block/genhd.c

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