device(disk->driverfs_dev);
1071         schedule_work(&disk->async_notify);
1072 }
1073 EXPORT_SYMBOL_GPL(genhd_media_change_notify);
1074 #endif  /*  0  */
1075
1076 dev_t blk_lookup_devt(const char *name, int partno)
1077 {
1078         dev_t devt = MKDEV(0, 0);
1079         struct class_dev_iter iter;
1080         struct device *dev;
1081
1082         class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1083         while ((dev = class_dev_iter_next(&iter))) {
1084                 struct gendisk *disk = dev_to_disk(dev);
1085                 struct hd_struct *part;
1086
1087                 if (strcmp(dev_name(dev), name))
1088                         continue;
1089
1090                 part = disk_get_part(disk, partno);
1091                 if (part) {
1092                         devt = part_devt(part);
1093                         disk_put_part(part);
1094                         break;
1095                 }
1096                 disk_put_part(part);
1097         }
1098         class_dev_iter_exit(&iter);
1099         return devt;
1100 }
1101 EXPORT_SYMBOL(blk_lookup_devt);
1102
1103 struct gendisk *alloc_disk(int minors)
1104 {
1105         return alloc_disk_node(minors, -1);
1106 }
1107 EXPORT_SYMBOL(alloc_disk);
1108
1109 struct gendisk *alloc_disk_node(int minors, int node_id)
1110 {
1111         struct gendisk *disk;
1112
1113         disk = kmalloc_node(sizeof(struct gendisk),
1114                                 GFP_KERNEL | __GFP_ZERO, node_id);
1115         if (disk) {
1116                 if (!init_part_stats(&disk->part0)) {
1117                         kfree(disk);
1118                         return NULL;
1119                 }
1120                 disk->node_id = node_id;
1121                 if (disk_expand_part_tbl(disk, 0)) {
1122                         free_part_stats(&disk->part0);
1123                         kfree(disk);
1124                         return NULL;
1125                 }
1126                 disk->part_tbl->part[0] = &disk->part0;
1127
1128                 disk->minors = minors;
1129                 rand_initialize_disk(disk);
1130                 disk_to_dev(disk)->class = &block_class;
1131                 disk_to_dev(disk)->type = &disk_type;
1132                 device_initialize(disk_to_dev(disk));
1133                 INIT_WORK(&disk->async_notify,
1134                         media_change_notify_thread);
1135         }
1136         return disk;
1137 }
1138 EXPORT_SYMBOL(alloc_disk_node);
1139
1140 struct kobject *get_disk(struct gendisk *disk)
1141 {
1142         struct module *owner;
1143         struct kobject *kobj;
1144
1145         if (!disk->fops)
1146                 return NULL;
1147         owner = disk->fops->owner;
1148         if (owner && !try_module_get(owner))
1149                 return NULL;
1150         kobj = kobject_get(&disk_to_dev(disk)->kobj);
1151         if (kobj == NULL) {
1152                 module_put(owner);
1153                 return NULL;
1154         }
1155         return kobj;
1156
1157 }
1158
1159 EXPORT_SYMBOL(get_disk);
1160
1161 void put_disk(struct gendisk *disk)
1162 {
1163         if (disk)
1164                 kobject_put(&disk_to_dev(disk)->kobj);
1165 }
1166
1167 EXPORT_SYMBOL(put_disk);
1168
1169 void set_device_ro(struct block_device *bdev, int flag)
1170 {
1171         bdev->bd_part->policy = flag;
1172 }
1173
1174 EXPORT_SYMBOL(set_device_ro);
1175
1176 void set_disk_ro(struct gendisk *disk, int flag)
1177 {
1178         struct disk_part_iter piter;
1179         struct hd_struct *part;
1180
1181         disk_part_iter_init(&piter, disk,
1182                             DISK_PITER_INCL_EMPTY | DISK_PITER_INCL_PART0);
1183         while ((part = disk_part_iter_next(&piter)))
1184                 part->policy = flag;
1185         disk_part_iter_exit(&piter);
1186 }
1187
1188 EXPORT_SYMBOL(set_disk_ro);
1189
1190 int bdev_read_only(struct block_device *bdev)
1191 {
1192         if (!bdev)
1193                 return 0;
1194         return bdev->bd_part->policy;
1195 }
1196
1197 EXPORT_SYMBOL(bdev_read_only);
1198
1199 int invalidate_partition(struct gendisk *disk, int partno)
1200 {
1201         int res = 0;
1202         struct block_device *bdev = bdget_disk(disk, partno);
1203         if (bdev) {
1204                 fsync_bdev(bdev);
1205                 res = __invalidate_device(bdev);
1206                 bdput(bdev);
1207         }
1208         return res;
1209 }
1210
1211 EXPORT_SYMBOL(invalidate_partition);