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