]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blame - block/genhd.c
net: sh_eth: use ioremap()
[mirror_ubuntu-zesty-kernel.git] / block / genhd.c
CommitLineData
1da177e4
LT
1/*
2 * gendisk handling
3 */
4
1da177e4
LT
5#include <linux/module.h>
6#include <linux/fs.h>
7#include <linux/genhd.h>
b446b60e 8#include <linux/kdev_t.h>
1da177e4
LT
9#include <linux/kernel.h>
10#include <linux/blkdev.h>
11#include <linux/init.h>
12#include <linux/spinlock.h>
f500975a 13#include <linux/proc_fs.h>
1da177e4
LT
14#include <linux/seq_file.h>
15#include <linux/slab.h>
16#include <linux/kmod.h>
17#include <linux/kobj_map.h>
2ef41634 18#include <linux/buffer_head.h>
58383af6 19#include <linux/mutex.h>
bcce3de1 20#include <linux/idr.h>
77ea887e 21#include <linux/log2.h>
1da177e4 22
ff88972c
AB
23#include "blk.h"
24
edfaa7c3 25static DEFINE_MUTEX(block_class_lock);
edfaa7c3 26struct kobject *block_depr;
1da177e4 27
bcce3de1
TH
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 */
34static DEFINE_MUTEX(ext_devt_mutex);
35static DEFINE_IDR(ext_devt_idr);
36
1826eadf
AB
37static struct device_type disk_type;
38
77ea887e
TH
39static void disk_add_events(struct gendisk *disk);
40static void disk_del_events(struct gendisk *disk);
41static void disk_release_events(struct gendisk *disk);
42
e71bf0d0
TH
43/**
44 * disk_get_part - get partition
45 * @disk: disk to look partition from
46 * @partno: partition number
47 *
48 * Look for partition @partno from @disk. If found, increment
49 * reference count and return it.
50 *
51 * CONTEXT:
52 * Don't care.
53 *
54 * RETURNS:
55 * Pointer to the found partition on success, NULL if not found.
56 */
57struct hd_struct *disk_get_part(struct gendisk *disk, int partno)
58{
540eed56
TH
59 struct hd_struct *part = NULL;
60 struct disk_part_tbl *ptbl;
e71bf0d0 61
540eed56 62 if (unlikely(partno < 0))
e71bf0d0 63 return NULL;
540eed56 64
e71bf0d0 65 rcu_read_lock();
540eed56
TH
66
67 ptbl = rcu_dereference(disk->part_tbl);
68 if (likely(partno < ptbl->len)) {
69 part = rcu_dereference(ptbl->part[partno]);
70 if (part)
71 get_device(part_to_dev(part));
72 }
73
e71bf0d0
TH
74 rcu_read_unlock();
75
76 return part;
77}
78EXPORT_SYMBOL_GPL(disk_get_part);
79
80/**
81 * disk_part_iter_init - initialize partition iterator
82 * @piter: iterator to initialize
83 * @disk: disk to iterate over
84 * @flags: DISK_PITER_* flags
85 *
86 * Initialize @piter so that it iterates over partitions of @disk.
87 *
88 * CONTEXT:
89 * Don't care.
90 */
91void disk_part_iter_init(struct disk_part_iter *piter, struct gendisk *disk,
92 unsigned int flags)
93{
540eed56
TH
94 struct disk_part_tbl *ptbl;
95
96 rcu_read_lock();
97 ptbl = rcu_dereference(disk->part_tbl);
98
e71bf0d0
TH
99 piter->disk = disk;
100 piter->part = NULL;
101
102 if (flags & DISK_PITER_REVERSE)
540eed56 103 piter->idx = ptbl->len - 1;
71982a40 104 else if (flags & (DISK_PITER_INCL_PART0 | DISK_PITER_INCL_EMPTY_PART0))
e71bf0d0 105 piter->idx = 0;
b5d0b9df
TH
106 else
107 piter->idx = 1;
e71bf0d0
TH
108
109 piter->flags = flags;
540eed56
TH
110
111 rcu_read_unlock();
e71bf0d0
TH
112}
113EXPORT_SYMBOL_GPL(disk_part_iter_init);
114
115/**
116 * disk_part_iter_next - proceed iterator to the next partition and return it
117 * @piter: iterator of interest
118 *
119 * Proceed @piter to the next partition and return it.
120 *
121 * CONTEXT:
122 * Don't care.
123 */
124struct hd_struct *disk_part_iter_next(struct disk_part_iter *piter)
125{
540eed56 126 struct disk_part_tbl *ptbl;
e71bf0d0
TH
127 int inc, end;
128
129 /* put the last partition */
130 disk_put_part(piter->part);
131 piter->part = NULL;
132
540eed56 133 /* get part_tbl */
e71bf0d0 134 rcu_read_lock();
540eed56 135 ptbl = rcu_dereference(piter->disk->part_tbl);
e71bf0d0
TH
136
137 /* determine iteration parameters */
138 if (piter->flags & DISK_PITER_REVERSE) {
139 inc = -1;
71982a40
TH
140 if (piter->flags & (DISK_PITER_INCL_PART0 |
141 DISK_PITER_INCL_EMPTY_PART0))
b5d0b9df
TH
142 end = -1;
143 else
144 end = 0;
e71bf0d0
TH
145 } else {
146 inc = 1;
540eed56 147 end = ptbl->len;
e71bf0d0
TH
148 }
149
150 /* iterate to the next partition */
151 for (; piter->idx != end; piter->idx += inc) {
152 struct hd_struct *part;
153
540eed56 154 part = rcu_dereference(ptbl->part[piter->idx]);
e71bf0d0
TH
155 if (!part)
156 continue;
71982a40
TH
157 if (!part->nr_sects &&
158 !(piter->flags & DISK_PITER_INCL_EMPTY) &&
159 !(piter->flags & DISK_PITER_INCL_EMPTY_PART0 &&
160 piter->idx == 0))
e71bf0d0
TH
161 continue;
162
ed9e1982 163 get_device(part_to_dev(part));
e71bf0d0
TH
164 piter->part = part;
165 piter->idx += inc;
166 break;
167 }
168
169 rcu_read_unlock();
170
171 return piter->part;
172}
173EXPORT_SYMBOL_GPL(disk_part_iter_next);
174
175/**
176 * disk_part_iter_exit - finish up partition iteration
177 * @piter: iter of interest
178 *
179 * Called when iteration is over. Cleans up @piter.
180 *
181 * CONTEXT:
182 * Don't care.
183 */
184void disk_part_iter_exit(struct disk_part_iter *piter)
185{
186 disk_put_part(piter->part);
187 piter->part = NULL;
188}
189EXPORT_SYMBOL_GPL(disk_part_iter_exit);
190
a6f23657
JA
191static inline int sector_in_part(struct hd_struct *part, sector_t sector)
192{
193 return part->start_sect <= sector &&
194 sector < part->start_sect + part->nr_sects;
195}
196
e71bf0d0
TH
197/**
198 * disk_map_sector_rcu - map sector to partition
199 * @disk: gendisk of interest
200 * @sector: sector to map
201 *
202 * Find out which partition @sector maps to on @disk. This is
203 * primarily used for stats accounting.
204 *
205 * CONTEXT:
206 * RCU read locked. The returned partition pointer is valid only
207 * while preemption is disabled.
208 *
209 * RETURNS:
074a7aca 210 * Found partition on success, part0 is returned if no partition matches
e71bf0d0
TH
211 */
212struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector)
213{
540eed56 214 struct disk_part_tbl *ptbl;
a6f23657 215 struct hd_struct *part;
e71bf0d0
TH
216 int i;
217
540eed56
TH
218 ptbl = rcu_dereference(disk->part_tbl);
219
a6f23657
JA
220 part = rcu_dereference(ptbl->last_lookup);
221 if (part && sector_in_part(part, sector))
222 return part;
223
540eed56 224 for (i = 1; i < ptbl->len; i++) {
a6f23657 225 part = rcu_dereference(ptbl->part[i]);
e71bf0d0 226
a6f23657
JA
227 if (part && sector_in_part(part, sector)) {
228 rcu_assign_pointer(ptbl->last_lookup, part);
e71bf0d0 229 return part;
a6f23657 230 }
e71bf0d0 231 }
074a7aca 232 return &disk->part0;
e71bf0d0
TH
233}
234EXPORT_SYMBOL_GPL(disk_map_sector_rcu);
235
1da177e4
LT
236/*
237 * Can be deleted altogether. Later.
238 *
239 */
240static struct blk_major_name {
241 struct blk_major_name *next;
242 int major;
243 char name[16];
68eef3b4 244} *major_names[BLKDEV_MAJOR_HASH_SIZE];
1da177e4
LT
245
246/* index in the above - for now: assume no multimajor ranges */
e61eb2e9 247static inline int major_to_index(unsigned major)
1da177e4 248{
68eef3b4 249 return major % BLKDEV_MAJOR_HASH_SIZE;
7170be5f
NH
250}
251
68eef3b4 252#ifdef CONFIG_PROC_FS
cf771cb5 253void blkdev_show(struct seq_file *seqf, off_t offset)
7170be5f 254{
68eef3b4 255 struct blk_major_name *dp;
7170be5f 256
68eef3b4 257 if (offset < BLKDEV_MAJOR_HASH_SIZE) {
edfaa7c3 258 mutex_lock(&block_class_lock);
68eef3b4 259 for (dp = major_names[offset]; dp; dp = dp->next)
cf771cb5 260 seq_printf(seqf, "%3d %s\n", dp->major, dp->name);
edfaa7c3 261 mutex_unlock(&block_class_lock);
1da177e4 262 }
1da177e4 263}
68eef3b4 264#endif /* CONFIG_PROC_FS */
1da177e4 265
9e8c0bcc
MN
266/**
267 * register_blkdev - register a new block device
268 *
269 * @major: the requested major device number [1..255]. If @major=0, try to
270 * allocate any unused major number.
271 * @name: the name of the new block device as a zero terminated string
272 *
273 * The @name must be unique within the system.
274 *
275 * The return value depends on the @major input parameter.
276 * - if a major device number was requested in range [1..255] then the
277 * function returns zero on success, or a negative error code
278 * - if any unused major number was requested with @major=0 parameter
279 * then the return value is the allocated major number in range
280 * [1..255] or a negative error code otherwise
281 */
1da177e4
LT
282int register_blkdev(unsigned int major, const char *name)
283{
284 struct blk_major_name **n, *p;
285 int index, ret = 0;
286
edfaa7c3 287 mutex_lock(&block_class_lock);
1da177e4
LT
288
289 /* temporary */
290 if (major == 0) {
291 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
292 if (major_names[index] == NULL)
293 break;
294 }
295
296 if (index == 0) {
297 printk("register_blkdev: failed to get major for %s\n",
298 name);
299 ret = -EBUSY;
300 goto out;
301 }
302 major = index;
303 ret = major;
304 }
305
306 p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
307 if (p == NULL) {
308 ret = -ENOMEM;
309 goto out;
310 }
311
312 p->major = major;
313 strlcpy(p->name, name, sizeof(p->name));
314 p->next = NULL;
315 index = major_to_index(major);
316
317 for (n = &major_names[index]; *n; n = &(*n)->next) {
318 if ((*n)->major == major)
319 break;
320 }
321 if (!*n)
322 *n = p;
323 else
324 ret = -EBUSY;
325
326 if (ret < 0) {
327 printk("register_blkdev: cannot get major %d for %s\n",
328 major, name);
329 kfree(p);
330 }
331out:
edfaa7c3 332 mutex_unlock(&block_class_lock);
1da177e4
LT
333 return ret;
334}
335
336EXPORT_SYMBOL(register_blkdev);
337
f4480240 338void unregister_blkdev(unsigned int major, const char *name)
1da177e4
LT
339{
340 struct blk_major_name **n;
341 struct blk_major_name *p = NULL;
342 int index = major_to_index(major);
1da177e4 343
edfaa7c3 344 mutex_lock(&block_class_lock);
1da177e4
LT
345 for (n = &major_names[index]; *n; n = &(*n)->next)
346 if ((*n)->major == major)
347 break;
294462a5
AM
348 if (!*n || strcmp((*n)->name, name)) {
349 WARN_ON(1);
294462a5 350 } else {
1da177e4
LT
351 p = *n;
352 *n = p->next;
353 }
edfaa7c3 354 mutex_unlock(&block_class_lock);
1da177e4 355 kfree(p);
1da177e4
LT
356}
357
358EXPORT_SYMBOL(unregister_blkdev);
359
360static struct kobj_map *bdev_map;
361
870d6656
TH
362/**
363 * blk_mangle_minor - scatter minor numbers apart
364 * @minor: minor number to mangle
365 *
366 * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
367 * is enabled. Mangling twice gives the original value.
368 *
369 * RETURNS:
370 * Mangled value.
371 *
372 * CONTEXT:
373 * Don't care.
374 */
375static int blk_mangle_minor(int minor)
376{
377#ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
378 int i;
379
380 for (i = 0; i < MINORBITS / 2; i++) {
381 int low = minor & (1 << i);
382 int high = minor & (1 << (MINORBITS - 1 - i));
383 int distance = MINORBITS - 1 - 2 * i;
384
385 minor ^= low | high; /* clear both bits */
386 low <<= distance; /* swap the positions */
387 high >>= distance;
388 minor |= low | high; /* and set */
389 }
390#endif
391 return minor;
392}
393
bcce3de1
TH
394/**
395 * blk_alloc_devt - allocate a dev_t for a partition
396 * @part: partition to allocate dev_t for
bcce3de1
TH
397 * @devt: out parameter for resulting dev_t
398 *
399 * Allocate a dev_t for block device.
400 *
401 * RETURNS:
402 * 0 on success, allocated dev_t is returned in *@devt. -errno on
403 * failure.
404 *
405 * CONTEXT:
406 * Might sleep.
407 */
408int blk_alloc_devt(struct hd_struct *part, dev_t *devt)
409{
410 struct gendisk *disk = part_to_disk(part);
411 int idx, rc;
412
413 /* in consecutive minor range? */
414 if (part->partno < disk->minors) {
415 *devt = MKDEV(disk->major, disk->first_minor + part->partno);
416 return 0;
417 }
418
419 /* allocate ext devt */
420 do {
421 if (!idr_pre_get(&ext_devt_idr, GFP_KERNEL))
422 return -ENOMEM;
423 rc = idr_get_new(&ext_devt_idr, part, &idx);
424 } while (rc == -EAGAIN);
425
426 if (rc)
427 return rc;
428
429 if (idx > MAX_EXT_DEVT) {
430 idr_remove(&ext_devt_idr, idx);
431 return -EBUSY;
432 }
433
870d6656 434 *devt = MKDEV(BLOCK_EXT_MAJOR, blk_mangle_minor(idx));
bcce3de1
TH
435 return 0;
436}
437
438/**
439 * blk_free_devt - free a dev_t
440 * @devt: dev_t to free
441 *
442 * Free @devt which was allocated using blk_alloc_devt().
443 *
444 * CONTEXT:
445 * Might sleep.
446 */
447void blk_free_devt(dev_t devt)
448{
449 might_sleep();
450
451 if (devt == MKDEV(0, 0))
452 return;
453
454 if (MAJOR(devt) == BLOCK_EXT_MAJOR) {
455 mutex_lock(&ext_devt_mutex);
870d6656 456 idr_remove(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
bcce3de1
TH
457 mutex_unlock(&ext_devt_mutex);
458 }
459}
460
1f014290
TH
461static char *bdevt_str(dev_t devt, char *buf)
462{
463 if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
464 char tbuf[BDEVT_SIZE];
465 snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
466 snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
467 } else
468 snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
469
470 return buf;
471}
472
1da177e4
LT
473/*
474 * Register device numbers dev..(dev+range-1)
475 * range must be nonzero
476 * The hash chain is sorted on range, so that subranges can override.
477 */
edfaa7c3 478void blk_register_region(dev_t devt, unsigned long range, struct module *module,
1da177e4
LT
479 struct kobject *(*probe)(dev_t, int *, void *),
480 int (*lock)(dev_t, void *), void *data)
481{
edfaa7c3 482 kobj_map(bdev_map, devt, range, module, probe, lock, data);
1da177e4
LT
483}
484
485EXPORT_SYMBOL(blk_register_region);
486
edfaa7c3 487void blk_unregister_region(dev_t devt, unsigned long range)
1da177e4 488{
edfaa7c3 489 kobj_unmap(bdev_map, devt, range);
1da177e4
LT
490}
491
492EXPORT_SYMBOL(blk_unregister_region);
493
cf771cb5 494static struct kobject *exact_match(dev_t devt, int *partno, void *data)
1da177e4
LT
495{
496 struct gendisk *p = data;
edfaa7c3 497
ed9e1982 498 return &disk_to_dev(p)->kobj;
1da177e4
LT
499}
500
edfaa7c3 501static int exact_lock(dev_t devt, void *data)
1da177e4
LT
502{
503 struct gendisk *p = data;
504
505 if (!get_disk(p))
506 return -1;
507 return 0;
508}
509
d2bf1b67
TH
510void register_disk(struct gendisk *disk)
511{
512 struct device *ddev = disk_to_dev(disk);
513 struct block_device *bdev;
514 struct disk_part_iter piter;
515 struct hd_struct *part;
516 int err;
517
518 ddev->parent = disk->driverfs_dev;
519
520 dev_set_name(ddev, disk->disk_name);
521
522 /* delay uevents, until we scanned partition table */
523 dev_set_uevent_suppress(ddev, 1);
524
525 if (device_add(ddev))
526 return;
527 if (!sysfs_deprecated) {
528 err = sysfs_create_link(block_depr, &ddev->kobj,
529 kobject_name(&ddev->kobj));
530 if (err) {
531 device_del(ddev);
532 return;
533 }
534 }
535 disk->part0.holder_dir = kobject_create_and_add("holders", &ddev->kobj);
536 disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
537
538 /* No minors to use for partitions */
539 if (!disk_partitionable(disk))
540 goto exit;
541
542 /* No such device (e.g., media were just removed) */
543 if (!get_capacity(disk))
544 goto exit;
545
546 bdev = bdget_disk(disk, 0);
547 if (!bdev)
548 goto exit;
549
550 bdev->bd_invalidated = 1;
551 err = blkdev_get(bdev, FMODE_READ, NULL);
552 if (err < 0)
553 goto exit;
554 blkdev_put(bdev, FMODE_READ);
555
556exit:
557 /* announce disk after possible partitions are created */
558 dev_set_uevent_suppress(ddev, 0);
559 kobject_uevent(&ddev->kobj, KOBJ_ADD);
560
561 /* announce possible partitions */
562 disk_part_iter_init(&piter, disk, 0);
563 while ((part = disk_part_iter_next(&piter)))
564 kobject_uevent(&part_to_dev(part)->kobj, KOBJ_ADD);
565 disk_part_iter_exit(&piter);
566}
567
1da177e4
LT
568/**
569 * add_disk - add partitioning information to kernel list
570 * @disk: per-device partitioning information
571 *
572 * This function registers the partitioning information in @disk
573 * with the kernel.
3e1a7ff8
TH
574 *
575 * FIXME: error handling
1da177e4
LT
576 */
577void add_disk(struct gendisk *disk)
578{
cf0ca9fe 579 struct backing_dev_info *bdi;
3e1a7ff8 580 dev_t devt;
6ffeea77 581 int retval;
cf0ca9fe 582
3e1a7ff8
TH
583 /* minors == 0 indicates to use ext devt from part0 and should
584 * be accompanied with EXT_DEVT flag. Make sure all
585 * parameters make sense.
586 */
587 WARN_ON(disk->minors && !(disk->major || disk->first_minor));
588 WARN_ON(!disk->minors && !(disk->flags & GENHD_FL_EXT_DEVT));
589
1da177e4 590 disk->flags |= GENHD_FL_UP;
3e1a7ff8
TH
591
592 retval = blk_alloc_devt(&disk->part0, &devt);
593 if (retval) {
594 WARN_ON(1);
595 return;
596 }
597 disk_to_dev(disk)->devt = devt;
598
599 /* ->major and ->first_minor aren't supposed to be
600 * dereferenced from here on, but set them just in case.
601 */
602 disk->major = MAJOR(devt);
603 disk->first_minor = MINOR(devt);
604
9f5e4865 605 /* Register BDI before referencing it from bdev */
01ea5063
SJK
606 bdi = &disk->queue->backing_dev_info;
607 bdi_register_dev(bdi, disk_devt(disk));
608
f331c029
TH
609 blk_register_region(disk_devt(disk), disk->minors, NULL,
610 exact_match, exact_lock, disk);
1da177e4
LT
611 register_disk(disk);
612 blk_register_queue(disk);
cf0ca9fe 613
ed9e1982
TH
614 retval = sysfs_create_link(&disk_to_dev(disk)->kobj, &bdi->dev->kobj,
615 "bdi");
6ffeea77 616 WARN_ON(retval);
1da177e4 617
77ea887e 618 disk_add_events(disk);
1da177e4 619}
1da177e4 620EXPORT_SYMBOL(add_disk);
1da177e4 621
d2bf1b67 622void del_gendisk(struct gendisk *disk)
1da177e4 623{
d2bf1b67
TH
624 struct disk_part_iter piter;
625 struct hd_struct *part;
626
77ea887e
TH
627 disk_del_events(disk);
628
d2bf1b67
TH
629 /* invalidate stuff */
630 disk_part_iter_init(&piter, disk,
631 DISK_PITER_INCL_EMPTY | DISK_PITER_REVERSE);
632 while ((part = disk_part_iter_next(&piter))) {
633 invalidate_partition(disk, part->partno);
634 delete_partition(disk, part->partno);
635 }
636 disk_part_iter_exit(&piter);
637
638 invalidate_partition(disk, 0);
639 blk_free_devt(disk_to_dev(disk)->devt);
640 set_capacity(disk, 0);
641 disk->flags &= ~GENHD_FL_UP;
642
ed9e1982 643 sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
cf0ca9fe 644 bdi_unregister(&disk->queue->backing_dev_info);
1da177e4 645 blk_unregister_queue(disk);
f331c029 646 blk_unregister_region(disk_devt(disk), disk->minors);
d2bf1b67
TH
647
648 part_stat_set_all(&disk->part0, 0);
649 disk->part0.stamp = 0;
650
651 kobject_put(disk->part0.holder_dir);
652 kobject_put(disk->slave_dir);
653 disk->driverfs_dev = NULL;
654 if (!sysfs_deprecated)
655 sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
656 device_del(disk_to_dev(disk));
1da177e4 657}
d2bf1b67 658EXPORT_SYMBOL(del_gendisk);
1da177e4 659
1da177e4
LT
660/**
661 * get_gendisk - get partitioning information for a given device
710027a4 662 * @devt: device to get partitioning information for
496aa8a9 663 * @partno: returned partition index
1da177e4
LT
664 *
665 * This function gets the structure containing partitioning
710027a4 666 * information for the given device @devt.
1da177e4 667 */
cf771cb5 668struct gendisk *get_gendisk(dev_t devt, int *partno)
1da177e4 669{
bcce3de1
TH
670 struct gendisk *disk = NULL;
671
672 if (MAJOR(devt) != BLOCK_EXT_MAJOR) {
673 struct kobject *kobj;
674
675 kobj = kobj_lookup(bdev_map, devt, partno);
676 if (kobj)
677 disk = dev_to_disk(kobj_to_dev(kobj));
678 } else {
679 struct hd_struct *part;
680
681 mutex_lock(&ext_devt_mutex);
870d6656 682 part = idr_find(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
bcce3de1
TH
683 if (part && get_disk(part_to_disk(part))) {
684 *partno = part->partno;
685 disk = part_to_disk(part);
686 }
687 mutex_unlock(&ext_devt_mutex);
688 }
edfaa7c3 689
bcce3de1 690 return disk;
1da177e4 691}
b6ac23af 692EXPORT_SYMBOL(get_gendisk);
1da177e4 693
f331c029
TH
694/**
695 * bdget_disk - do bdget() by gendisk and partition number
696 * @disk: gendisk of interest
697 * @partno: partition number
698 *
699 * Find partition @partno from @disk, do bdget() on it.
700 *
701 * CONTEXT:
702 * Don't care.
703 *
704 * RETURNS:
705 * Resulting block_device on success, NULL on failure.
706 */
aeb3d3a8 707struct block_device *bdget_disk(struct gendisk *disk, int partno)
f331c029 708{
548b10eb
TH
709 struct hd_struct *part;
710 struct block_device *bdev = NULL;
f331c029 711
548b10eb 712 part = disk_get_part(disk, partno);
2bbedcb4 713 if (part)
548b10eb
TH
714 bdev = bdget(part_devt(part));
715 disk_put_part(part);
f331c029 716
548b10eb 717 return bdev;
f331c029
TH
718}
719EXPORT_SYMBOL(bdget_disk);
720
5c6f35c5
GKH
721/*
722 * print a full list of all partitions - intended for places where the root
723 * filesystem can't be mounted and thus to give the victim some idea of what
724 * went wrong
725 */
726void __init printk_all_partitions(void)
727{
def4e38d
TH
728 struct class_dev_iter iter;
729 struct device *dev;
730
731 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
732 while ((dev = class_dev_iter_next(&iter))) {
733 struct gendisk *disk = dev_to_disk(dev);
e71bf0d0
TH
734 struct disk_part_iter piter;
735 struct hd_struct *part;
1f014290
TH
736 char name_buf[BDEVNAME_SIZE];
737 char devt_buf[BDEVT_SIZE];
b5af921e 738 u8 uuid[PARTITION_META_INFO_UUIDLTH * 2 + 1];
def4e38d
TH
739
740 /*
741 * Don't show empty devices or things that have been
25985edc 742 * suppressed
def4e38d
TH
743 */
744 if (get_capacity(disk) == 0 ||
745 (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
746 continue;
747
748 /*
749 * Note, unlike /proc/partitions, I am showing the
750 * numbers in hex - the same format as the root=
751 * option takes.
752 */
074a7aca
TH
753 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0);
754 while ((part = disk_part_iter_next(&piter))) {
755 bool is_part0 = part == &disk->part0;
def4e38d 756
b5af921e
WD
757 uuid[0] = 0;
758 if (part->info)
759 part_unpack_uuid(part->info->uuid, uuid);
760
761 printk("%s%s %10llu %s %s", is_part0 ? "" : " ",
1f014290 762 bdevt_str(part_devt(part), devt_buf),
f331c029 763 (unsigned long long)part->nr_sects >> 1,
b5af921e 764 disk_name(disk, part->partno, name_buf), uuid);
074a7aca
TH
765 if (is_part0) {
766 if (disk->driverfs_dev != NULL &&
767 disk->driverfs_dev->driver != NULL)
768 printk(" driver: %s\n",
769 disk->driverfs_dev->driver->name);
770 else
771 printk(" (driver?)\n");
772 } else
773 printk("\n");
774 }
e71bf0d0 775 disk_part_iter_exit(&piter);
def4e38d
TH
776 }
777 class_dev_iter_exit(&iter);
dd2a345f
DG
778}
779
1da177e4
LT
780#ifdef CONFIG_PROC_FS
781/* iterator */
def4e38d 782static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
68c4d4a7 783{
def4e38d
TH
784 loff_t skip = *pos;
785 struct class_dev_iter *iter;
786 struct device *dev;
68c4d4a7 787
aeb3d3a8 788 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
def4e38d
TH
789 if (!iter)
790 return ERR_PTR(-ENOMEM);
791
792 seqf->private = iter;
793 class_dev_iter_init(iter, &block_class, NULL, &disk_type);
794 do {
795 dev = class_dev_iter_next(iter);
796 if (!dev)
797 return NULL;
798 } while (skip--);
799
800 return dev_to_disk(dev);
68c4d4a7
GKH
801}
802
def4e38d 803static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
1da177e4 804{
edfaa7c3 805 struct device *dev;
1da177e4 806
def4e38d
TH
807 (*pos)++;
808 dev = class_dev_iter_next(seqf->private);
2ac3cee5 809 if (dev)
68c4d4a7 810 return dev_to_disk(dev);
2ac3cee5 811
1da177e4
LT
812 return NULL;
813}
814
def4e38d 815static void disk_seqf_stop(struct seq_file *seqf, void *v)
27f30251 816{
def4e38d 817 struct class_dev_iter *iter = seqf->private;
27f30251 818
def4e38d
TH
819 /* stop is called even after start failed :-( */
820 if (iter) {
821 class_dev_iter_exit(iter);
822 kfree(iter);
5c0ef6d0 823 }
1da177e4
LT
824}
825
def4e38d 826static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
1da177e4 827{
def4e38d
TH
828 static void *p;
829
830 p = disk_seqf_start(seqf, pos);
b9f985b6 831 if (!IS_ERR_OR_NULL(p) && !*pos)
def4e38d
TH
832 seq_puts(seqf, "major minor #blocks name\n\n");
833 return p;
1da177e4
LT
834}
835
cf771cb5 836static int show_partition(struct seq_file *seqf, void *v)
1da177e4
LT
837{
838 struct gendisk *sgp = v;
e71bf0d0
TH
839 struct disk_part_iter piter;
840 struct hd_struct *part;
1da177e4
LT
841 char buf[BDEVNAME_SIZE];
842
1da177e4 843 /* Don't show non-partitionable removeable devices or empty devices */
b5d0b9df 844 if (!get_capacity(sgp) || (!disk_partitionable(sgp) &&
f331c029 845 (sgp->flags & GENHD_FL_REMOVABLE)))
1da177e4
LT
846 return 0;
847 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
848 return 0;
849
850 /* show the full disk and all non-0 size partitions of it */
074a7aca 851 disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0);
e71bf0d0 852 while ((part = disk_part_iter_next(&piter)))
1f014290 853 seq_printf(seqf, "%4d %7d %10llu %s\n",
f331c029
TH
854 MAJOR(part_devt(part)), MINOR(part_devt(part)),
855 (unsigned long long)part->nr_sects >> 1,
856 disk_name(sgp, part->partno, buf));
e71bf0d0 857 disk_part_iter_exit(&piter);
1da177e4
LT
858
859 return 0;
860}
861
f500975a 862static const struct seq_operations partitions_op = {
def4e38d
TH
863 .start = show_partition_start,
864 .next = disk_seqf_next,
865 .stop = disk_seqf_stop,
edfaa7c3 866 .show = show_partition
1da177e4 867};
f500975a
AD
868
869static int partitions_open(struct inode *inode, struct file *file)
870{
871 return seq_open(file, &partitions_op);
872}
873
874static const struct file_operations proc_partitions_operations = {
875 .open = partitions_open,
876 .read = seq_read,
877 .llseek = seq_lseek,
878 .release = seq_release,
879};
1da177e4
LT
880#endif
881
882
cf771cb5 883static struct kobject *base_probe(dev_t devt, int *partno, void *data)
1da177e4 884{
edfaa7c3 885 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
1da177e4 886 /* Make old-style 2.4 aliases work */
edfaa7c3 887 request_module("block-major-%d", MAJOR(devt));
1da177e4
LT
888 return NULL;
889}
890
891static int __init genhd_device_init(void)
892{
e105b8bf
DW
893 int error;
894
895 block_class.dev_kobj = sysfs_dev_block_kobj;
896 error = class_register(&block_class);
ee27a558
RM
897 if (unlikely(error))
898 return error;
edfaa7c3 899 bdev_map = kobj_map_init(base_probe, &block_class_lock);
1da177e4 900 blk_dev_init();
edfaa7c3 901
561ec68e
ZY
902 register_blkdev(BLOCK_EXT_MAJOR, "blkext");
903
edfaa7c3 904 /* create top-level block dir */
e52eec13
AK
905 if (!sysfs_deprecated)
906 block_depr = kobject_create_and_add("block", NULL);
830d3cfb 907 return 0;
1da177e4
LT
908}
909
910subsys_initcall(genhd_device_init);
911
edfaa7c3
KS
912static ssize_t disk_range_show(struct device *dev,
913 struct device_attribute *attr, char *buf)
1da177e4 914{
edfaa7c3 915 struct gendisk *disk = dev_to_disk(dev);
1da177e4 916
edfaa7c3 917 return sprintf(buf, "%d\n", disk->minors);
1da177e4
LT
918}
919
1f014290
TH
920static ssize_t disk_ext_range_show(struct device *dev,
921 struct device_attribute *attr, char *buf)
922{
923 struct gendisk *disk = dev_to_disk(dev);
924
b5d0b9df 925 return sprintf(buf, "%d\n", disk_max_parts(disk));
1f014290
TH
926}
927
edfaa7c3
KS
928static ssize_t disk_removable_show(struct device *dev,
929 struct device_attribute *attr, char *buf)
a7fd6706 930{
edfaa7c3 931 struct gendisk *disk = dev_to_disk(dev);
a7fd6706 932
edfaa7c3
KS
933 return sprintf(buf, "%d\n",
934 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
a7fd6706
KS
935}
936
1c9ce527
KS
937static ssize_t disk_ro_show(struct device *dev,
938 struct device_attribute *attr, char *buf)
939{
940 struct gendisk *disk = dev_to_disk(dev);
941
b7db9956 942 return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
1c9ce527
KS
943}
944
edfaa7c3
KS
945static ssize_t disk_capability_show(struct device *dev,
946 struct device_attribute *attr, char *buf)
86ce18d7 947{
edfaa7c3
KS
948 struct gendisk *disk = dev_to_disk(dev);
949
950 return sprintf(buf, "%x\n", disk->flags);
86ce18d7 951}
edfaa7c3 952
c72758f3
MP
953static ssize_t disk_alignment_offset_show(struct device *dev,
954 struct device_attribute *attr,
955 char *buf)
956{
957 struct gendisk *disk = dev_to_disk(dev);
958
959 return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue));
960}
961
86b37281
MP
962static ssize_t disk_discard_alignment_show(struct device *dev,
963 struct device_attribute *attr,
964 char *buf)
965{
966 struct gendisk *disk = dev_to_disk(dev);
967
dd3d145d 968 return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue));
86b37281
MP
969}
970
edfaa7c3 971static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL);
1f014290 972static DEVICE_ATTR(ext_range, S_IRUGO, disk_ext_range_show, NULL);
edfaa7c3 973static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL);
1c9ce527 974static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL);
e5610521 975static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL);
c72758f3 976static DEVICE_ATTR(alignment_offset, S_IRUGO, disk_alignment_offset_show, NULL);
86b37281
MP
977static DEVICE_ATTR(discard_alignment, S_IRUGO, disk_discard_alignment_show,
978 NULL);
edfaa7c3 979static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL);
074a7aca 980static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL);
316d315b 981static DEVICE_ATTR(inflight, S_IRUGO, part_inflight_show, NULL);
c17bb495 982#ifdef CONFIG_FAIL_MAKE_REQUEST
edfaa7c3 983static struct device_attribute dev_attr_fail =
eddb2e26 984 __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store);
c17bb495 985#endif
581d4e28
JA
986#ifdef CONFIG_FAIL_IO_TIMEOUT
987static struct device_attribute dev_attr_fail_timeout =
988 __ATTR(io-timeout-fail, S_IRUGO|S_IWUSR, part_timeout_show,
989 part_timeout_store);
990#endif
edfaa7c3
KS
991
992static struct attribute *disk_attrs[] = {
993 &dev_attr_range.attr,
1f014290 994 &dev_attr_ext_range.attr,
edfaa7c3 995 &dev_attr_removable.attr,
1c9ce527 996 &dev_attr_ro.attr,
edfaa7c3 997 &dev_attr_size.attr,
c72758f3 998 &dev_attr_alignment_offset.attr,
86b37281 999 &dev_attr_discard_alignment.attr,
edfaa7c3
KS
1000 &dev_attr_capability.attr,
1001 &dev_attr_stat.attr,
316d315b 1002 &dev_attr_inflight.attr,
edfaa7c3
KS
1003#ifdef CONFIG_FAIL_MAKE_REQUEST
1004 &dev_attr_fail.attr,
581d4e28
JA
1005#endif
1006#ifdef CONFIG_FAIL_IO_TIMEOUT
1007 &dev_attr_fail_timeout.attr,
edfaa7c3
KS
1008#endif
1009 NULL
1010};
1011
1012static struct attribute_group disk_attr_group = {
1013 .attrs = disk_attrs,
1014};
1015
a4dbd674 1016static const struct attribute_group *disk_attr_groups[] = {
edfaa7c3
KS
1017 &disk_attr_group,
1018 NULL
1da177e4
LT
1019};
1020
540eed56
TH
1021/**
1022 * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way
1023 * @disk: disk to replace part_tbl for
1024 * @new_ptbl: new part_tbl to install
1025 *
1026 * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The
1027 * original ptbl is freed using RCU callback.
1028 *
1029 * LOCKING:
1030 * Matching bd_mutx locked.
1031 */
1032static void disk_replace_part_tbl(struct gendisk *disk,
1033 struct disk_part_tbl *new_ptbl)
1034{
1035 struct disk_part_tbl *old_ptbl = disk->part_tbl;
1036
1037 rcu_assign_pointer(disk->part_tbl, new_ptbl);
a6f23657
JA
1038
1039 if (old_ptbl) {
1040 rcu_assign_pointer(old_ptbl->last_lookup, NULL);
57bdfbf9 1041 kfree_rcu(old_ptbl, rcu_head);
a6f23657 1042 }
540eed56
TH
1043}
1044
1045/**
1046 * disk_expand_part_tbl - expand disk->part_tbl
1047 * @disk: disk to expand part_tbl for
1048 * @partno: expand such that this partno can fit in
1049 *
1050 * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl
1051 * uses RCU to allow unlocked dereferencing for stats and other stuff.
1052 *
1053 * LOCKING:
1054 * Matching bd_mutex locked, might sleep.
1055 *
1056 * RETURNS:
1057 * 0 on success, -errno on failure.
1058 */
1059int disk_expand_part_tbl(struct gendisk *disk, int partno)
1060{
1061 struct disk_part_tbl *old_ptbl = disk->part_tbl;
1062 struct disk_part_tbl *new_ptbl;
1063 int len = old_ptbl ? old_ptbl->len : 0;
1064 int target = partno + 1;
1065 size_t size;
1066 int i;
1067
1068 /* disk_max_parts() is zero during initialization, ignore if so */
1069 if (disk_max_parts(disk) && target > disk_max_parts(disk))
1070 return -EINVAL;
1071
1072 if (target <= len)
1073 return 0;
1074
1075 size = sizeof(*new_ptbl) + target * sizeof(new_ptbl->part[0]);
1076 new_ptbl = kzalloc_node(size, GFP_KERNEL, disk->node_id);
1077 if (!new_ptbl)
1078 return -ENOMEM;
1079
540eed56
TH
1080 new_ptbl->len = target;
1081
1082 for (i = 0; i < len; i++)
1083 rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]);
1084
1085 disk_replace_part_tbl(disk, new_ptbl);
1086 return 0;
1087}
1088
edfaa7c3 1089static void disk_release(struct device *dev)
1da177e4 1090{
edfaa7c3
KS
1091 struct gendisk *disk = dev_to_disk(dev);
1092
77ea887e 1093 disk_release_events(disk);
1da177e4 1094 kfree(disk->random);
540eed56 1095 disk_replace_part_tbl(disk, NULL);
074a7aca 1096 free_part_stats(&disk->part0);
6d1d8050 1097 free_part_info(&disk->part0);
1da177e4
LT
1098 kfree(disk);
1099}
edfaa7c3
KS
1100struct class block_class = {
1101 .name = "block",
1da177e4
LT
1102};
1103
e454cea2 1104static char *block_devnode(struct device *dev, mode_t *mode)
b03f38b6
KS
1105{
1106 struct gendisk *disk = dev_to_disk(dev);
1107
e454cea2
KS
1108 if (disk->devnode)
1109 return disk->devnode(disk, mode);
b03f38b6
KS
1110 return NULL;
1111}
1112
1826eadf 1113static struct device_type disk_type = {
edfaa7c3
KS
1114 .name = "disk",
1115 .groups = disk_attr_groups,
1116 .release = disk_release,
e454cea2 1117 .devnode = block_devnode,
1da177e4
LT
1118};
1119
a6e2ba88 1120#ifdef CONFIG_PROC_FS
cf771cb5
TH
1121/*
1122 * aggregate disk stat collector. Uses the same stats that the sysfs
1123 * entries do, above, but makes them available through one seq_file.
1124 *
1125 * The output looks suspiciously like /proc/partitions with a bunch of
1126 * extra fields.
1127 */
1128static int diskstats_show(struct seq_file *seqf, void *v)
1da177e4
LT
1129{
1130 struct gendisk *gp = v;
e71bf0d0
TH
1131 struct disk_part_iter piter;
1132 struct hd_struct *hd;
1da177e4 1133 char buf[BDEVNAME_SIZE];
c9959059 1134 int cpu;
1da177e4
LT
1135
1136 /*
ed9e1982 1137 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
cf771cb5 1138 seq_puts(seqf, "major minor name"
1da177e4
LT
1139 " rio rmerge rsect ruse wio wmerge "
1140 "wsect wuse running use aveq"
1141 "\n\n");
1142 */
9f5e4865 1143
71982a40 1144 disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0);
e71bf0d0 1145 while ((hd = disk_part_iter_next(&piter))) {
074a7aca 1146 cpu = part_stat_lock();
c9959059 1147 part_round_stats(cpu, hd);
074a7aca 1148 part_stat_unlock();
f95fe9cf
HP
1149 seq_printf(seqf, "%4d %7d %s %lu %lu %lu "
1150 "%u %lu %lu %lu %u %u %u %u\n",
f331c029
TH
1151 MAJOR(part_devt(hd)), MINOR(part_devt(hd)),
1152 disk_name(gp, hd->partno, buf),
53f22956
LY
1153 part_stat_read(hd, ios[READ]),
1154 part_stat_read(hd, merges[READ]),
f95fe9cf 1155 part_stat_read(hd, sectors[READ]),
53f22956
LY
1156 jiffies_to_msecs(part_stat_read(hd, ticks[READ])),
1157 part_stat_read(hd, ios[WRITE]),
1158 part_stat_read(hd, merges[WRITE]),
f95fe9cf 1159 part_stat_read(hd, sectors[WRITE]),
53f22956 1160 jiffies_to_msecs(part_stat_read(hd, ticks[WRITE])),
316d315b 1161 part_in_flight(hd),
28f39d55
JM
1162 jiffies_to_msecs(part_stat_read(hd, io_ticks)),
1163 jiffies_to_msecs(part_stat_read(hd, time_in_queue))
1164 );
1da177e4 1165 }
e71bf0d0 1166 disk_part_iter_exit(&piter);
9f5e4865 1167
1da177e4
LT
1168 return 0;
1169}
1170
31d85ab2 1171static const struct seq_operations diskstats_op = {
def4e38d
TH
1172 .start = disk_seqf_start,
1173 .next = disk_seqf_next,
1174 .stop = disk_seqf_stop,
1da177e4
LT
1175 .show = diskstats_show
1176};
f500975a 1177
31d85ab2
AD
1178static int diskstats_open(struct inode *inode, struct file *file)
1179{
1180 return seq_open(file, &diskstats_op);
1181}
1182
1183static const struct file_operations proc_diskstats_operations = {
1184 .open = diskstats_open,
1185 .read = seq_read,
1186 .llseek = seq_lseek,
1187 .release = seq_release,
1188};
1189
f500975a
AD
1190static int __init proc_genhd_init(void)
1191{
31d85ab2 1192 proc_create("diskstats", 0, NULL, &proc_diskstats_operations);
f500975a
AD
1193 proc_create("partitions", 0, NULL, &proc_partitions_operations);
1194 return 0;
1195}
1196module_init(proc_genhd_init);
a6e2ba88 1197#endif /* CONFIG_PROC_FS */
1da177e4 1198
cf771cb5 1199dev_t blk_lookup_devt(const char *name, int partno)
a142be85 1200{
def4e38d
TH
1201 dev_t devt = MKDEV(0, 0);
1202 struct class_dev_iter iter;
1203 struct device *dev;
a142be85 1204
def4e38d
TH
1205 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1206 while ((dev = class_dev_iter_next(&iter))) {
a142be85 1207 struct gendisk *disk = dev_to_disk(dev);
548b10eb 1208 struct hd_struct *part;
a142be85 1209
3ada8b7e 1210 if (strcmp(dev_name(dev), name))
f331c029 1211 continue;
f331c029 1212
41b8c853
NB
1213 if (partno < disk->minors) {
1214 /* We need to return the right devno, even
1215 * if the partition doesn't exist yet.
1216 */
1217 devt = MKDEV(MAJOR(dev->devt),
1218 MINOR(dev->devt) + partno);
1219 break;
1220 }
548b10eb 1221 part = disk_get_part(disk, partno);
2bbedcb4 1222 if (part) {
f331c029 1223 devt = part_devt(part);
e71bf0d0 1224 disk_put_part(part);
548b10eb 1225 break;
def4e38d 1226 }
548b10eb 1227 disk_put_part(part);
5c0ef6d0 1228 }
def4e38d 1229 class_dev_iter_exit(&iter);
edfaa7c3
KS
1230 return devt;
1231}
edfaa7c3
KS
1232EXPORT_SYMBOL(blk_lookup_devt);
1233
1da177e4
LT
1234struct gendisk *alloc_disk(int minors)
1235{
1946089a
CL
1236 return alloc_disk_node(minors, -1);
1237}
689d6fac 1238EXPORT_SYMBOL(alloc_disk);
1946089a
CL
1239
1240struct gendisk *alloc_disk_node(int minors, int node_id)
1241{
1242 struct gendisk *disk;
1243
94f6030c
CL
1244 disk = kmalloc_node(sizeof(struct gendisk),
1245 GFP_KERNEL | __GFP_ZERO, node_id);
1da177e4 1246 if (disk) {
074a7aca 1247 if (!init_part_stats(&disk->part0)) {
1da177e4
LT
1248 kfree(disk);
1249 return NULL;
1250 }
bf91db18 1251 disk->node_id = node_id;
540eed56
TH
1252 if (disk_expand_part_tbl(disk, 0)) {
1253 free_part_stats(&disk->part0);
b5d0b9df
TH
1254 kfree(disk);
1255 return NULL;
1da177e4 1256 }
540eed56 1257 disk->part_tbl->part[0] = &disk->part0;
6c23a968
JA
1258
1259 hd_ref_init(&disk->part0);
b5d0b9df 1260
1da177e4 1261 disk->minors = minors;
1da177e4 1262 rand_initialize_disk(disk);
ed9e1982
TH
1263 disk_to_dev(disk)->class = &block_class;
1264 disk_to_dev(disk)->type = &disk_type;
1265 device_initialize(disk_to_dev(disk));
1da177e4
LT
1266 }
1267 return disk;
1268}
1946089a 1269EXPORT_SYMBOL(alloc_disk_node);
1da177e4
LT
1270
1271struct kobject *get_disk(struct gendisk *disk)
1272{
1273 struct module *owner;
1274 struct kobject *kobj;
1275
1276 if (!disk->fops)
1277 return NULL;
1278 owner = disk->fops->owner;
1279 if (owner && !try_module_get(owner))
1280 return NULL;
ed9e1982 1281 kobj = kobject_get(&disk_to_dev(disk)->kobj);
1da177e4
LT
1282 if (kobj == NULL) {
1283 module_put(owner);
1284 return NULL;
1285 }
1286 return kobj;
1287
1288}
1289
1290EXPORT_SYMBOL(get_disk);
1291
1292void put_disk(struct gendisk *disk)
1293{
1294 if (disk)
ed9e1982 1295 kobject_put(&disk_to_dev(disk)->kobj);
1da177e4
LT
1296}
1297
1298EXPORT_SYMBOL(put_disk);
1299
e3264a4d
HR
1300static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1301{
1302 char event[] = "DISK_RO=1";
1303 char *envp[] = { event, NULL };
1304
1305 if (!ro)
1306 event[8] = '0';
1307 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1308}
1309
1da177e4
LT
1310void set_device_ro(struct block_device *bdev, int flag)
1311{
b7db9956 1312 bdev->bd_part->policy = flag;
1da177e4
LT
1313}
1314
1315EXPORT_SYMBOL(set_device_ro);
1316
1317void set_disk_ro(struct gendisk *disk, int flag)
1318{
e71bf0d0
TH
1319 struct disk_part_iter piter;
1320 struct hd_struct *part;
1321
e3264a4d
HR
1322 if (disk->part0.policy != flag) {
1323 set_disk_ro_uevent(disk, flag);
1324 disk->part0.policy = flag;
1325 }
1326
1327 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY);
e71bf0d0
TH
1328 while ((part = disk_part_iter_next(&piter)))
1329 part->policy = flag;
1330 disk_part_iter_exit(&piter);
1da177e4
LT
1331}
1332
1333EXPORT_SYMBOL(set_disk_ro);
1334
1335int bdev_read_only(struct block_device *bdev)
1336{
1337 if (!bdev)
1338 return 0;
b7db9956 1339 return bdev->bd_part->policy;
1da177e4
LT
1340}
1341
1342EXPORT_SYMBOL(bdev_read_only);
1343
cf771cb5 1344int invalidate_partition(struct gendisk *disk, int partno)
1da177e4
LT
1345{
1346 int res = 0;
cf771cb5 1347 struct block_device *bdev = bdget_disk(disk, partno);
1da177e4 1348 if (bdev) {
2ef41634 1349 fsync_bdev(bdev);
93b270f7 1350 res = __invalidate_device(bdev, true);
1da177e4
LT
1351 bdput(bdev);
1352 }
1353 return res;
1354}
1355
1356EXPORT_SYMBOL(invalidate_partition);
77ea887e
TH
1357
1358/*
1359 * Disk events - monitor disk events like media change and eject request.
1360 */
1361struct disk_events {
1362 struct list_head node; /* all disk_event's */
1363 struct gendisk *disk; /* the associated disk */
1364 spinlock_t lock;
1365
fdd514e1 1366 struct mutex block_mutex; /* protects blocking */
77ea887e
TH
1367 int block; /* event blocking depth */
1368 unsigned int pending; /* events already sent out */
1369 unsigned int clearing; /* events being cleared */
1370
1371 long poll_msecs; /* interval, -1 for default */
1372 struct delayed_work dwork;
1373};
1374
1375static const char *disk_events_strs[] = {
1376 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "media_change",
1377 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "eject_request",
1378};
1379
1380static char *disk_uevents[] = {
1381 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "DISK_MEDIA_CHANGE=1",
1382 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "DISK_EJECT_REQUEST=1",
1383};
1384
1385/* list of all disk_events */
1386static DEFINE_MUTEX(disk_events_mutex);
1387static LIST_HEAD(disk_events);
1388
1389/* disable in-kernel polling by default */
1390static unsigned long disk_events_dfl_poll_msecs = 0;
1391
1392static unsigned long disk_events_poll_jiffies(struct gendisk *disk)
1393{
1394 struct disk_events *ev = disk->ev;
1395 long intv_msecs = 0;
1396
1397 /*
1398 * If device-specific poll interval is set, always use it. If
1399 * the default is being used, poll iff there are events which
1400 * can't be monitored asynchronously.
1401 */
1402 if (ev->poll_msecs >= 0)
1403 intv_msecs = ev->poll_msecs;
1404 else if (disk->events & ~disk->async_events)
1405 intv_msecs = disk_events_dfl_poll_msecs;
1406
1407 return msecs_to_jiffies(intv_msecs);
1408}
1409
c3af54af
TH
1410/**
1411 * disk_block_events - block and flush disk event checking
1412 * @disk: disk to block events for
1413 *
1414 * On return from this function, it is guaranteed that event checking
1415 * isn't in progress and won't happen until unblocked by
1416 * disk_unblock_events(). Events blocking is counted and the actual
1417 * unblocking happens after the matching number of unblocks are done.
1418 *
1419 * Note that this intentionally does not block event checking from
1420 * disk_clear_events().
1421 *
1422 * CONTEXT:
1423 * Might sleep.
1424 */
1425void disk_block_events(struct gendisk *disk)
77ea887e
TH
1426{
1427 struct disk_events *ev = disk->ev;
1428 unsigned long flags;
1429 bool cancel;
1430
c3af54af
TH
1431 if (!ev)
1432 return;
1433
fdd514e1
TH
1434 /*
1435 * Outer mutex ensures that the first blocker completes canceling
1436 * the event work before further blockers are allowed to finish.
1437 */
1438 mutex_lock(&ev->block_mutex);
1439
77ea887e
TH
1440 spin_lock_irqsave(&ev->lock, flags);
1441 cancel = !ev->block++;
1442 spin_unlock_irqrestore(&ev->lock, flags);
1443
c3af54af
TH
1444 if (cancel)
1445 cancel_delayed_work_sync(&disk->ev->dwork);
fdd514e1
TH
1446
1447 mutex_unlock(&ev->block_mutex);
77ea887e
TH
1448}
1449
1450static void __disk_unblock_events(struct gendisk *disk, bool check_now)
1451{
1452 struct disk_events *ev = disk->ev;
1453 unsigned long intv;
1454 unsigned long flags;
1455
1456 spin_lock_irqsave(&ev->lock, flags);
1457
1458 if (WARN_ON_ONCE(ev->block <= 0))
1459 goto out_unlock;
1460
1461 if (--ev->block)
1462 goto out_unlock;
1463
1464 /*
1465 * Not exactly a latency critical operation, set poll timer
1466 * slack to 25% and kick event check.
1467 */
1468 intv = disk_events_poll_jiffies(disk);
1469 set_timer_slack(&ev->dwork.timer, intv / 4);
1470 if (check_now)
1471 queue_delayed_work(system_nrt_wq, &ev->dwork, 0);
1472 else if (intv)
1473 queue_delayed_work(system_nrt_wq, &ev->dwork, intv);
1474out_unlock:
1475 spin_unlock_irqrestore(&ev->lock, flags);
1476}
1477
77ea887e
TH
1478/**
1479 * disk_unblock_events - unblock disk event checking
1480 * @disk: disk to unblock events for
1481 *
1482 * Undo disk_block_events(). When the block count reaches zero, it
1483 * starts events polling if configured.
1484 *
1485 * CONTEXT:
1486 * Don't care. Safe to call from irq context.
1487 */
1488void disk_unblock_events(struct gendisk *disk)
1489{
1490 if (disk->ev)
facc31dd 1491 __disk_unblock_events(disk, false);
77ea887e
TH
1492}
1493
1494/**
85ef06d1
TH
1495 * disk_flush_events - schedule immediate event checking and flushing
1496 * @disk: disk to check and flush events for
1497 * @mask: events to flush
77ea887e 1498 *
85ef06d1
TH
1499 * Schedule immediate event checking on @disk if not blocked. Events in
1500 * @mask are scheduled to be cleared from the driver. Note that this
1501 * doesn't clear the events from @disk->ev.
77ea887e
TH
1502 *
1503 * CONTEXT:
85ef06d1 1504 * If @mask is non-zero must be called with bdev->bd_mutex held.
77ea887e 1505 */
85ef06d1 1506void disk_flush_events(struct gendisk *disk, unsigned int mask)
77ea887e 1507{
a9dce2a3 1508 struct disk_events *ev = disk->ev;
a9dce2a3
TH
1509
1510 if (!ev)
1511 return;
1512
85ef06d1
TH
1513 spin_lock_irq(&ev->lock);
1514 ev->clearing |= mask;
a9dce2a3
TH
1515 if (!ev->block) {
1516 cancel_delayed_work(&ev->dwork);
1517 queue_delayed_work(system_nrt_wq, &ev->dwork, 0);
77ea887e 1518 }
85ef06d1 1519 spin_unlock_irq(&ev->lock);
77ea887e 1520}
77ea887e
TH
1521
1522/**
1523 * disk_clear_events - synchronously check, clear and return pending events
1524 * @disk: disk to fetch and clear events from
1525 * @mask: mask of events to be fetched and clearted
1526 *
1527 * Disk events are synchronously checked and pending events in @mask
1528 * are cleared and returned. This ignores the block count.
1529 *
1530 * CONTEXT:
1531 * Might sleep.
1532 */
1533unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask)
1534{
1535 const struct block_device_operations *bdops = disk->fops;
1536 struct disk_events *ev = disk->ev;
1537 unsigned int pending;
1538
1539 if (!ev) {
1540 /* for drivers still using the old ->media_changed method */
1541 if ((mask & DISK_EVENT_MEDIA_CHANGE) &&
1542 bdops->media_changed && bdops->media_changed(disk))
1543 return DISK_EVENT_MEDIA_CHANGE;
1544 return 0;
1545 }
1546
1547 /* tell the workfn about the events being cleared */
1548 spin_lock_irq(&ev->lock);
1549 ev->clearing |= mask;
1550 spin_unlock_irq(&ev->lock);
1551
1552 /* uncondtionally schedule event check and wait for it to finish */
c3af54af 1553 disk_block_events(disk);
77ea887e
TH
1554 queue_delayed_work(system_nrt_wq, &ev->dwork, 0);
1555 flush_delayed_work(&ev->dwork);
1556 __disk_unblock_events(disk, false);
1557
1558 /* then, fetch and clear pending events */
1559 spin_lock_irq(&ev->lock);
1560 WARN_ON_ONCE(ev->clearing & mask); /* cleared by workfn */
1561 pending = ev->pending & mask;
1562 ev->pending &= ~mask;
1563 spin_unlock_irq(&ev->lock);
1564
1565 return pending;
1566}
1567
1568static void disk_events_workfn(struct work_struct *work)
1569{
1570 struct delayed_work *dwork = to_delayed_work(work);
1571 struct disk_events *ev = container_of(dwork, struct disk_events, dwork);
1572 struct gendisk *disk = ev->disk;
1573 char *envp[ARRAY_SIZE(disk_uevents) + 1] = { };
1574 unsigned int clearing = ev->clearing;
1575 unsigned int events;
1576 unsigned long intv;
1577 int nr_events = 0, i;
1578
1579 /* check events */
1580 events = disk->fops->check_events(disk, clearing);
1581
1582 /* accumulate pending events and schedule next poll if necessary */
1583 spin_lock_irq(&ev->lock);
1584
1585 events &= ~ev->pending;
1586 ev->pending |= events;
1587 ev->clearing &= ~clearing;
1588
1589 intv = disk_events_poll_jiffies(disk);
1590 if (!ev->block && intv)
1591 queue_delayed_work(system_nrt_wq, &ev->dwork, intv);
1592
1593 spin_unlock_irq(&ev->lock);
1594
7c88a168
TH
1595 /*
1596 * Tell userland about new events. Only the events listed in
1597 * @disk->events are reported. Unlisted events are processed the
1598 * same internally but never get reported to userland.
1599 */
77ea887e 1600 for (i = 0; i < ARRAY_SIZE(disk_uevents); i++)
7c88a168 1601 if (events & disk->events & (1 << i))
77ea887e
TH
1602 envp[nr_events++] = disk_uevents[i];
1603
1604 if (nr_events)
1605 kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp);
1606}
1607
1608/*
1609 * A disk events enabled device has the following sysfs nodes under
1610 * its /sys/block/X/ directory.
1611 *
1612 * events : list of all supported events
1613 * events_async : list of events which can be detected w/o polling
1614 * events_poll_msecs : polling interval, 0: disable, -1: system default
1615 */
1616static ssize_t __disk_events_show(unsigned int events, char *buf)
1617{
1618 const char *delim = "";
1619 ssize_t pos = 0;
1620 int i;
1621
1622 for (i = 0; i < ARRAY_SIZE(disk_events_strs); i++)
1623 if (events & (1 << i)) {
1624 pos += sprintf(buf + pos, "%s%s",
1625 delim, disk_events_strs[i]);
1626 delim = " ";
1627 }
1628 if (pos)
1629 pos += sprintf(buf + pos, "\n");
1630 return pos;
1631}
1632
1633static ssize_t disk_events_show(struct device *dev,
1634 struct device_attribute *attr, char *buf)
1635{
1636 struct gendisk *disk = dev_to_disk(dev);
1637
1638 return __disk_events_show(disk->events, buf);
1639}
1640
1641static ssize_t disk_events_async_show(struct device *dev,
1642 struct device_attribute *attr, char *buf)
1643{
1644 struct gendisk *disk = dev_to_disk(dev);
1645
1646 return __disk_events_show(disk->async_events, buf);
1647}
1648
1649static ssize_t disk_events_poll_msecs_show(struct device *dev,
1650 struct device_attribute *attr,
1651 char *buf)
1652{
1653 struct gendisk *disk = dev_to_disk(dev);
1654
1655 return sprintf(buf, "%ld\n", disk->ev->poll_msecs);
1656}
1657
1658static ssize_t disk_events_poll_msecs_store(struct device *dev,
1659 struct device_attribute *attr,
1660 const char *buf, size_t count)
1661{
1662 struct gendisk *disk = dev_to_disk(dev);
1663 long intv;
1664
1665 if (!count || !sscanf(buf, "%ld", &intv))
1666 return -EINVAL;
1667
1668 if (intv < 0 && intv != -1)
1669 return -EINVAL;
1670
c3af54af 1671 disk_block_events(disk);
77ea887e
TH
1672 disk->ev->poll_msecs = intv;
1673 __disk_unblock_events(disk, true);
1674
1675 return count;
1676}
1677
1678static const DEVICE_ATTR(events, S_IRUGO, disk_events_show, NULL);
1679static const DEVICE_ATTR(events_async, S_IRUGO, disk_events_async_show, NULL);
1680static const DEVICE_ATTR(events_poll_msecs, S_IRUGO|S_IWUSR,
1681 disk_events_poll_msecs_show,
1682 disk_events_poll_msecs_store);
1683
1684static const struct attribute *disk_events_attrs[] = {
1685 &dev_attr_events.attr,
1686 &dev_attr_events_async.attr,
1687 &dev_attr_events_poll_msecs.attr,
1688 NULL,
1689};
1690
1691/*
1692 * The default polling interval can be specified by the kernel
1693 * parameter block.events_dfl_poll_msecs which defaults to 0
1694 * (disable). This can also be modified runtime by writing to
1695 * /sys/module/block/events_dfl_poll_msecs.
1696 */
1697static int disk_events_set_dfl_poll_msecs(const char *val,
1698 const struct kernel_param *kp)
1699{
1700 struct disk_events *ev;
1701 int ret;
1702
1703 ret = param_set_ulong(val, kp);
1704 if (ret < 0)
1705 return ret;
1706
1707 mutex_lock(&disk_events_mutex);
1708
1709 list_for_each_entry(ev, &disk_events, node)
85ef06d1 1710 disk_flush_events(ev->disk, 0);
77ea887e
TH
1711
1712 mutex_unlock(&disk_events_mutex);
1713
1714 return 0;
1715}
1716
1717static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops = {
1718 .set = disk_events_set_dfl_poll_msecs,
1719 .get = param_get_ulong,
1720};
1721
1722#undef MODULE_PARAM_PREFIX
1723#define MODULE_PARAM_PREFIX "block."
1724
1725module_param_cb(events_dfl_poll_msecs, &disk_events_dfl_poll_msecs_param_ops,
1726 &disk_events_dfl_poll_msecs, 0644);
1727
1728/*
1729 * disk_{add|del|release}_events - initialize and destroy disk_events.
1730 */
1731static void disk_add_events(struct gendisk *disk)
1732{
1733 struct disk_events *ev;
1734
75e3f3ee 1735 if (!disk->fops->check_events)
77ea887e
TH
1736 return;
1737
1738 ev = kzalloc(sizeof(*ev), GFP_KERNEL);
1739 if (!ev) {
1740 pr_warn("%s: failed to initialize events\n", disk->disk_name);
1741 return;
1742 }
1743
1744 if (sysfs_create_files(&disk_to_dev(disk)->kobj,
1745 disk_events_attrs) < 0) {
1746 pr_warn("%s: failed to create sysfs files for events\n",
1747 disk->disk_name);
1748 kfree(ev);
1749 return;
1750 }
1751
1752 disk->ev = ev;
1753
1754 INIT_LIST_HEAD(&ev->node);
1755 ev->disk = disk;
1756 spin_lock_init(&ev->lock);
fdd514e1 1757 mutex_init(&ev->block_mutex);
77ea887e
TH
1758 ev->block = 1;
1759 ev->poll_msecs = -1;
1760 INIT_DELAYED_WORK(&ev->dwork, disk_events_workfn);
1761
1762 mutex_lock(&disk_events_mutex);
1763 list_add_tail(&ev->node, &disk_events);
1764 mutex_unlock(&disk_events_mutex);
1765
1766 /*
1767 * Block count is initialized to 1 and the following initial
1768 * unblock kicks it into action.
1769 */
1770 __disk_unblock_events(disk, true);
1771}
1772
1773static void disk_del_events(struct gendisk *disk)
1774{
1775 if (!disk->ev)
1776 return;
1777
c3af54af 1778 disk_block_events(disk);
77ea887e
TH
1779
1780 mutex_lock(&disk_events_mutex);
1781 list_del_init(&disk->ev->node);
1782 mutex_unlock(&disk_events_mutex);
1783
1784 sysfs_remove_files(&disk_to_dev(disk)->kobj, disk_events_attrs);
1785}
1786
1787static void disk_release_events(struct gendisk *disk)
1788{
1789 /* the block count should be 1 from disk_del_events() */
1790 WARN_ON_ONCE(disk->ev && disk->ev->block != 1);
1791 kfree(disk->ev);
1792}