drivers/block/genhd.c

   1 /*
   2  *  gendisk handling
   3  */
   4
   5 #include <linux/config.h>
   6 #include <linux/module.h>
   7 #include <linux/fs.h>
   8 #include <linux/genhd.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
  18 #define MAX_PROBE_HASH 255      /* random */
  19
  20 static struct subsystem block_subsys;
  21
  22 static DECLARE_MUTEX(block_subsys_sem);
  23
  24 /*
  25  * Can be deleted altogether. Later.
  26  *
  27  */
  28 static struct blk_major_name {
  29         struct blk_major_name *next;
  30         int major;
  31         char name[16];
  32 } *major_names[MAX_PROBE_HASH];
  33
  34 /* index in the above - for now: assume no multimajor ranges */
  35 static inline int major_to_index(int major)
  36 {
  37         return major % MAX_PROBE_HASH;
  38 }
  39
  40 #ifdef CONFIG_PROC_FS
  41 /* get block device names in somewhat random order */
  42 int get_blkdev_list(char *p)
  43 {
  44         struct blk_major_name *n;
  45         int i, len;
  46
  47         len = sprintf(p, "\nBlock devices:\n");
  48
  49         down(&block_subsys_sem);
  50         for (i = 0; i < ARRAY_SIZE(major_names); i++) {
  51                 for (n = major_names[i]; n; n = n->next)
  52                         len += sprintf(p+len, "%3d %s\n",
  53                                        n->major, n->name);
  54         }
  55         up(&block_subsys_sem);
  56
  57         return len;
  58 }
  59 #endif
  60
  61 int register_blkdev(unsigned int major, const char *name)
  62 {
  63         struct blk_major_name **n, *p;
  64         int index, ret = 0;
  65
  66         down(&block_subsys_sem);
  67
  68         /* temporary */
  69         if (major == 0) {
  70                 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
  71                         if (major_names[index] == NULL)
  72                                 break;
  73                 }
  74
  75                 if (index == 0) {
  76                         printk("register_blkdev: failed to get major for %s\n",
  77                                name);
  78                         ret = -EBUSY;
  79                         goto out;
  80                 }
  81                 major = index;
  82                 ret = major;
  83         }
  84
  85         p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
  86         if (p == NULL) {
  87                 ret = -ENOMEM;
  88                 goto out;
  89         }
  90
  91         p->major = major;
  92         strlcpy(p->name, name, sizeof(p->name));
  93         p->next = NULL;
  94         index = major_to_index(major);
  95
  96         for (n = &major_names[index]; *n; n = &(*n)->next) {
  97                 if ((*n)->major == major)
  98                         break;
  99         }
 100         if (!*n)
 101                 *n = p;
 102         else
 103                 ret = -EBUSY;
 104
 105         if (ret < 0) {
 106                 printk("register_blkdev: cannot get major %d for %s\n",
 107                        major, name);
 108                 kfree(p);
 109         }
 110 out:
 111         up(&block_subsys_sem);
 112         return ret;
 113 }
 114
 115 EXPORT_SYMBOL(register_blkdev);
 116
 117 /* todo: make void - error printk here */
 118 int unregister_blkdev(unsigned int major, const char *name)
 119 {
 120         struct blk_major_name **n;
 121         struct blk_major_name *p = NULL;
 122         int index = major_to_index(major);
 123         int ret = 0;
 124
 125         down(&block_subsys_sem);
 126         for (n = &major_names[index]; *n; n = &(*n)->next)
 127                 if ((*n)->major == major)
 128                         break;
 129         if (!*n || strcmp((*n)->name, name))
 130                 ret = -EINVAL;
 131         else {
 132                 p = *n;
 133                 *n = p->next;
 134         }
 135         up(&block_subsys_sem);
 136         kfree(p);
 137
 138         return ret;
 139 }
 140
 141 EXPORT_SYMBOL(unregister_blkdev);
 142
 143 static struct kobj_map *bdev_map;
 144
 145 /*
 146  * Register device numbers dev..(dev+range-1)
 147  * range must be nonzero
 148  * The hash chain is sorted on range, so that subranges can override.
 149  */
 150 void blk_register_region(dev_t dev, unsigned long range, struct module *module,
 151                          struct kobject *(*probe)(dev_t, int *, void *),
 152                          int (*lock)(dev_t, void *), void *data)
 153 {
 154         kobj_map(bdev_map, dev, range, module, probe, lock, data);
 155 }
 156
 157 EXPORT_SYMBOL(blk_register_region);
 158
 159 void blk_unregister_region(dev_t dev, unsigned long range)
 160 {
 161         kobj_unmap(bdev_map, dev, range);
 162 }
 163
 164 EXPORT_SYMBOL(blk_unregister_region);
 165
 166 static struct kobject *exact_match(dev_t dev, int *part, void *data)
 167 {
 168         struct gendisk *p = data;
 169         return &p->kobj;
 170 }
 171
 172 static int exact_lock(dev_t dev, void *data)
 173 {
 174         struct gendisk *p = data;
 175
 176         if (!get_disk(p))
 177                 return -1;
 178         return 0;
 179 }
 180
 181 /**
 182  * add_disk - add partitioning information to kernel list
 183  * @disk: per-device partitioning information
 184  *
 185  * This function registers the partitioning information in @disk
 186  * with the kernel.
 187  */
 188 void add_disk(struct gendisk *disk)
 189 {
 190         disk->flags |= GENHD_FL_UP;
 191         blk_register_region(MKDEV(disk->major, disk->first_minor),
 192                             disk->minors, NULL, exact_match, exact_lock, disk);
 193         register_disk(disk);
 194         blk_register_queue(disk);
 195 }
 196
 197 EXPORT_SYMBOL(add_disk);
 198 EXPORT_SYMBOL(del_gendisk);     /* in partitions/check.c */
 199
 200 void unlink_gendisk(struct gendisk *disk)
 201 {
 202         blk_unregister_queue(disk);
 203         blk_unregister_region(MKDEV(disk->major, disk->first_minor),
 204                               disk->minors);
 205 }
 206
 207 #define to_disk(obj) container_of(obj,struct gendisk,kobj)
 208
 209 /**
 210  * get_gendisk - get partitioning information for a given device
 211  * @dev: device to get partitioning information for
 212  *
 213  * This function gets the structure containing partitioning
 214  * information for the given device @dev.
 215  */
 216 struct gendisk *get_gendisk(dev_t dev, int *part)
 217 {
 218         struct kobject *kobj = kobj_lookup(bdev_map, dev, part);
 219         return  kobj ? to_disk(kobj) : NULL;
 220 }
 221
 222 #ifdef CONFIG_PROC_FS
 223 /* iterator */
 224 static void *part_start(struct seq_file *part, loff_t *pos)
 225 {
 226         struct list_head *p;
 227         loff_t l = *pos;
 228
 229         down(&block_subsys_sem);
 230         list_for_each(p, &block_subsys.kset.list)
 231                 if (!l--)
 232                         return list_entry(p, struct gendisk, kobj.entry);
 233         return NULL;
 234 }
 235
 236 static void *part_next(struct seq_file *part, void *v, loff_t *pos)
 237 {
 238         struct list_head *p = ((struct gendisk *)v)->kobj.entry.next;
 239         ++*pos;
 240         return p==&block_subsys.kset.list ? NULL :
 241                 list_entry(p, struct gendisk, kobj.entry);
 242 }
 243
 244 static void part_stop(struct seq_file *part, void *v)
 245 {
 246         up(&block_subsys_sem);
 247 }
 248
 249 static int show_partition(struct seq_file *part, void *v)
 250 {
 251         struct gendisk *sgp = v;
 252         int n;
 253         char buf[BDEVNAME_SIZE];
 254
 255         if (&sgp->kobj.entry == block_subsys.kset.list.next)
 256                 seq_puts(part, "major minor  #blocks  name\n\n");
 257
 258         /* Don't show non-partitionable removeable devices or empty devices */
 259         if (!get_capacity(sgp) ||
 260                         (sgp->minors == 1 && (sgp->flags & GENHD_FL_REMOVABLE)))
 261                 return 0;
 262         if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
 263                 return 0;
 264
 265         /* show the full disk and all non-0 size partitions of it */
 266         seq_printf(part, "%4d  %4d %10llu %s\n",
 267                 sgp->major, sgp->first_minor,
 268                 (unsigned long long)get_capacity(sgp) >> 1,
 269                 disk_name(sgp, 0, buf));
 270         for (n = 0; n < sgp->minors - 1; n++) {
 271                 if (!sgp->part[n])
 272                         continue;
 273                 if (sgp->part[n]->nr_sects == 0)
 274                         continue;
 275                 seq_printf(part, "%4d  %4d %10llu %s\n",
 276                         sgp->major, n + 1 + sgp->first_minor,
 277                         (unsigned long long)sgp->part[n]->nr_sects >> 1 ,
 278                         disk_name(sgp, n + 1, buf));
 279         }
 280
 281         return 0;
 282 }
 283
 284 struct seq_operations partitions_op = {
 285         .start =part_start,
 286         .next = part_next,
 287         .stop = part_stop,
 288         .show = show_partition
 289 };
 290 #endif
 291
 292
 293 extern int blk_dev_init(void);
 294
 295 static struct kobject *base_probe(dev_t dev, int *part, void *data)
 296 {
 297         if (request_module("block-major-%d-%d", MAJOR(dev), MINOR(dev)) > 0)
 298                 /* Make old-style 2.4 aliases work */
 299                 request_module("block-major-%d", MAJOR(dev));
 300         return NULL;
 301 }
 302
 303 static int __init genhd_device_init(void)
 304 {
 305         bdev_map = kobj_map_init(base_probe, &block_subsys_sem);
 306         blk_dev_init();
 307         subsystem_register(&block_subsys);
 308         return 0;
 309 }
 310
 311 subsys_initcall(genhd_device_init);
 312
 313
 314
 315 /*
 316  * kobject & sysfs bindings for block devices
 317  */
 318 static ssize_t disk_attr_show(struct kobject *kobj, struct attribute *attr,
 319                               char *page)
 320 {
 321         struct gendisk *disk = to_disk(kobj);
 322         struct disk_attribute *disk_attr =
 323                 container_of(attr,struct disk_attribute,attr);
 324         ssize_t ret = 0;
 325
 326         if (disk_attr->show)
 327                 ret = disk_attr->show(disk,page);
 328         return ret;
 329 }
 330
 331 static struct sysfs_ops disk_sysfs_ops = {
 332         .show   = &disk_attr_show,
 333 };
 334
 335 static ssize_t disk_dev_read(struct gendisk * disk, char *page)
 336 {
 337         dev_t base = MKDEV(disk->major, disk->first_minor);
 338         return print_dev_t(page, base);
 339 }
 340 static ssize_t disk_range_read(struct gendisk * disk, char *page)
 341 {
 342         return sprintf(page, "%d\n", disk->minors);
 343 }
 344 static ssize_t disk_removable_read(struct gendisk * disk, char *page)
 345 {
 346         return sprintf(page, "%d\n",
 347                        (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
 348
 349 }
 350 static ssize_t disk_size_read(struct gendisk * disk, char *page)
 351 {
 352         return sprintf(page, "%llu\n", (unsigned long long)get_capacity(disk));
 353 }
 354
 355 static ssize_t disk_stats_read(struct gendisk * disk, char *page)
 356 {
 357         preempt_disable();
 358         disk_round_stats(disk);
 359         preempt_enable();
 360         return sprintf(page,
 361                 "%8u %8u %8llu %8u "
 362                 "%8u %8u %8llu %8u "
 363                 "%8u %8u %8u"
 364                 "\n",
 365                 disk_stat_read(disk, reads), disk_stat_read(disk, read_merges),
 366                 (unsigned long long)disk_stat_read(disk, read_sectors),
 367                 jiffies_to_msecs(disk_stat_read(disk, read_ticks)),
 368                 disk_stat_read(disk, writes),
 369                 disk_stat_read(disk, write_merges),
 370                 (unsigned long long)disk_stat_read(disk, write_sectors),
 371                 jiffies_to_msecs(disk_stat_read(disk, write_ticks)),
 372                 disk->in_flight,
 373                 jiffies_to_msecs(disk_stat_read(disk, io_ticks)),
 374                 jiffies_to_msecs(disk_stat_read(disk, time_in_queue)));
 375 }
 376 static struct disk_attribute disk_attr_dev = {
 377         .attr = {.name = "dev", .mode = S_IRUGO },
 378         .show   = disk_dev_read
 379 };
 380 static struct disk_attribute disk_attr_range = {
 381         .attr = {.name = "range", .mode = S_IRUGO },
 382         .show   = disk_range_read
 383 };
 384 static struct disk_attribute disk_attr_removable = {
 385         .attr = {.name = "removable", .mode = S_IRUGO },
 386         .show   = disk_removable_read
 387 };
 388 static struct disk_attribute disk_attr_size = {
 389         .attr = {.name = "size", .mode = S_IRUGO },
 390         .show   = disk_size_read
 391 };
 392 static struct disk_attribute disk_attr_stat = {
 393         .attr = {.name = "stat", .mode = S_IRUGO },
 394         .show   = disk_stats_read
 395 };
 396
 397 static struct attribute * default_attrs[] = {
 398         &disk_attr_dev.attr,
 399         &disk_attr_range.attr,
 400         &disk_attr_removable.attr,
 401         &disk_attr_size.attr,
 402         &disk_attr_stat.attr,
 403         NULL,
 404 };
 405
 406 static void disk_release(struct kobject * kobj)
 407 {
 408         struct gendisk *disk = to_disk(kobj);
 409         kfree(disk->random);
 410         kfree(disk->part);
 411         free_disk_stats(disk);
 412         kfree(disk);
 413 }
 414
 415 static struct kobj_type ktype_block = {
 416         .release        = disk_release,
 417         .sysfs_ops      = &disk_sysfs_ops,
 418         .default_attrs  = default_attrs,
 419 };
 420
 421 extern struct kobj_type ktype_part;
 422
 423 static int block_hotplug_filter(struct kset *kset, struct kobject *kobj)
 424 {
 425         struct kobj_type *ktype = get_ktype(kobj);
 426
 427         return ((ktype == &ktype_block) || (ktype == &ktype_part));
 428 }
 429
 430 static int block_hotplug(struct kset *kset, struct kobject *kobj, char **envp,
 431                          int num_envp, char *buffer, int buffer_size)
 432 {
 433         struct kobj_type *ktype = get_ktype(kobj);
 434         struct device *physdev;
 435         struct gendisk *disk;
 436         struct hd_struct *part;
 437         int length = 0;
 438         int i = 0;
 439
 440         if (ktype == &ktype_block) {
 441                 disk = container_of(kobj, struct gendisk, kobj);
 442                 add_hotplug_env_var(envp, num_envp, &i, buffer, buffer_size,
 443                                     &length, "MINOR=%u", disk->first_minor);
 444         } else if (ktype == &ktype_part) {
 445                 disk = container_of(kobj->parent, struct gendisk, kobj);
 446                 part = container_of(kobj, struct hd_struct, kobj);
 447                 add_hotplug_env_var(envp, num_envp, &i, buffer, buffer_size,
 448                                     &length, "MINOR=%u",
 449                                     disk->first_minor + part->partno);
 450         } else
 451                 return 0;
 452
 453         add_hotplug_env_var(envp, num_envp, &i, buffer, buffer_size, &length,
 454                             "MAJOR=%u", disk->major);
 455
 456         /* add physical device, backing this device  */
 457         physdev = disk->driverfs_dev;
 458         if (physdev) {
 459                 char *path = kobject_get_path(&physdev->kobj, GFP_KERNEL);
 460
 461                 add_hotplug_env_var(envp, num_envp, &i, buffer, buffer_size,
 462                                     &length, "PHYSDEVPATH=%s", path);
 463                 kfree(path);
 464
 465                 if (physdev->bus)
 466                         add_hotplug_env_var(envp, num_envp, &i,
 467                                             buffer, buffer_size, &length,
 468                                             "PHYSDEVBUS=%s",
 469                                             physdev->bus->name);
 470
 471                 if (physdev->driver)
 472                         add_hotplug_env_var(envp, num_envp, &i,
 473                                             buffer, buffer_size, &length,
 474                                             "PHYSDEVDRIVER=%s",
 475                                             physdev->driver->name);
 476         }
 477
 478         /* terminate, set to next free slot, shrink available space */
 479         envp[i] = NULL;
 480         envp = &envp[i];
 481         num_envp -= i;
 482         buffer = &buffer[length];
 483         buffer_size -= length;
 484
 485         return 0;
 486 }
 487
 488 static struct kset_hotplug_ops block_hotplug_ops = {
 489         .filter         = block_hotplug_filter,
 490         .hotplug        = block_hotplug,
 491 };
 492
 493 /* declare block_subsys. */
 494 static decl_subsys(block, &ktype_block, &block_hotplug_ops);
 495
 496
 497 /*
 498  * aggregate disk stat collector.  Uses the same stats that the sysfs
 499  * entries do, above, but makes them available through one seq_file.
 500  * Watching a few disks may be efficient through sysfs, but watching
 501  * all of them will be more efficient through this interface.
 502  *
 503  * The output looks suspiciously like /proc/partitions with a bunch of
 504  * extra fields.
 505  */
 506
 507 /* iterator */
 508 static void *diskstats_start(struct seq_file *part, loff_t *pos)
 509 {
 510         loff_t k = *pos;
 511         struct list_head *p;
 512
 513         down(&block_subsys_sem);
 514         list_for_each(p, &block_subsys.kset.list)
 515                 if (!k--)
 516                         return list_entry(p, struct gendisk, kobj.entry);
 517         return NULL;
 518 }
 519
 520 static void *diskstats_next(struct seq_file *part, void *v, loff_t *pos)
 521 {
 522         struct list_head *p = ((struct gendisk *)v)->kobj.entry.next;
 523         ++*pos;
 524         return p==&block_subsys.kset.list ? NULL :
 525                 list_entry(p, struct gendisk, kobj.entry);
 526 }
 527
 528 static void diskstats_stop(struct seq_file *part, void *v)
 529 {
 530         up(&block_subsys_sem);
 531 }
 532
 533 static int diskstats_show(struct seq_file *s, void *v)
 534 {
 535         struct gendisk *gp = v;
 536         char buf[BDEVNAME_SIZE];
 537         int n = 0;
 538
 539         /*
 540         if (&sgp->kobj.entry == block_subsys.kset.list.next)
 541                 seq_puts(s,     "major minor name"
 542                                 "     rio rmerge rsect ruse wio wmerge "
 543                                 "wsect wuse running use aveq"
 544                                 "\n\n");
 545         */
 546
 547         preempt_disable();
 548         disk_round_stats(gp);
 549         preempt_enable();
 550         seq_printf(s, "%4d %4d %s %u %u %llu %u %u %u %llu %u %u %u %u\n",
 551                 gp->major, n + gp->first_minor, disk_name(gp, n, buf),
 552                 disk_stat_read(gp, reads), disk_stat_read(gp, read_merges),
 553                 (unsigned long long)disk_stat_read(gp, read_sectors),
 554                 jiffies_to_msecs(disk_stat_read(gp, read_ticks)),
 555                 disk_stat_read(gp, writes), disk_stat_read(gp, write_merges),
 556                 (unsigned long long)disk_stat_read(gp, write_sectors),
 557                 jiffies_to_msecs(disk_stat_read(gp, write_ticks)),
 558                 gp->in_flight,
 559                 jiffies_to_msecs(disk_stat_read(gp, io_ticks)),
 560                 jiffies_to_msecs(disk_stat_read(gp, time_in_queue)));
 561
 562         /* now show all non-0 size partitions of it */
 563         for (n = 0; n < gp->minors - 1; n++) {
 564                 struct hd_struct *hd = gp->part[n];
 565
 566                 if (hd && hd->nr_sects)
 567                         seq_printf(s, "%4d %4d %s %u %u %u %u\n",
 568                                 gp->major, n + gp->first_minor + 1,
 569                                 disk_name(gp, n + 1, buf),
 570                                 hd->reads, hd->read_sectors,
 571                                 hd->writes, hd->write_sectors);
 572         }
 573
 574         return 0;
 575 }
 576
 577 struct seq_operations diskstats_op = {
 578         .start  = diskstats_start,
 579         .next   = diskstats_next,
 580         .stop   = diskstats_stop,
 581         .show   = diskstats_show
 582 };
 583
 584
 585 struct gendisk *alloc_disk(int minors)
 586 {
 587         struct gendisk *disk = kmalloc(sizeof(struct gendisk), GFP_KERNEL);
 588         if (disk) {
 589                 memset(disk, 0, sizeof(struct gendisk));
 590                 if (!init_disk_stats(disk)) {
 591                         kfree(disk);
 592                         return NULL;
 593                 }
 594                 if (minors > 1) {
 595                         int size = (minors - 1) * sizeof(struct hd_struct *);
 596                         disk->part = kmalloc(size, GFP_KERNEL);
 597                         if (!disk->part) {
 598                                 kfree(disk);
 599                                 return NULL;
 600                         }
 601                         memset(disk->part, 0, size);
 602                 }
 603                 disk->minors = minors;
 604                 kobj_set_kset_s(disk,block_subsys);
 605                 kobject_init(&disk->kobj);
 606                 rand_initialize_disk(disk);
 607         }
 608         return disk;
 609 }
 610
 611 EXPORT_SYMBOL(alloc_disk);
 612
 613 struct kobject *get_disk(struct gendisk *disk)
 614 {
 615         struct module *owner;
 616         struct kobject *kobj;
 617
 618         if (!disk->fops)
 619                 return NULL;
 620         owner = disk->fops->owner;
 621         if (owner && !try_module_get(owner))
 622                 return NULL;
 623         kobj = kobject_get(&disk->kobj);
 624         if (kobj == NULL) {
 625                 module_put(owner);
 626                 return NULL;
 627         }
 628         return kobj;
 629
 630 }
 631
 632 EXPORT_SYMBOL(get_disk);
 633
 634 void put_disk(struct gendisk *disk)
 635 {
 636         if (disk)
 637                 kobject_put(&disk->kobj);
 638 }
 639
 640 EXPORT_SYMBOL(put_disk);
 641
 642 void set_device_ro(struct block_device *bdev, int flag)
 643 {
 644         if (bdev->bd_contains != bdev)
 645                 bdev->bd_part->policy = flag;
 646         else
 647                 bdev->bd_disk->policy = flag;
 648 }
 649
 650 EXPORT_SYMBOL(set_device_ro);
 651
 652 void set_disk_ro(struct gendisk *disk, int flag)
 653 {
 654         int i;
 655         disk->policy = flag;
 656         for (i = 0; i < disk->minors - 1; i++)
 657                 if (disk->part[i]) disk->part[i]->policy = flag;
 658 }
 659
 660 EXPORT_SYMBOL(set_disk_ro);
 661
 662 int bdev_read_only(struct block_device *bdev)
 663 {
 664         if (!bdev)
 665                 return 0;
 666         else if (bdev->bd_contains != bdev)
 667                 return bdev->bd_part->policy;
 668         else
 669                 return bdev->bd_disk->policy;
 670 }
 671
 672 EXPORT_SYMBOL(bdev_read_only);
 673
 674 int invalidate_partition(struct gendisk *disk, int index)
 675 {
 676         int res = 0;
 677         struct block_device *bdev = bdget_disk(disk, index);
 678         if (bdev) {
 679                 res = __invalidate_device(bdev, 1);
 680                 bdput(bdev);
 681         }
 682         return res;
 683 }
 684
 685 EXPORT_SYMBOL(invalidate_partition);