1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * gendisk handling 4 * 5 * Portions Copyright (C) 2020 Christoph Hellwig 6 */ 7 8 #include <linux/module.h> 9 #include <linux/ctype.h> 10 #include <linux/fs.h> 11 #include <linux/genhd.h> 12 #include <linux/kdev_t.h> 13 #include <linux/kernel.h> 14 #include <linux/blkdev.h> 15 #include <linux/backing-dev.h> 16 #include <linux/init.h> 17 #include <linux/spinlock.h> 18 #include <linux/proc_fs.h> 19 #include <linux/seq_file.h> 20 #include <linux/slab.h> 21 #include <linux/kmod.h> 22 #include <linux/mutex.h> 23 #include <linux/idr.h> 24 #include <linux/log2.h> 25 #include <linux/pm_runtime.h> 26 #include <linux/badblocks.h> 27 28 #include "blk.h" 29 30 static struct kobject *block_depr; 31 32 /* 33 * Unique, monotonically increasing sequential number associated with block 34 * devices instances (i.e. incremented each time a device is attached). 35 * Associating uevents with block devices in userspace is difficult and racy: 36 * the uevent netlink socket is lossy, and on slow and overloaded systems has 37 * a very high latency. 38 * Block devices do not have exclusive owners in userspace, any process can set 39 * one up (e.g. loop devices). Moreover, device names can be reused (e.g. loop0 40 * can be reused again and again). 41 * A userspace process setting up a block device and watching for its events 42 * cannot thus reliably tell whether an event relates to the device it just set 43 * up or another earlier instance with the same name. 44 * This sequential number allows userspace processes to solve this problem, and 45 * uniquely associate an uevent to the lifetime to a device. 46 */ 47 static atomic64_t diskseq; 48 49 /* for extended dynamic devt allocation, currently only one major is used */ 50 #define NR_EXT_DEVT (1 << MINORBITS) 51 static DEFINE_IDA(ext_devt_ida); 52 53 void set_capacity(struct gendisk *disk, sector_t sectors) 54 { 55 struct block_device *bdev = disk->part0; 56 57 spin_lock(&bdev->bd_size_lock); 58 i_size_write(bdev->bd_inode, (loff_t)sectors << SECTOR_SHIFT); 59 spin_unlock(&bdev->bd_size_lock); 60 } 61 EXPORT_SYMBOL(set_capacity); 62 63 /* 64 * Set disk capacity and notify if the size is not currently zero and will not 65 * be set to zero. Returns true if a uevent was sent, otherwise false. 66 */ 67 bool set_capacity_and_notify(struct gendisk *disk, sector_t size) 68 { 69 sector_t capacity = get_capacity(disk); 70 char *envp[] = { "RESIZE=1", NULL }; 71 72 set_capacity(disk, size); 73 74 /* 75 * Only print a message and send a uevent if the gendisk is user visible 76 * and alive. This avoids spamming the log and udev when setting the 77 * initial capacity during probing. 78 */ 79 if (size == capacity || 80 !disk_live(disk) || 81 (disk->flags & GENHD_FL_HIDDEN)) 82 return false; 83 84 pr_info("%s: detected capacity change from %lld to %lld\n", 85 disk->disk_name, capacity, size); 86 87 /* 88 * Historically we did not send a uevent for changes to/from an empty 89 * device. 90 */ 91 if (!capacity || !size) 92 return false; 93 kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp); 94 return true; 95 } 96 EXPORT_SYMBOL_GPL(set_capacity_and_notify); 97 98 /* 99 * Format the device name of the indicated block device into the supplied buffer 100 * and return a pointer to that same buffer for convenience. 101 * 102 * Note: do not use this in new code, use the %pg specifier to sprintf and 103 * printk insted. 104 */ 105 const char *bdevname(struct block_device *bdev, char *buf) 106 { 107 struct gendisk *hd = bdev->bd_disk; 108 int partno = bdev->bd_partno; 109 110 if (!partno) 111 snprintf(buf, BDEVNAME_SIZE, "%s", hd->disk_name); 112 else if (isdigit(hd->disk_name[strlen(hd->disk_name)-1])) 113 snprintf(buf, BDEVNAME_SIZE, "%sp%d", hd->disk_name, partno); 114 else 115 snprintf(buf, BDEVNAME_SIZE, "%s%d", hd->disk_name, partno); 116 117 return buf; 118 } 119 EXPORT_SYMBOL(bdevname); 120 121 static void part_stat_read_all(struct block_device *part, 122 struct disk_stats *stat) 123 { 124 int cpu; 125 126 memset(stat, 0, sizeof(struct disk_stats)); 127 for_each_possible_cpu(cpu) { 128 struct disk_stats *ptr = per_cpu_ptr(part->bd_stats, cpu); 129 int group; 130 131 for (group = 0; group < NR_STAT_GROUPS; group++) { 132 stat->nsecs[group] += ptr->nsecs[group]; 133 stat->sectors[group] += ptr->sectors[group]; 134 stat->ios[group] += ptr->ios[group]; 135 stat->merges[group] += ptr->merges[group]; 136 } 137 138 stat->io_ticks += ptr->io_ticks; 139 } 140 } 141 142 static unsigned int part_in_flight(struct block_device *part) 143 { 144 unsigned int inflight = 0; 145 int cpu; 146 147 for_each_possible_cpu(cpu) { 148 inflight += part_stat_local_read_cpu(part, in_flight[0], cpu) + 149 part_stat_local_read_cpu(part, in_flight[1], cpu); 150 } 151 if ((int)inflight < 0) 152 inflight = 0; 153 154 return inflight; 155 } 156 157 static void part_in_flight_rw(struct block_device *part, 158 unsigned int inflight[2]) 159 { 160 int cpu; 161 162 inflight[0] = 0; 163 inflight[1] = 0; 164 for_each_possible_cpu(cpu) { 165 inflight[0] += part_stat_local_read_cpu(part, in_flight[0], cpu); 166 inflight[1] += part_stat_local_read_cpu(part, in_flight[1], cpu); 167 } 168 if ((int)inflight[0] < 0) 169 inflight[0] = 0; 170 if ((int)inflight[1] < 0) 171 inflight[1] = 0; 172 } 173 174 /* 175 * Can be deleted altogether. Later. 176 * 177 */ 178 #define BLKDEV_MAJOR_HASH_SIZE 255 179 static struct blk_major_name { 180 struct blk_major_name *next; 181 int major; 182 char name[16]; 183 void (*probe)(dev_t devt); 184 } *major_names[BLKDEV_MAJOR_HASH_SIZE]; 185 static DEFINE_MUTEX(major_names_lock); 186 static DEFINE_SPINLOCK(major_names_spinlock); 187 188 /* index in the above - for now: assume no multimajor ranges */ 189 static inline int major_to_index(unsigned major) 190 { 191 return major % BLKDEV_MAJOR_HASH_SIZE; 192 } 193 194 #ifdef CONFIG_PROC_FS 195 void blkdev_show(struct seq_file *seqf, off_t offset) 196 { 197 struct blk_major_name *dp; 198 199 spin_lock(&major_names_spinlock); 200 for (dp = major_names[major_to_index(offset)]; dp; dp = dp->next) 201 if (dp->major == offset) 202 seq_printf(seqf, "%3d %s\n", dp->major, dp->name); 203 spin_unlock(&major_names_spinlock); 204 } 205 #endif /* CONFIG_PROC_FS */ 206 207 /** 208 * __register_blkdev - register a new block device 209 * 210 * @major: the requested major device number [1..BLKDEV_MAJOR_MAX-1]. If 211 * @major = 0, try to allocate any unused major number. 212 * @name: the name of the new block device as a zero terminated string 213 * @probe: allback that is called on access to any minor number of @major 214 * 215 * The @name must be unique within the system. 216 * 217 * The return value depends on the @major input parameter: 218 * 219 * - if a major device number was requested in range [1..BLKDEV_MAJOR_MAX-1] 220 * then the function returns zero on success, or a negative error code 221 * - if any unused major number was requested with @major = 0 parameter 222 * then the return value is the allocated major number in range 223 * [1..BLKDEV_MAJOR_MAX-1] or a negative error code otherwise 224 * 225 * See Documentation/admin-guide/devices.txt for the list of allocated 226 * major numbers. 227 * 228 * Use register_blkdev instead for any new code. 229 */ 230 int __register_blkdev(unsigned int major, const char *name, 231 void (*probe)(dev_t devt)) 232 { 233 struct blk_major_name **n, *p; 234 int index, ret = 0; 235 236 mutex_lock(&major_names_lock); 237 238 /* temporary */ 239 if (major == 0) { 240 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) { 241 if (major_names[index] == NULL) 242 break; 243 } 244 245 if (index == 0) { 246 printk("%s: failed to get major for %s\n", 247 __func__, name); 248 ret = -EBUSY; 249 goto out; 250 } 251 major = index; 252 ret = major; 253 } 254 255 if (major >= BLKDEV_MAJOR_MAX) { 256 pr_err("%s: major requested (%u) is greater than the maximum (%u) for %s\n", 257 __func__, major, BLKDEV_MAJOR_MAX-1, name); 258 259 ret = -EINVAL; 260 goto out; 261 } 262 263 p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL); 264 if (p == NULL) { 265 ret = -ENOMEM; 266 goto out; 267 } 268 269 p->major = major; 270 p->probe = probe; 271 strlcpy(p->name, name, sizeof(p->name)); 272 p->next = NULL; 273 index = major_to_index(major); 274 275 spin_lock(&major_names_spinlock); 276 for (n = &major_names[index]; *n; n = &(*n)->next) { 277 if ((*n)->major == major) 278 break; 279 } 280 if (!*n) 281 *n = p; 282 else 283 ret = -EBUSY; 284 spin_unlock(&major_names_spinlock); 285 286 if (ret < 0) { 287 printk("register_blkdev: cannot get major %u for %s\n", 288 major, name); 289 kfree(p); 290 } 291 out: 292 mutex_unlock(&major_names_lock); 293 return ret; 294 } 295 EXPORT_SYMBOL(__register_blkdev); 296 297 void unregister_blkdev(unsigned int major, const char *name) 298 { 299 struct blk_major_name **n; 300 struct blk_major_name *p = NULL; 301 int index = major_to_index(major); 302 303 mutex_lock(&major_names_lock); 304 spin_lock(&major_names_spinlock); 305 for (n = &major_names[index]; *n; n = &(*n)->next) 306 if ((*n)->major == major) 307 break; 308 if (!*n || strcmp((*n)->name, name)) { 309 WARN_ON(1); 310 } else { 311 p = *n; 312 *n = p->next; 313 } 314 spin_unlock(&major_names_spinlock); 315 mutex_unlock(&major_names_lock); 316 kfree(p); 317 } 318 319 EXPORT_SYMBOL(unregister_blkdev); 320 321 int blk_alloc_ext_minor(void) 322 { 323 int idx; 324 325 idx = ida_alloc_range(&ext_devt_ida, 0, NR_EXT_DEVT, GFP_KERNEL); 326 if (idx == -ENOSPC) 327 return -EBUSY; 328 return idx; 329 } 330 331 void blk_free_ext_minor(unsigned int minor) 332 { 333 ida_free(&ext_devt_ida, minor); 334 } 335 336 static char *bdevt_str(dev_t devt, char *buf) 337 { 338 if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) { 339 char tbuf[BDEVT_SIZE]; 340 snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt)); 341 snprintf(buf, BDEVT_SIZE, "%-9s", tbuf); 342 } else 343 snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt)); 344 345 return buf; 346 } 347 348 void disk_uevent(struct gendisk *disk, enum kobject_action action) 349 { 350 struct block_device *part; 351 unsigned long idx; 352 353 rcu_read_lock(); 354 xa_for_each(&disk->part_tbl, idx, part) { 355 if (bdev_is_partition(part) && !bdev_nr_sectors(part)) 356 continue; 357 if (!kobject_get_unless_zero(&part->bd_device.kobj)) 358 continue; 359 360 rcu_read_unlock(); 361 kobject_uevent(bdev_kobj(part), action); 362 put_device(&part->bd_device); 363 rcu_read_lock(); 364 } 365 rcu_read_unlock(); 366 } 367 EXPORT_SYMBOL_GPL(disk_uevent); 368 369 static void disk_scan_partitions(struct gendisk *disk) 370 { 371 struct block_device *bdev; 372 373 if (!get_capacity(disk) || !disk_part_scan_enabled(disk)) 374 return; 375 376 set_bit(GD_NEED_PART_SCAN, &disk->state); 377 bdev = blkdev_get_by_dev(disk_devt(disk), FMODE_READ, NULL); 378 if (!IS_ERR(bdev)) 379 blkdev_put(bdev, FMODE_READ); 380 } 381 382 /** 383 * device_add_disk - add disk information to kernel list 384 * @parent: parent device for the disk 385 * @disk: per-device partitioning information 386 * @groups: Additional per-device sysfs groups 387 * 388 * This function registers the partitioning information in @disk 389 * with the kernel. 390 */ 391 int device_add_disk(struct device *parent, struct gendisk *disk, 392 const struct attribute_group **groups) 393 394 { 395 struct device *ddev = disk_to_dev(disk); 396 int ret; 397 398 /* 399 * The disk queue should now be all set with enough information about 400 * the device for the elevator code to pick an adequate default 401 * elevator if one is needed, that is, for devices requesting queue 402 * registration. 403 */ 404 elevator_init_mq(disk->queue); 405 406 /* 407 * If the driver provides an explicit major number it also must provide 408 * the number of minors numbers supported, and those will be used to 409 * setup the gendisk. 410 * Otherwise just allocate the device numbers for both the whole device 411 * and all partitions from the extended dev_t space. 412 */ 413 if (disk->major) { 414 if (WARN_ON(!disk->minors)) 415 return -EINVAL; 416 417 if (disk->minors > DISK_MAX_PARTS) { 418 pr_err("block: can't allocate more than %d partitions\n", 419 DISK_MAX_PARTS); 420 disk->minors = DISK_MAX_PARTS; 421 } 422 } else { 423 if (WARN_ON(disk->minors)) 424 return -EINVAL; 425 426 ret = blk_alloc_ext_minor(); 427 if (ret < 0) 428 return ret; 429 disk->major = BLOCK_EXT_MAJOR; 430 disk->first_minor = ret; 431 disk->flags |= GENHD_FL_EXT_DEVT; 432 } 433 434 ret = disk_alloc_events(disk); 435 if (ret) 436 goto out_free_ext_minor; 437 438 /* delay uevents, until we scanned partition table */ 439 dev_set_uevent_suppress(ddev, 1); 440 441 ddev->parent = parent; 442 ddev->groups = groups; 443 dev_set_name(ddev, "%s", disk->disk_name); 444 if (!(disk->flags & GENHD_FL_HIDDEN)) 445 ddev->devt = MKDEV(disk->major, disk->first_minor); 446 ret = device_add(ddev); 447 if (ret) 448 goto out_disk_release_events; 449 if (!sysfs_deprecated) { 450 ret = sysfs_create_link(block_depr, &ddev->kobj, 451 kobject_name(&ddev->kobj)); 452 if (ret) 453 goto out_device_del; 454 } 455 456 /* 457 * avoid probable deadlock caused by allocating memory with 458 * GFP_KERNEL in runtime_resume callback of its all ancestor 459 * devices 460 */ 461 pm_runtime_set_memalloc_noio(ddev, true); 462 463 ret = blk_integrity_add(disk); 464 if (ret) 465 goto out_del_block_link; 466 467 disk->part0->bd_holder_dir = 468 kobject_create_and_add("holders", &ddev->kobj); 469 if (!disk->part0->bd_holder_dir) 470 goto out_del_integrity; 471 disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj); 472 if (!disk->slave_dir) 473 goto out_put_holder_dir; 474 475 ret = bd_register_pending_holders(disk); 476 if (ret < 0) 477 goto out_put_slave_dir; 478 479 ret = blk_register_queue(disk); 480 if (ret) 481 goto out_put_slave_dir; 482 483 if (disk->flags & GENHD_FL_HIDDEN) { 484 /* 485 * Don't let hidden disks show up in /proc/partitions, 486 * and don't bother scanning for partitions either. 487 */ 488 disk->flags |= GENHD_FL_SUPPRESS_PARTITION_INFO; 489 disk->flags |= GENHD_FL_NO_PART_SCAN; 490 } else { 491 ret = bdi_register(disk->bdi, "%u:%u", 492 disk->major, disk->first_minor); 493 if (ret) 494 goto out_unregister_queue; 495 bdi_set_owner(disk->bdi, ddev); 496 ret = sysfs_create_link(&ddev->kobj, 497 &disk->bdi->dev->kobj, "bdi"); 498 if (ret) 499 goto out_unregister_bdi; 500 501 bdev_add(disk->part0, ddev->devt); 502 disk_scan_partitions(disk); 503 504 /* 505 * Announce the disk and partitions after all partitions are 506 * created. (for hidden disks uevents remain suppressed forever) 507 */ 508 dev_set_uevent_suppress(ddev, 0); 509 disk_uevent(disk, KOBJ_ADD); 510 } 511 512 disk_update_readahead(disk); 513 disk_add_events(disk); 514 return 0; 515 516 out_unregister_bdi: 517 if (!(disk->flags & GENHD_FL_HIDDEN)) 518 bdi_unregister(disk->bdi); 519 out_unregister_queue: 520 blk_unregister_queue(disk); 521 out_put_slave_dir: 522 kobject_put(disk->slave_dir); 523 out_put_holder_dir: 524 kobject_put(disk->part0->bd_holder_dir); 525 out_del_integrity: 526 blk_integrity_del(disk); 527 out_del_block_link: 528 if (!sysfs_deprecated) 529 sysfs_remove_link(block_depr, dev_name(ddev)); 530 out_device_del: 531 device_del(ddev); 532 out_disk_release_events: 533 disk_release_events(disk); 534 out_free_ext_minor: 535 if (disk->major == BLOCK_EXT_MAJOR) 536 blk_free_ext_minor(disk->first_minor); 537 return WARN_ON_ONCE(ret); /* keep until all callers handle errors */ 538 } 539 EXPORT_SYMBOL(device_add_disk); 540 541 /** 542 * del_gendisk - remove the gendisk 543 * @disk: the struct gendisk to remove 544 * 545 * Removes the gendisk and all its associated resources. This deletes the 546 * partitions associated with the gendisk, and unregisters the associated 547 * request_queue. 548 * 549 * This is the counter to the respective __device_add_disk() call. 550 * 551 * The final removal of the struct gendisk happens when its refcount reaches 0 552 * with put_disk(), which should be called after del_gendisk(), if 553 * __device_add_disk() was used. 554 * 555 * Drivers exist which depend on the release of the gendisk to be synchronous, 556 * it should not be deferred. 557 * 558 * Context: can sleep 559 */ 560 void del_gendisk(struct gendisk *disk) 561 { 562 might_sleep(); 563 564 if (WARN_ON_ONCE(!disk_live(disk) && !(disk->flags & GENHD_FL_HIDDEN))) 565 return; 566 567 blk_integrity_del(disk); 568 disk_del_events(disk); 569 570 mutex_lock(&disk->open_mutex); 571 remove_inode_hash(disk->part0->bd_inode); 572 blk_drop_partitions(disk); 573 mutex_unlock(&disk->open_mutex); 574 575 fsync_bdev(disk->part0); 576 __invalidate_device(disk->part0, true); 577 578 set_capacity(disk, 0); 579 580 if (!(disk->flags & GENHD_FL_HIDDEN)) { 581 sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi"); 582 583 /* 584 * Unregister bdi before releasing device numbers (as they can 585 * get reused and we'd get clashes in sysfs). 586 */ 587 bdi_unregister(disk->bdi); 588 } 589 590 blk_unregister_queue(disk); 591 592 kobject_put(disk->part0->bd_holder_dir); 593 kobject_put(disk->slave_dir); 594 595 part_stat_set_all(disk->part0, 0); 596 disk->part0->bd_stamp = 0; 597 if (!sysfs_deprecated) 598 sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk))); 599 pm_runtime_set_memalloc_noio(disk_to_dev(disk), false); 600 device_del(disk_to_dev(disk)); 601 } 602 EXPORT_SYMBOL(del_gendisk); 603 604 /* sysfs access to bad-blocks list. */ 605 static ssize_t disk_badblocks_show(struct device *dev, 606 struct device_attribute *attr, 607 char *page) 608 { 609 struct gendisk *disk = dev_to_disk(dev); 610 611 if (!disk->bb) 612 return sprintf(page, "\n"); 613 614 return badblocks_show(disk->bb, page, 0); 615 } 616 617 static ssize_t disk_badblocks_store(struct device *dev, 618 struct device_attribute *attr, 619 const char *page, size_t len) 620 { 621 struct gendisk *disk = dev_to_disk(dev); 622 623 if (!disk->bb) 624 return -ENXIO; 625 626 return badblocks_store(disk->bb, page, len, 0); 627 } 628 629 void blk_request_module(dev_t devt) 630 { 631 unsigned int major = MAJOR(devt); 632 struct blk_major_name **n; 633 634 mutex_lock(&major_names_lock); 635 for (n = &major_names[major_to_index(major)]; *n; n = &(*n)->next) { 636 if ((*n)->major == major && (*n)->probe) { 637 (*n)->probe(devt); 638 mutex_unlock(&major_names_lock); 639 return; 640 } 641 } 642 mutex_unlock(&major_names_lock); 643 644 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0) 645 /* Make old-style 2.4 aliases work */ 646 request_module("block-major-%d", MAJOR(devt)); 647 } 648 649 /* 650 * print a full list of all partitions - intended for places where the root 651 * filesystem can't be mounted and thus to give the victim some idea of what 652 * went wrong 653 */ 654 void __init printk_all_partitions(void) 655 { 656 struct class_dev_iter iter; 657 struct device *dev; 658 659 class_dev_iter_init(&iter, &block_class, NULL, &disk_type); 660 while ((dev = class_dev_iter_next(&iter))) { 661 struct gendisk *disk = dev_to_disk(dev); 662 struct block_device *part; 663 char devt_buf[BDEVT_SIZE]; 664 unsigned long idx; 665 666 /* 667 * Don't show empty devices or things that have been 668 * suppressed 669 */ 670 if (get_capacity(disk) == 0 || 671 (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)) 672 continue; 673 674 /* 675 * Note, unlike /proc/partitions, I am showing the numbers in 676 * hex - the same format as the root= option takes. 677 */ 678 rcu_read_lock(); 679 xa_for_each(&disk->part_tbl, idx, part) { 680 if (!bdev_nr_sectors(part)) 681 continue; 682 printk("%s%s %10llu %pg %s", 683 bdev_is_partition(part) ? " " : "", 684 bdevt_str(part->bd_dev, devt_buf), 685 bdev_nr_sectors(part) >> 1, part, 686 part->bd_meta_info ? 687 part->bd_meta_info->uuid : ""); 688 if (bdev_is_partition(part)) 689 printk("\n"); 690 else if (dev->parent && dev->parent->driver) 691 printk(" driver: %s\n", 692 dev->parent->driver->name); 693 else 694 printk(" (driver?)\n"); 695 } 696 rcu_read_unlock(); 697 } 698 class_dev_iter_exit(&iter); 699 } 700 701 #ifdef CONFIG_PROC_FS 702 /* iterator */ 703 static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos) 704 { 705 loff_t skip = *pos; 706 struct class_dev_iter *iter; 707 struct device *dev; 708 709 iter = kmalloc(sizeof(*iter), GFP_KERNEL); 710 if (!iter) 711 return ERR_PTR(-ENOMEM); 712 713 seqf->private = iter; 714 class_dev_iter_init(iter, &block_class, NULL, &disk_type); 715 do { 716 dev = class_dev_iter_next(iter); 717 if (!dev) 718 return NULL; 719 } while (skip--); 720 721 return dev_to_disk(dev); 722 } 723 724 static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos) 725 { 726 struct device *dev; 727 728 (*pos)++; 729 dev = class_dev_iter_next(seqf->private); 730 if (dev) 731 return dev_to_disk(dev); 732 733 return NULL; 734 } 735 736 static void disk_seqf_stop(struct seq_file *seqf, void *v) 737 { 738 struct class_dev_iter *iter = seqf->private; 739 740 /* stop is called even after start failed :-( */ 741 if (iter) { 742 class_dev_iter_exit(iter); 743 kfree(iter); 744 seqf->private = NULL; 745 } 746 } 747 748 static void *show_partition_start(struct seq_file *seqf, loff_t *pos) 749 { 750 void *p; 751 752 p = disk_seqf_start(seqf, pos); 753 if (!IS_ERR_OR_NULL(p) && !*pos) 754 seq_puts(seqf, "major minor #blocks name\n\n"); 755 return p; 756 } 757 758 static int show_partition(struct seq_file *seqf, void *v) 759 { 760 struct gendisk *sgp = v; 761 struct block_device *part; 762 unsigned long idx; 763 764 /* Don't show non-partitionable removeable devices or empty devices */ 765 if (!get_capacity(sgp) || (!disk_max_parts(sgp) && 766 (sgp->flags & GENHD_FL_REMOVABLE))) 767 return 0; 768 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO) 769 return 0; 770 771 rcu_read_lock(); 772 xa_for_each(&sgp->part_tbl, idx, part) { 773 if (!bdev_nr_sectors(part)) 774 continue; 775 seq_printf(seqf, "%4d %7d %10llu %pg\n", 776 MAJOR(part->bd_dev), MINOR(part->bd_dev), 777 bdev_nr_sectors(part) >> 1, part); 778 } 779 rcu_read_unlock(); 780 return 0; 781 } 782 783 static const struct seq_operations partitions_op = { 784 .start = show_partition_start, 785 .next = disk_seqf_next, 786 .stop = disk_seqf_stop, 787 .show = show_partition 788 }; 789 #endif 790 791 static int __init genhd_device_init(void) 792 { 793 int error; 794 795 block_class.dev_kobj = sysfs_dev_block_kobj; 796 error = class_register(&block_class); 797 if (unlikely(error)) 798 return error; 799 blk_dev_init(); 800 801 register_blkdev(BLOCK_EXT_MAJOR, "blkext"); 802 803 /* create top-level block dir */ 804 if (!sysfs_deprecated) 805 block_depr = kobject_create_and_add("block", NULL); 806 return 0; 807 } 808 809 subsys_initcall(genhd_device_init); 810 811 static ssize_t disk_range_show(struct device *dev, 812 struct device_attribute *attr, char *buf) 813 { 814 struct gendisk *disk = dev_to_disk(dev); 815 816 return sprintf(buf, "%d\n", disk->minors); 817 } 818 819 static ssize_t disk_ext_range_show(struct device *dev, 820 struct device_attribute *attr, char *buf) 821 { 822 struct gendisk *disk = dev_to_disk(dev); 823 824 return sprintf(buf, "%d\n", disk_max_parts(disk)); 825 } 826 827 static ssize_t disk_removable_show(struct device *dev, 828 struct device_attribute *attr, char *buf) 829 { 830 struct gendisk *disk = dev_to_disk(dev); 831 832 return sprintf(buf, "%d\n", 833 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0)); 834 } 835 836 static ssize_t disk_hidden_show(struct device *dev, 837 struct device_attribute *attr, char *buf) 838 { 839 struct gendisk *disk = dev_to_disk(dev); 840 841 return sprintf(buf, "%d\n", 842 (disk->flags & GENHD_FL_HIDDEN ? 1 : 0)); 843 } 844 845 static ssize_t disk_ro_show(struct device *dev, 846 struct device_attribute *attr, char *buf) 847 { 848 struct gendisk *disk = dev_to_disk(dev); 849 850 return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0); 851 } 852 853 ssize_t part_size_show(struct device *dev, 854 struct device_attribute *attr, char *buf) 855 { 856 return sprintf(buf, "%llu\n", bdev_nr_sectors(dev_to_bdev(dev))); 857 } 858 859 ssize_t part_stat_show(struct device *dev, 860 struct device_attribute *attr, char *buf) 861 { 862 struct block_device *bdev = dev_to_bdev(dev); 863 struct request_queue *q = bdev->bd_disk->queue; 864 struct disk_stats stat; 865 unsigned int inflight; 866 867 part_stat_read_all(bdev, &stat); 868 if (queue_is_mq(q)) 869 inflight = blk_mq_in_flight(q, bdev); 870 else 871 inflight = part_in_flight(bdev); 872 873 return sprintf(buf, 874 "%8lu %8lu %8llu %8u " 875 "%8lu %8lu %8llu %8u " 876 "%8u %8u %8u " 877 "%8lu %8lu %8llu %8u " 878 "%8lu %8u" 879 "\n", 880 stat.ios[STAT_READ], 881 stat.merges[STAT_READ], 882 (unsigned long long)stat.sectors[STAT_READ], 883 (unsigned int)div_u64(stat.nsecs[STAT_READ], NSEC_PER_MSEC), 884 stat.ios[STAT_WRITE], 885 stat.merges[STAT_WRITE], 886 (unsigned long long)stat.sectors[STAT_WRITE], 887 (unsigned int)div_u64(stat.nsecs[STAT_WRITE], NSEC_PER_MSEC), 888 inflight, 889 jiffies_to_msecs(stat.io_ticks), 890 (unsigned int)div_u64(stat.nsecs[STAT_READ] + 891 stat.nsecs[STAT_WRITE] + 892 stat.nsecs[STAT_DISCARD] + 893 stat.nsecs[STAT_FLUSH], 894 NSEC_PER_MSEC), 895 stat.ios[STAT_DISCARD], 896 stat.merges[STAT_DISCARD], 897 (unsigned long long)stat.sectors[STAT_DISCARD], 898 (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], NSEC_PER_MSEC), 899 stat.ios[STAT_FLUSH], 900 (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], NSEC_PER_MSEC)); 901 } 902 903 ssize_t part_inflight_show(struct device *dev, struct device_attribute *attr, 904 char *buf) 905 { 906 struct block_device *bdev = dev_to_bdev(dev); 907 struct request_queue *q = bdev->bd_disk->queue; 908 unsigned int inflight[2]; 909 910 if (queue_is_mq(q)) 911 blk_mq_in_flight_rw(q, bdev, inflight); 912 else 913 part_in_flight_rw(bdev, inflight); 914 915 return sprintf(buf, "%8u %8u\n", inflight[0], inflight[1]); 916 } 917 918 static ssize_t disk_capability_show(struct device *dev, 919 struct device_attribute *attr, char *buf) 920 { 921 struct gendisk *disk = dev_to_disk(dev); 922 923 return sprintf(buf, "%x\n", disk->flags); 924 } 925 926 static ssize_t disk_alignment_offset_show(struct device *dev, 927 struct device_attribute *attr, 928 char *buf) 929 { 930 struct gendisk *disk = dev_to_disk(dev); 931 932 return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue)); 933 } 934 935 static ssize_t disk_discard_alignment_show(struct device *dev, 936 struct device_attribute *attr, 937 char *buf) 938 { 939 struct gendisk *disk = dev_to_disk(dev); 940 941 return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue)); 942 } 943 944 static ssize_t diskseq_show(struct device *dev, 945 struct device_attribute *attr, char *buf) 946 { 947 struct gendisk *disk = dev_to_disk(dev); 948 949 return sprintf(buf, "%llu\n", disk->diskseq); 950 } 951 952 static DEVICE_ATTR(range, 0444, disk_range_show, NULL); 953 static DEVICE_ATTR(ext_range, 0444, disk_ext_range_show, NULL); 954 static DEVICE_ATTR(removable, 0444, disk_removable_show, NULL); 955 static DEVICE_ATTR(hidden, 0444, disk_hidden_show, NULL); 956 static DEVICE_ATTR(ro, 0444, disk_ro_show, NULL); 957 static DEVICE_ATTR(size, 0444, part_size_show, NULL); 958 static DEVICE_ATTR(alignment_offset, 0444, disk_alignment_offset_show, NULL); 959 static DEVICE_ATTR(discard_alignment, 0444, disk_discard_alignment_show, NULL); 960 static DEVICE_ATTR(capability, 0444, disk_capability_show, NULL); 961 static DEVICE_ATTR(stat, 0444, part_stat_show, NULL); 962 static DEVICE_ATTR(inflight, 0444, part_inflight_show, NULL); 963 static DEVICE_ATTR(badblocks, 0644, disk_badblocks_show, disk_badblocks_store); 964 static DEVICE_ATTR(diskseq, 0444, diskseq_show, NULL); 965 966 #ifdef CONFIG_FAIL_MAKE_REQUEST 967 ssize_t part_fail_show(struct device *dev, 968 struct device_attribute *attr, char *buf) 969 { 970 return sprintf(buf, "%d\n", dev_to_bdev(dev)->bd_make_it_fail); 971 } 972 973 ssize_t part_fail_store(struct device *dev, 974 struct device_attribute *attr, 975 const char *buf, size_t count) 976 { 977 int i; 978 979 if (count > 0 && sscanf(buf, "%d", &i) > 0) 980 dev_to_bdev(dev)->bd_make_it_fail = i; 981 982 return count; 983 } 984 985 static struct device_attribute dev_attr_fail = 986 __ATTR(make-it-fail, 0644, part_fail_show, part_fail_store); 987 #endif /* CONFIG_FAIL_MAKE_REQUEST */ 988 989 #ifdef CONFIG_FAIL_IO_TIMEOUT 990 static struct device_attribute dev_attr_fail_timeout = 991 __ATTR(io-timeout-fail, 0644, part_timeout_show, part_timeout_store); 992 #endif 993 994 static struct attribute *disk_attrs[] = { 995 &dev_attr_range.attr, 996 &dev_attr_ext_range.attr, 997 &dev_attr_removable.attr, 998 &dev_attr_hidden.attr, 999 &dev_attr_ro.attr, 1000 &dev_attr_size.attr, 1001 &dev_attr_alignment_offset.attr, 1002 &dev_attr_discard_alignment.attr, 1003 &dev_attr_capability.attr, 1004 &dev_attr_stat.attr, 1005 &dev_attr_inflight.attr, 1006 &dev_attr_badblocks.attr, 1007 &dev_attr_events.attr, 1008 &dev_attr_events_async.attr, 1009 &dev_attr_events_poll_msecs.attr, 1010 &dev_attr_diskseq.attr, 1011 #ifdef CONFIG_FAIL_MAKE_REQUEST 1012 &dev_attr_fail.attr, 1013 #endif 1014 #ifdef CONFIG_FAIL_IO_TIMEOUT 1015 &dev_attr_fail_timeout.attr, 1016 #endif 1017 NULL 1018 }; 1019 1020 static umode_t disk_visible(struct kobject *kobj, struct attribute *a, int n) 1021 { 1022 struct device *dev = container_of(kobj, typeof(*dev), kobj); 1023 struct gendisk *disk = dev_to_disk(dev); 1024 1025 if (a == &dev_attr_badblocks.attr && !disk->bb) 1026 return 0; 1027 return a->mode; 1028 } 1029 1030 static struct attribute_group disk_attr_group = { 1031 .attrs = disk_attrs, 1032 .is_visible = disk_visible, 1033 }; 1034 1035 static const struct attribute_group *disk_attr_groups[] = { 1036 &disk_attr_group, 1037 NULL 1038 }; 1039 1040 /** 1041 * disk_release - releases all allocated resources of the gendisk 1042 * @dev: the device representing this disk 1043 * 1044 * This function releases all allocated resources of the gendisk. 1045 * 1046 * Drivers which used __device_add_disk() have a gendisk with a request_queue 1047 * assigned. Since the request_queue sits on top of the gendisk for these 1048 * drivers we also call blk_put_queue() for them, and we expect the 1049 * request_queue refcount to reach 0 at this point, and so the request_queue 1050 * will also be freed prior to the disk. 1051 * 1052 * Context: can sleep 1053 */ 1054 static void disk_release(struct device *dev) 1055 { 1056 struct gendisk *disk = dev_to_disk(dev); 1057 1058 might_sleep(); 1059 1060 disk_release_events(disk); 1061 kfree(disk->random); 1062 xa_destroy(&disk->part_tbl); 1063 disk->queue->disk = NULL; 1064 blk_put_queue(disk->queue); 1065 iput(disk->part0->bd_inode); /* frees the disk */ 1066 } 1067 1068 static int block_uevent(struct device *dev, struct kobj_uevent_env *env) 1069 { 1070 struct gendisk *disk = dev_to_disk(dev); 1071 1072 return add_uevent_var(env, "DISKSEQ=%llu", disk->diskseq); 1073 } 1074 1075 struct class block_class = { 1076 .name = "block", 1077 .dev_uevent = block_uevent, 1078 }; 1079 1080 static char *block_devnode(struct device *dev, umode_t *mode, 1081 kuid_t *uid, kgid_t *gid) 1082 { 1083 struct gendisk *disk = dev_to_disk(dev); 1084 1085 if (disk->fops->devnode) 1086 return disk->fops->devnode(disk, mode); 1087 return NULL; 1088 } 1089 1090 const struct device_type disk_type = { 1091 .name = "disk", 1092 .groups = disk_attr_groups, 1093 .release = disk_release, 1094 .devnode = block_devnode, 1095 }; 1096 1097 #ifdef CONFIG_PROC_FS 1098 /* 1099 * aggregate disk stat collector. Uses the same stats that the sysfs 1100 * entries do, above, but makes them available through one seq_file. 1101 * 1102 * The output looks suspiciously like /proc/partitions with a bunch of 1103 * extra fields. 1104 */ 1105 static int diskstats_show(struct seq_file *seqf, void *v) 1106 { 1107 struct gendisk *gp = v; 1108 struct block_device *hd; 1109 unsigned int inflight; 1110 struct disk_stats stat; 1111 unsigned long idx; 1112 1113 /* 1114 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next) 1115 seq_puts(seqf, "major minor name" 1116 " rio rmerge rsect ruse wio wmerge " 1117 "wsect wuse running use aveq" 1118 "\n\n"); 1119 */ 1120 1121 rcu_read_lock(); 1122 xa_for_each(&gp->part_tbl, idx, hd) { 1123 if (bdev_is_partition(hd) && !bdev_nr_sectors(hd)) 1124 continue; 1125 part_stat_read_all(hd, &stat); 1126 if (queue_is_mq(gp->queue)) 1127 inflight = blk_mq_in_flight(gp->queue, hd); 1128 else 1129 inflight = part_in_flight(hd); 1130 1131 seq_printf(seqf, "%4d %7d %pg " 1132 "%lu %lu %lu %u " 1133 "%lu %lu %lu %u " 1134 "%u %u %u " 1135 "%lu %lu %lu %u " 1136 "%lu %u" 1137 "\n", 1138 MAJOR(hd->bd_dev), MINOR(hd->bd_dev), hd, 1139 stat.ios[STAT_READ], 1140 stat.merges[STAT_READ], 1141 stat.sectors[STAT_READ], 1142 (unsigned int)div_u64(stat.nsecs[STAT_READ], 1143 NSEC_PER_MSEC), 1144 stat.ios[STAT_WRITE], 1145 stat.merges[STAT_WRITE], 1146 stat.sectors[STAT_WRITE], 1147 (unsigned int)div_u64(stat.nsecs[STAT_WRITE], 1148 NSEC_PER_MSEC), 1149 inflight, 1150 jiffies_to_msecs(stat.io_ticks), 1151 (unsigned int)div_u64(stat.nsecs[STAT_READ] + 1152 stat.nsecs[STAT_WRITE] + 1153 stat.nsecs[STAT_DISCARD] + 1154 stat.nsecs[STAT_FLUSH], 1155 NSEC_PER_MSEC), 1156 stat.ios[STAT_DISCARD], 1157 stat.merges[STAT_DISCARD], 1158 stat.sectors[STAT_DISCARD], 1159 (unsigned int)div_u64(stat.nsecs[STAT_DISCARD], 1160 NSEC_PER_MSEC), 1161 stat.ios[STAT_FLUSH], 1162 (unsigned int)div_u64(stat.nsecs[STAT_FLUSH], 1163 NSEC_PER_MSEC) 1164 ); 1165 } 1166 rcu_read_unlock(); 1167 1168 return 0; 1169 } 1170 1171 static const struct seq_operations diskstats_op = { 1172 .start = disk_seqf_start, 1173 .next = disk_seqf_next, 1174 .stop = disk_seqf_stop, 1175 .show = diskstats_show 1176 }; 1177 1178 static int __init proc_genhd_init(void) 1179 { 1180 proc_create_seq("diskstats", 0, NULL, &diskstats_op); 1181 proc_create_seq("partitions", 0, NULL, &partitions_op); 1182 return 0; 1183 } 1184 module_init(proc_genhd_init); 1185 #endif /* CONFIG_PROC_FS */ 1186 1187 dev_t part_devt(struct gendisk *disk, u8 partno) 1188 { 1189 struct block_device *part; 1190 dev_t devt = 0; 1191 1192 rcu_read_lock(); 1193 part = xa_load(&disk->part_tbl, partno); 1194 if (part) 1195 devt = part->bd_dev; 1196 rcu_read_unlock(); 1197 1198 return devt; 1199 } 1200 1201 dev_t blk_lookup_devt(const char *name, int partno) 1202 { 1203 dev_t devt = MKDEV(0, 0); 1204 struct class_dev_iter iter; 1205 struct device *dev; 1206 1207 class_dev_iter_init(&iter, &block_class, NULL, &disk_type); 1208 while ((dev = class_dev_iter_next(&iter))) { 1209 struct gendisk *disk = dev_to_disk(dev); 1210 1211 if (strcmp(dev_name(dev), name)) 1212 continue; 1213 1214 if (partno < disk->minors) { 1215 /* We need to return the right devno, even 1216 * if the partition doesn't exist yet. 1217 */ 1218 devt = MKDEV(MAJOR(dev->devt), 1219 MINOR(dev->devt) + partno); 1220 } else { 1221 devt = part_devt(disk, partno); 1222 if (devt) 1223 break; 1224 } 1225 } 1226 class_dev_iter_exit(&iter); 1227 return devt; 1228 } 1229 1230 struct gendisk *__alloc_disk_node(struct request_queue *q, int node_id, 1231 struct lock_class_key *lkclass) 1232 { 1233 struct gendisk *disk; 1234 1235 if (!blk_get_queue(q)) 1236 return NULL; 1237 1238 disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id); 1239 if (!disk) 1240 goto out_put_queue; 1241 1242 disk->bdi = bdi_alloc(node_id); 1243 if (!disk->bdi) 1244 goto out_free_disk; 1245 1246 disk->part0 = bdev_alloc(disk, 0); 1247 if (!disk->part0) 1248 goto out_free_bdi; 1249 1250 disk->node_id = node_id; 1251 mutex_init(&disk->open_mutex); 1252 xa_init(&disk->part_tbl); 1253 if (xa_insert(&disk->part_tbl, 0, disk->part0, GFP_KERNEL)) 1254 goto out_destroy_part_tbl; 1255 1256 rand_initialize_disk(disk); 1257 disk_to_dev(disk)->class = &block_class; 1258 disk_to_dev(disk)->type = &disk_type; 1259 device_initialize(disk_to_dev(disk)); 1260 inc_diskseq(disk); 1261 disk->queue = q; 1262 q->disk = disk; 1263 lockdep_init_map(&disk->lockdep_map, "(bio completion)", lkclass, 0); 1264 #ifdef CONFIG_BLOCK_HOLDER_DEPRECATED 1265 INIT_LIST_HEAD(&disk->slave_bdevs); 1266 #endif 1267 return disk; 1268 1269 out_destroy_part_tbl: 1270 xa_destroy(&disk->part_tbl); 1271 iput(disk->part0->bd_inode); 1272 out_free_bdi: 1273 bdi_put(disk->bdi); 1274 out_free_disk: 1275 kfree(disk); 1276 out_put_queue: 1277 blk_put_queue(q); 1278 return NULL; 1279 } 1280 EXPORT_SYMBOL(__alloc_disk_node); 1281 1282 struct gendisk *__blk_alloc_disk(int node, struct lock_class_key *lkclass) 1283 { 1284 struct request_queue *q; 1285 struct gendisk *disk; 1286 1287 q = blk_alloc_queue(node); 1288 if (!q) 1289 return NULL; 1290 1291 disk = __alloc_disk_node(q, node, lkclass); 1292 if (!disk) { 1293 blk_cleanup_queue(q); 1294 return NULL; 1295 } 1296 return disk; 1297 } 1298 EXPORT_SYMBOL(__blk_alloc_disk); 1299 1300 /** 1301 * put_disk - decrements the gendisk refcount 1302 * @disk: the struct gendisk to decrement the refcount for 1303 * 1304 * This decrements the refcount for the struct gendisk. When this reaches 0 1305 * we'll have disk_release() called. 1306 * 1307 * Context: Any context, but the last reference must not be dropped from 1308 * atomic context. 1309 */ 1310 void put_disk(struct gendisk *disk) 1311 { 1312 if (disk) 1313 put_device(disk_to_dev(disk)); 1314 } 1315 EXPORT_SYMBOL(put_disk); 1316 1317 /** 1318 * blk_cleanup_disk - shutdown a gendisk allocated by blk_alloc_disk 1319 * @disk: gendisk to shutdown 1320 * 1321 * Mark the queue hanging off @disk DYING, drain all pending requests, then mark 1322 * the queue DEAD, destroy and put it and the gendisk structure. 1323 * 1324 * Context: can sleep 1325 */ 1326 void blk_cleanup_disk(struct gendisk *disk) 1327 { 1328 blk_cleanup_queue(disk->queue); 1329 put_disk(disk); 1330 } 1331 EXPORT_SYMBOL(blk_cleanup_disk); 1332 1333 static void set_disk_ro_uevent(struct gendisk *gd, int ro) 1334 { 1335 char event[] = "DISK_RO=1"; 1336 char *envp[] = { event, NULL }; 1337 1338 if (!ro) 1339 event[8] = '0'; 1340 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp); 1341 } 1342 1343 /** 1344 * set_disk_ro - set a gendisk read-only 1345 * @disk: gendisk to operate on 1346 * @read_only: %true to set the disk read-only, %false set the disk read/write 1347 * 1348 * This function is used to indicate whether a given disk device should have its 1349 * read-only flag set. set_disk_ro() is typically used by device drivers to 1350 * indicate whether the underlying physical device is write-protected. 1351 */ 1352 void set_disk_ro(struct gendisk *disk, bool read_only) 1353 { 1354 if (read_only) { 1355 if (test_and_set_bit(GD_READ_ONLY, &disk->state)) 1356 return; 1357 } else { 1358 if (!test_and_clear_bit(GD_READ_ONLY, &disk->state)) 1359 return; 1360 } 1361 set_disk_ro_uevent(disk, read_only); 1362 } 1363 EXPORT_SYMBOL(set_disk_ro); 1364 1365 int bdev_read_only(struct block_device *bdev) 1366 { 1367 return bdev->bd_read_only || get_disk_ro(bdev->bd_disk); 1368 } 1369 EXPORT_SYMBOL(bdev_read_only); 1370 1371 void inc_diskseq(struct gendisk *disk) 1372 { 1373 disk->diskseq = atomic64_inc_return(&diskseq); 1374 } 1375