blob: dadf42b454a383299231fa46abee1631aab53196 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * gendisk handling
3 */
4
Linus Torvalds1da177e2005-04-16 15:20:36 -07005#include <linux/module.h>
6#include <linux/fs.h>
7#include <linux/genhd.h>
Andrew Mortonb446b602007-02-20 13:57:48 -08008#include <linux/kdev_t.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -07009#include <linux/kernel.h>
10#include <linux/blkdev.h>
11#include <linux/init.h>
12#include <linux/spinlock.h>
Alexey Dobriyanf5009752008-10-04 23:53:21 +040013#include <linux/proc_fs.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070014#include <linux/seq_file.h>
15#include <linux/slab.h>
16#include <linux/kmod.h>
17#include <linux/kobj_map.h>
Jes Sorensen58383af2006-02-06 14:12:43 -080018#include <linux/mutex.h>
Tejun Heobcce3de2008-08-25 19:47:22 +090019#include <linux/idr.h>
Tejun Heo77ea8872010-12-08 20:57:37 +010020#include <linux/log2.h>
Ming Lei25e823c2013-02-22 16:34:13 -080021#include <linux/pm_runtime.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070022
Adrian Bunkff889722008-03-04 11:23:45 +010023#include "blk.h"
24
Kay Sieversedfaa7c2007-05-21 22:08:01 +020025static DEFINE_MUTEX(block_class_lock);
Kay Sieversedfaa7c2007-05-21 22:08:01 +020026struct kobject *block_depr;
Linus Torvalds1da177e2005-04-16 15:20:36 -070027
Tejun Heobcce3de2008-08-25 19:47:22 +090028/* for extended dynamic devt allocation, currently only one major is used */
Tejun Heoce23bba2013-02-27 17:03:56 -080029#define NR_EXT_DEVT (1 << MINORBITS)
Tejun Heobcce3de2008-08-25 19:47:22 +090030
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
Adrian Bunk1826ead2008-03-04 11:23:46 +010037static struct device_type disk_type;
38
Derek Basehore12c2bdb2012-12-18 12:27:20 -080039static void disk_check_events(struct disk_events *ev,
40 unsigned int *clearing_ptr);
Stanislaw Gruszka9f53d2fe2012-03-02 10:43:28 +010041static void disk_alloc_events(struct gendisk *disk);
Tejun Heo77ea8872010-12-08 20:57:37 +010042static void disk_add_events(struct gendisk *disk);
43static void disk_del_events(struct gendisk *disk);
44static void disk_release_events(struct gendisk *disk);
45
Tejun Heoe71bf0d2008-09-03 09:03:02 +020046/**
47 * disk_get_part - get partition
48 * @disk: disk to look partition from
49 * @partno: partition number
50 *
51 * Look for partition @partno from @disk. If found, increment
52 * reference count and return it.
53 *
54 * CONTEXT:
55 * Don't care.
56 *
57 * RETURNS:
58 * Pointer to the found partition on success, NULL if not found.
59 */
60struct hd_struct *disk_get_part(struct gendisk *disk, int partno)
61{
Tejun Heo540eed52008-08-25 19:56:15 +090062 struct hd_struct *part = NULL;
63 struct disk_part_tbl *ptbl;
Tejun Heoe71bf0d2008-09-03 09:03:02 +020064
Tejun Heo540eed52008-08-25 19:56:15 +090065 if (unlikely(partno < 0))
Tejun Heoe71bf0d2008-09-03 09:03:02 +020066 return NULL;
Tejun Heo540eed52008-08-25 19:56:15 +090067
Tejun Heoe71bf0d2008-09-03 09:03:02 +020068 rcu_read_lock();
Tejun Heo540eed52008-08-25 19:56:15 +090069
70 ptbl = rcu_dereference(disk->part_tbl);
71 if (likely(partno < ptbl->len)) {
72 part = rcu_dereference(ptbl->part[partno]);
73 if (part)
74 get_device(part_to_dev(part));
75 }
76
Tejun Heoe71bf0d2008-09-03 09:03:02 +020077 rcu_read_unlock();
78
79 return part;
80}
81EXPORT_SYMBOL_GPL(disk_get_part);
82
83/**
84 * disk_part_iter_init - initialize partition iterator
85 * @piter: iterator to initialize
86 * @disk: disk to iterate over
87 * @flags: DISK_PITER_* flags
88 *
89 * Initialize @piter so that it iterates over partitions of @disk.
90 *
91 * CONTEXT:
92 * Don't care.
93 */
94void disk_part_iter_init(struct disk_part_iter *piter, struct gendisk *disk,
95 unsigned int flags)
96{
Tejun Heo540eed52008-08-25 19:56:15 +090097 struct disk_part_tbl *ptbl;
98
99 rcu_read_lock();
100 ptbl = rcu_dereference(disk->part_tbl);
101
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200102 piter->disk = disk;
103 piter->part = NULL;
104
105 if (flags & DISK_PITER_REVERSE)
Tejun Heo540eed52008-08-25 19:56:15 +0900106 piter->idx = ptbl->len - 1;
Tejun Heo71982a42009-04-17 08:34:48 +0200107 else if (flags & (DISK_PITER_INCL_PART0 | DISK_PITER_INCL_EMPTY_PART0))
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200108 piter->idx = 0;
Tejun Heob5d0b9d2008-09-03 09:06:42 +0200109 else
110 piter->idx = 1;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200111
112 piter->flags = flags;
Tejun Heo540eed52008-08-25 19:56:15 +0900113
114 rcu_read_unlock();
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200115}
116EXPORT_SYMBOL_GPL(disk_part_iter_init);
117
118/**
119 * disk_part_iter_next - proceed iterator to the next partition and return it
120 * @piter: iterator of interest
121 *
122 * Proceed @piter to the next partition and return it.
123 *
124 * CONTEXT:
125 * Don't care.
126 */
127struct hd_struct *disk_part_iter_next(struct disk_part_iter *piter)
128{
Tejun Heo540eed52008-08-25 19:56:15 +0900129 struct disk_part_tbl *ptbl;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200130 int inc, end;
131
132 /* put the last partition */
133 disk_put_part(piter->part);
134 piter->part = NULL;
135
Tejun Heo540eed52008-08-25 19:56:15 +0900136 /* get part_tbl */
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200137 rcu_read_lock();
Tejun Heo540eed52008-08-25 19:56:15 +0900138 ptbl = rcu_dereference(piter->disk->part_tbl);
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200139
140 /* determine iteration parameters */
141 if (piter->flags & DISK_PITER_REVERSE) {
142 inc = -1;
Tejun Heo71982a42009-04-17 08:34:48 +0200143 if (piter->flags & (DISK_PITER_INCL_PART0 |
144 DISK_PITER_INCL_EMPTY_PART0))
Tejun Heob5d0b9d2008-09-03 09:06:42 +0200145 end = -1;
146 else
147 end = 0;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200148 } else {
149 inc = 1;
Tejun Heo540eed52008-08-25 19:56:15 +0900150 end = ptbl->len;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200151 }
152
153 /* iterate to the next partition */
154 for (; piter->idx != end; piter->idx += inc) {
155 struct hd_struct *part;
156
Tejun Heo540eed52008-08-25 19:56:15 +0900157 part = rcu_dereference(ptbl->part[piter->idx]);
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200158 if (!part)
159 continue;
Vivek Goyalc83f6bf2012-08-01 12:24:18 +0200160 if (!part_nr_sects_read(part) &&
Tejun Heo71982a42009-04-17 08:34:48 +0200161 !(piter->flags & DISK_PITER_INCL_EMPTY) &&
162 !(piter->flags & DISK_PITER_INCL_EMPTY_PART0 &&
163 piter->idx == 0))
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200164 continue;
165
Tejun Heoed9e1982008-08-25 19:56:05 +0900166 get_device(part_to_dev(part));
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200167 piter->part = part;
168 piter->idx += inc;
169 break;
170 }
171
172 rcu_read_unlock();
173
174 return piter->part;
175}
176EXPORT_SYMBOL_GPL(disk_part_iter_next);
177
178/**
179 * disk_part_iter_exit - finish up partition iteration
180 * @piter: iter of interest
181 *
182 * Called when iteration is over. Cleans up @piter.
183 *
184 * CONTEXT:
185 * Don't care.
186 */
187void disk_part_iter_exit(struct disk_part_iter *piter)
188{
189 disk_put_part(piter->part);
190 piter->part = NULL;
191}
192EXPORT_SYMBOL_GPL(disk_part_iter_exit);
193
Jens Axboea6f23652008-10-24 12:52:42 +0200194static inline int sector_in_part(struct hd_struct *part, sector_t sector)
195{
196 return part->start_sect <= sector &&
Vivek Goyalc83f6bf2012-08-01 12:24:18 +0200197 sector < part->start_sect + part_nr_sects_read(part);
Jens Axboea6f23652008-10-24 12:52:42 +0200198}
199
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200200/**
201 * disk_map_sector_rcu - map sector to partition
202 * @disk: gendisk of interest
203 * @sector: sector to map
204 *
205 * Find out which partition @sector maps to on @disk. This is
206 * primarily used for stats accounting.
207 *
208 * CONTEXT:
209 * RCU read locked. The returned partition pointer is valid only
210 * while preemption is disabled.
211 *
212 * RETURNS:
Tejun Heo074a7ac2008-08-25 19:56:14 +0900213 * Found partition on success, part0 is returned if no partition matches
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200214 */
215struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector)
216{
Tejun Heo540eed52008-08-25 19:56:15 +0900217 struct disk_part_tbl *ptbl;
Jens Axboea6f23652008-10-24 12:52:42 +0200218 struct hd_struct *part;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200219 int i;
220
Tejun Heo540eed52008-08-25 19:56:15 +0900221 ptbl = rcu_dereference(disk->part_tbl);
222
Jens Axboea6f23652008-10-24 12:52:42 +0200223 part = rcu_dereference(ptbl->last_lookup);
224 if (part && sector_in_part(part, sector))
225 return part;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200226
Jens Axboea6f23652008-10-24 12:52:42 +0200227 for (i = 1; i < ptbl->len; i++) {
228 part = rcu_dereference(ptbl->part[i]);
229
230 if (part && sector_in_part(part, sector)) {
231 rcu_assign_pointer(ptbl->last_lookup, part);
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200232 return part;
Jens Axboea6f23652008-10-24 12:52:42 +0200233 }
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200234 }
Tejun Heo074a7ac2008-08-25 19:56:14 +0900235 return &disk->part0;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200236}
237EXPORT_SYMBOL_GPL(disk_map_sector_rcu);
238
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239/*
240 * Can be deleted altogether. Later.
241 *
242 */
243static struct blk_major_name {
244 struct blk_major_name *next;
245 int major;
246 char name[16];
Joe Korty68eef3b2006-03-31 02:30:32 -0800247} *major_names[BLKDEV_MAJOR_HASH_SIZE];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248
249/* index in the above - for now: assume no multimajor ranges */
Yang Zhange61eb2e2010-12-17 09:00:18 +0100250static inline int major_to_index(unsigned major)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251{
Joe Korty68eef3b2006-03-31 02:30:32 -0800252 return major % BLKDEV_MAJOR_HASH_SIZE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253}
254
Joe Korty68eef3b2006-03-31 02:30:32 -0800255#ifdef CONFIG_PROC_FS
Tejun Heocf771cb2008-09-03 09:01:09 +0200256void blkdev_show(struct seq_file *seqf, off_t offset)
Neil Horman7170be52006-01-14 13:20:38 -0800257{
Joe Korty68eef3b2006-03-31 02:30:32 -0800258 struct blk_major_name *dp;
Neil Horman7170be52006-01-14 13:20:38 -0800259
Joe Korty68eef3b2006-03-31 02:30:32 -0800260 if (offset < BLKDEV_MAJOR_HASH_SIZE) {
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200261 mutex_lock(&block_class_lock);
Joe Korty68eef3b2006-03-31 02:30:32 -0800262 for (dp = major_names[offset]; dp; dp = dp->next)
Tejun Heocf771cb2008-09-03 09:01:09 +0200263 seq_printf(seqf, "%3d %s\n", dp->major, dp->name);
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200264 mutex_unlock(&block_class_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700265 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266}
Joe Korty68eef3b2006-03-31 02:30:32 -0800267#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268
Márton Németh9e8c0bc2009-02-20 08:12:51 +0100269/**
270 * register_blkdev - register a new block device
271 *
272 * @major: the requested major device number [1..255]. If @major=0, try to
273 * allocate any unused major number.
274 * @name: the name of the new block device as a zero terminated string
275 *
276 * The @name must be unique within the system.
277 *
278 * The return value depends on the @major input parameter.
279 * - if a major device number was requested in range [1..255] then the
280 * function returns zero on success, or a negative error code
281 * - if any unused major number was requested with @major=0 parameter
282 * then the return value is the allocated major number in range
283 * [1..255] or a negative error code otherwise
284 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285int register_blkdev(unsigned int major, const char *name)
286{
287 struct blk_major_name **n, *p;
288 int index, ret = 0;
289
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200290 mutex_lock(&block_class_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700291
292 /* temporary */
293 if (major == 0) {
294 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
295 if (major_names[index] == NULL)
296 break;
297 }
298
299 if (index == 0) {
300 printk("register_blkdev: failed to get major for %s\n",
301 name);
302 ret = -EBUSY;
303 goto out;
304 }
305 major = index;
306 ret = major;
307 }
308
309 p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
310 if (p == NULL) {
311 ret = -ENOMEM;
312 goto out;
313 }
314
315 p->major = major;
316 strlcpy(p->name, name, sizeof(p->name));
317 p->next = NULL;
318 index = major_to_index(major);
319
320 for (n = &major_names[index]; *n; n = &(*n)->next) {
321 if ((*n)->major == major)
322 break;
323 }
324 if (!*n)
325 *n = p;
326 else
327 ret = -EBUSY;
328
329 if (ret < 0) {
330 printk("register_blkdev: cannot get major %d for %s\n",
331 major, name);
332 kfree(p);
333 }
334out:
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200335 mutex_unlock(&block_class_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336 return ret;
337}
338
339EXPORT_SYMBOL(register_blkdev);
340
Akinobu Mitaf4480242007-07-17 04:03:47 -0700341void unregister_blkdev(unsigned int major, const char *name)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342{
343 struct blk_major_name **n;
344 struct blk_major_name *p = NULL;
345 int index = major_to_index(major);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200347 mutex_lock(&block_class_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348 for (n = &major_names[index]; *n; n = &(*n)->next)
349 if ((*n)->major == major)
350 break;
Akinobu Mita294462a2007-07-17 04:03:45 -0700351 if (!*n || strcmp((*n)->name, name)) {
352 WARN_ON(1);
Akinobu Mita294462a2007-07-17 04:03:45 -0700353 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354 p = *n;
355 *n = p->next;
356 }
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200357 mutex_unlock(&block_class_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700358 kfree(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359}
360
361EXPORT_SYMBOL(unregister_blkdev);
362
363static struct kobj_map *bdev_map;
364
Tejun Heobcce3de2008-08-25 19:47:22 +0900365/**
Tejun Heo870d6652008-08-25 19:47:25 +0900366 * blk_mangle_minor - scatter minor numbers apart
367 * @minor: minor number to mangle
368 *
369 * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
370 * is enabled. Mangling twice gives the original value.
371 *
372 * RETURNS:
373 * Mangled value.
374 *
375 * CONTEXT:
376 * Don't care.
377 */
378static int blk_mangle_minor(int minor)
379{
380#ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
381 int i;
382
383 for (i = 0; i < MINORBITS / 2; i++) {
384 int low = minor & (1 << i);
385 int high = minor & (1 << (MINORBITS - 1 - i));
386 int distance = MINORBITS - 1 - 2 * i;
387
388 minor ^= low | high; /* clear both bits */
389 low <<= distance; /* swap the positions */
390 high >>= distance;
391 minor |= low | high; /* and set */
392 }
393#endif
394 return minor;
395}
396
397/**
Tejun Heobcce3de2008-08-25 19:47:22 +0900398 * blk_alloc_devt - allocate a dev_t for a partition
399 * @part: partition to allocate dev_t for
Tejun Heobcce3de2008-08-25 19:47:22 +0900400 * @devt: out parameter for resulting dev_t
401 *
402 * Allocate a dev_t for block device.
403 *
404 * RETURNS:
405 * 0 on success, allocated dev_t is returned in *@devt. -errno on
406 * failure.
407 *
408 * CONTEXT:
409 * Might sleep.
410 */
411int blk_alloc_devt(struct hd_struct *part, dev_t *devt)
412{
413 struct gendisk *disk = part_to_disk(part);
Tejun Heobab998d2013-02-27 17:03:57 -0800414 int idx;
Tejun Heobcce3de2008-08-25 19:47:22 +0900415
416 /* in consecutive minor range? */
417 if (part->partno < disk->minors) {
418 *devt = MKDEV(disk->major, disk->first_minor + part->partno);
419 return 0;
420 }
421
422 /* allocate ext devt */
Tejun Heobab998d2013-02-27 17:03:57 -0800423 mutex_lock(&ext_devt_mutex);
424 idx = idr_alloc(&ext_devt_idr, part, 0, NR_EXT_DEVT, GFP_KERNEL);
425 mutex_unlock(&ext_devt_mutex);
426 if (idx < 0)
427 return idx == -ENOSPC ? -EBUSY : idx;
Tejun Heobcce3de2008-08-25 19:47:22 +0900428
Tejun Heo870d6652008-08-25 19:47:25 +0900429 *devt = MKDEV(BLOCK_EXT_MAJOR, blk_mangle_minor(idx));
Tejun Heobcce3de2008-08-25 19:47:22 +0900430 return 0;
431}
432
433/**
434 * blk_free_devt - free a dev_t
435 * @devt: dev_t to free
436 *
437 * Free @devt which was allocated using blk_alloc_devt().
438 *
439 * CONTEXT:
440 * Might sleep.
441 */
442void blk_free_devt(dev_t devt)
443{
444 might_sleep();
445
446 if (devt == MKDEV(0, 0))
447 return;
448
449 if (MAJOR(devt) == BLOCK_EXT_MAJOR) {
450 mutex_lock(&ext_devt_mutex);
Tejun Heo870d6652008-08-25 19:47:25 +0900451 idr_remove(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
Tejun Heobcce3de2008-08-25 19:47:22 +0900452 mutex_unlock(&ext_devt_mutex);
453 }
454}
455
Tejun Heo1f014292008-08-25 19:47:23 +0900456static char *bdevt_str(dev_t devt, char *buf)
457{
458 if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
459 char tbuf[BDEVT_SIZE];
460 snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
461 snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
462 } else
463 snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
464
465 return buf;
466}
467
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468/*
469 * Register device numbers dev..(dev+range-1)
470 * range must be nonzero
471 * The hash chain is sorted on range, so that subranges can override.
472 */
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200473void blk_register_region(dev_t devt, unsigned long range, struct module *module,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474 struct kobject *(*probe)(dev_t, int *, void *),
475 int (*lock)(dev_t, void *), void *data)
476{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200477 kobj_map(bdev_map, devt, range, module, probe, lock, data);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700478}
479
480EXPORT_SYMBOL(blk_register_region);
481
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200482void blk_unregister_region(dev_t devt, unsigned long range)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200484 kobj_unmap(bdev_map, devt, range);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485}
486
487EXPORT_SYMBOL(blk_unregister_region);
488
Tejun Heocf771cb2008-09-03 09:01:09 +0200489static struct kobject *exact_match(dev_t devt, int *partno, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490{
491 struct gendisk *p = data;
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200492
Tejun Heoed9e1982008-08-25 19:56:05 +0900493 return &disk_to_dev(p)->kobj;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494}
495
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200496static int exact_lock(dev_t devt, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497{
498 struct gendisk *p = data;
499
500 if (!get_disk(p))
501 return -1;
502 return 0;
503}
504
Al Viro4752bc32011-09-16 00:21:54 -0400505static void register_disk(struct gendisk *disk)
Tejun Heod2bf1b62010-12-08 20:57:36 +0100506{
507 struct device *ddev = disk_to_dev(disk);
508 struct block_device *bdev;
509 struct disk_part_iter piter;
510 struct hd_struct *part;
511 int err;
512
513 ddev->parent = disk->driverfs_dev;
514
Kees Cookffc8b302013-07-03 15:01:14 -0700515 dev_set_name(ddev, "%s", disk->disk_name);
Tejun Heod2bf1b62010-12-08 20:57:36 +0100516
517 /* delay uevents, until we scanned partition table */
518 dev_set_uevent_suppress(ddev, 1);
519
520 if (device_add(ddev))
521 return;
522 if (!sysfs_deprecated) {
523 err = sysfs_create_link(block_depr, &ddev->kobj,
524 kobject_name(&ddev->kobj));
525 if (err) {
526 device_del(ddev);
527 return;
528 }
529 }
Ming Lei25e823c2013-02-22 16:34:13 -0800530
531 /*
532 * avoid probable deadlock caused by allocating memory with
533 * GFP_KERNEL in runtime_resume callback of its all ancestor
534 * devices
535 */
536 pm_runtime_set_memalloc_noio(ddev, true);
537
Tejun Heod2bf1b62010-12-08 20:57:36 +0100538 disk->part0.holder_dir = kobject_create_and_add("holders", &ddev->kobj);
539 disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
540
541 /* No minors to use for partitions */
Tejun Heod27769e2011-08-23 20:01:04 +0200542 if (!disk_part_scan_enabled(disk))
Tejun Heod2bf1b62010-12-08 20:57:36 +0100543 goto exit;
544
545 /* No such device (e.g., media were just removed) */
546 if (!get_capacity(disk))
547 goto exit;
548
549 bdev = bdget_disk(disk, 0);
550 if (!bdev)
551 goto exit;
552
553 bdev->bd_invalidated = 1;
554 err = blkdev_get(bdev, FMODE_READ, NULL);
555 if (err < 0)
556 goto exit;
557 blkdev_put(bdev, FMODE_READ);
558
559exit:
560 /* announce disk after possible partitions are created */
561 dev_set_uevent_suppress(ddev, 0);
562 kobject_uevent(&ddev->kobj, KOBJ_ADD);
563
564 /* announce possible partitions */
565 disk_part_iter_init(&piter, disk, 0);
566 while ((part = disk_part_iter_next(&piter)))
567 kobject_uevent(&part_to_dev(part)->kobj, KOBJ_ADD);
568 disk_part_iter_exit(&piter);
569}
570
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571/**
572 * add_disk - add partitioning information to kernel list
573 * @disk: per-device partitioning information
574 *
575 * This function registers the partitioning information in @disk
576 * with the kernel.
Tejun Heo3e1a7ff2008-08-25 19:56:17 +0900577 *
578 * FIXME: error handling
Linus Torvalds1da177e2005-04-16 15:20:36 -0700579 */
580void add_disk(struct gendisk *disk)
581{
Peter Zijlstracf0ca9f2008-04-30 00:54:32 -0700582 struct backing_dev_info *bdi;
Tejun Heo3e1a7ff2008-08-25 19:56:17 +0900583 dev_t devt;
Greg Kroah-Hartman6ffeea72008-05-22 17:21:08 -0400584 int retval;
Peter Zijlstracf0ca9f2008-04-30 00:54:32 -0700585
Tejun Heo3e1a7ff2008-08-25 19:56:17 +0900586 /* minors == 0 indicates to use ext devt from part0 and should
587 * be accompanied with EXT_DEVT flag. Make sure all
588 * parameters make sense.
589 */
590 WARN_ON(disk->minors && !(disk->major || disk->first_minor));
591 WARN_ON(!disk->minors && !(disk->flags & GENHD_FL_EXT_DEVT));
592
Linus Torvalds1da177e2005-04-16 15:20:36 -0700593 disk->flags |= GENHD_FL_UP;
Tejun Heo3e1a7ff2008-08-25 19:56:17 +0900594
595 retval = blk_alloc_devt(&disk->part0, &devt);
596 if (retval) {
597 WARN_ON(1);
598 return;
599 }
600 disk_to_dev(disk)->devt = devt;
601
602 /* ->major and ->first_minor aren't supposed to be
603 * dereferenced from here on, but set them just in case.
604 */
605 disk->major = MAJOR(devt);
606 disk->first_minor = MINOR(devt);
607
Stanislaw Gruszka9f53d2fe2012-03-02 10:43:28 +0100608 disk_alloc_events(disk);
609
Wanlong Gao9f5e4862011-06-13 10:45:43 +0200610 /* Register BDI before referencing it from bdev */
Signed-off-by: Jan Kara01ea5062010-09-16 20:36:36 +0200611 bdi = &disk->queue->backing_dev_info;
612 bdi_register_dev(bdi, disk_devt(disk));
613
Tejun Heof331c022008-09-03 09:01:48 +0200614 blk_register_region(disk_devt(disk), disk->minors, NULL,
615 exact_match, exact_lock, disk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616 register_disk(disk);
617 blk_register_queue(disk);
Peter Zijlstracf0ca9f2008-04-30 00:54:32 -0700618
Tejun Heo523e1d32011-10-19 14:31:07 +0200619 /*
620 * Take an extra ref on queue which will be put on disk_release()
621 * so that it sticks around as long as @disk is there.
622 */
Tejun Heo09ac46c2011-12-14 00:33:38 +0100623 WARN_ON_ONCE(!blk_get_queue(disk->queue));
Tejun Heo523e1d32011-10-19 14:31:07 +0200624
Tejun Heoed9e1982008-08-25 19:56:05 +0900625 retval = sysfs_create_link(&disk_to_dev(disk)->kobj, &bdi->dev->kobj,
626 "bdi");
Greg Kroah-Hartman6ffeea72008-05-22 17:21:08 -0400627 WARN_ON(retval);
Tejun Heo77ea8872010-12-08 20:57:37 +0100628
629 disk_add_events(disk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700630}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631EXPORT_SYMBOL(add_disk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632
Tejun Heod2bf1b62010-12-08 20:57:36 +0100633void del_gendisk(struct gendisk *disk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634{
Tejun Heod2bf1b62010-12-08 20:57:36 +0100635 struct disk_part_iter piter;
636 struct hd_struct *part;
637
Tejun Heo77ea8872010-12-08 20:57:37 +0100638 disk_del_events(disk);
639
Tejun Heod2bf1b62010-12-08 20:57:36 +0100640 /* invalidate stuff */
641 disk_part_iter_init(&piter, disk,
642 DISK_PITER_INCL_EMPTY | DISK_PITER_REVERSE);
643 while ((part = disk_part_iter_next(&piter))) {
644 invalidate_partition(disk, part->partno);
645 delete_partition(disk, part->partno);
646 }
647 disk_part_iter_exit(&piter);
648
649 invalidate_partition(disk, 0);
Tejun Heod2bf1b62010-12-08 20:57:36 +0100650 set_capacity(disk, 0);
651 disk->flags &= ~GENHD_FL_UP;
652
Tejun Heoed9e1982008-08-25 19:56:05 +0900653 sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
Peter Zijlstracf0ca9f2008-04-30 00:54:32 -0700654 bdi_unregister(&disk->queue->backing_dev_info);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655 blk_unregister_queue(disk);
Tejun Heof331c022008-09-03 09:01:48 +0200656 blk_unregister_region(disk_devt(disk), disk->minors);
Tejun Heod2bf1b62010-12-08 20:57:36 +0100657
658 part_stat_set_all(&disk->part0, 0);
659 disk->part0.stamp = 0;
660
661 kobject_put(disk->part0.holder_dir);
662 kobject_put(disk->slave_dir);
663 disk->driverfs_dev = NULL;
664 if (!sysfs_deprecated)
665 sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
Ming Lei25e823c2013-02-22 16:34:13 -0800666 pm_runtime_set_memalloc_noio(disk_to_dev(disk), false);
Tejun Heod2bf1b62010-12-08 20:57:36 +0100667 device_del(disk_to_dev(disk));
Tomas Henzl7b74e912013-02-27 17:03:32 -0800668 blk_free_devt(disk_to_dev(disk)->devt);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669}
Tejun Heod2bf1b62010-12-08 20:57:36 +0100670EXPORT_SYMBOL(del_gendisk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671
Linus Torvalds1da177e2005-04-16 15:20:36 -0700672/**
673 * get_gendisk - get partitioning information for a given device
Randy Dunlap710027a2008-08-19 20:13:11 +0200674 * @devt: device to get partitioning information for
Randy Dunlap496aa8a2008-10-16 07:46:23 +0200675 * @partno: returned partition index
Linus Torvalds1da177e2005-04-16 15:20:36 -0700676 *
677 * This function gets the structure containing partitioning
Randy Dunlap710027a2008-08-19 20:13:11 +0200678 * information for the given device @devt.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679 */
Tejun Heocf771cb2008-09-03 09:01:09 +0200680struct gendisk *get_gendisk(dev_t devt, int *partno)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700681{
Tejun Heobcce3de2008-08-25 19:47:22 +0900682 struct gendisk *disk = NULL;
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200683
Tejun Heobcce3de2008-08-25 19:47:22 +0900684 if (MAJOR(devt) != BLOCK_EXT_MAJOR) {
685 struct kobject *kobj;
686
687 kobj = kobj_lookup(bdev_map, devt, partno);
688 if (kobj)
689 disk = dev_to_disk(kobj_to_dev(kobj));
690 } else {
691 struct hd_struct *part;
692
693 mutex_lock(&ext_devt_mutex);
Tejun Heo870d6652008-08-25 19:47:25 +0900694 part = idr_find(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
Tejun Heobcce3de2008-08-25 19:47:22 +0900695 if (part && get_disk(part_to_disk(part))) {
696 *partno = part->partno;
697 disk = part_to_disk(part);
698 }
699 mutex_unlock(&ext_devt_mutex);
700 }
701
702 return disk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700703}
Divyesh Shahb6ac23af2010-04-15 08:54:59 +0200704EXPORT_SYMBOL(get_gendisk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705
Tejun Heof331c022008-09-03 09:01:48 +0200706/**
707 * bdget_disk - do bdget() by gendisk and partition number
708 * @disk: gendisk of interest
709 * @partno: partition number
710 *
711 * Find partition @partno from @disk, do bdget() on it.
712 *
713 * CONTEXT:
714 * Don't care.
715 *
716 * RETURNS:
717 * Resulting block_device on success, NULL on failure.
718 */
Harvey Harrisonaeb3d3a2008-08-28 09:27:42 +0200719struct block_device *bdget_disk(struct gendisk *disk, int partno)
Tejun Heof331c022008-09-03 09:01:48 +0200720{
Tejun Heo548b10e2008-08-29 09:01:47 +0200721 struct hd_struct *part;
722 struct block_device *bdev = NULL;
Tejun Heof331c022008-09-03 09:01:48 +0200723
Tejun Heo548b10e2008-08-29 09:01:47 +0200724 part = disk_get_part(disk, partno);
Tejun Heo2bbedcb2008-08-29 11:41:51 +0200725 if (part)
Tejun Heo548b10e2008-08-29 09:01:47 +0200726 bdev = bdget(part_devt(part));
727 disk_put_part(part);
Tejun Heof331c022008-09-03 09:01:48 +0200728
Tejun Heo548b10e2008-08-29 09:01:47 +0200729 return bdev;
Tejun Heof331c022008-09-03 09:01:48 +0200730}
731EXPORT_SYMBOL(bdget_disk);
732
Dave Gilbertdd2a3452007-05-09 02:33:24 -0700733/*
734 * print a full list of all partitions - intended for places where the root
735 * filesystem can't be mounted and thus to give the victim some idea of what
736 * went wrong
737 */
738void __init printk_all_partitions(void)
739{
Tejun Heodef4e382008-09-03 08:57:12 +0200740 struct class_dev_iter iter;
741 struct device *dev;
742
743 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
744 while ((dev = class_dev_iter_next(&iter))) {
745 struct gendisk *disk = dev_to_disk(dev);
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200746 struct disk_part_iter piter;
747 struct hd_struct *part;
Tejun Heo1f014292008-08-25 19:47:23 +0900748 char name_buf[BDEVNAME_SIZE];
749 char devt_buf[BDEVT_SIZE];
Tejun Heodef4e382008-09-03 08:57:12 +0200750
751 /*
752 * Don't show empty devices or things that have been
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300753 * suppressed
Tejun Heodef4e382008-09-03 08:57:12 +0200754 */
755 if (get_capacity(disk) == 0 ||
756 (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
757 continue;
758
759 /*
760 * Note, unlike /proc/partitions, I am showing the
761 * numbers in hex - the same format as the root=
762 * option takes.
763 */
Tejun Heo074a7ac2008-08-25 19:56:14 +0900764 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0);
765 while ((part = disk_part_iter_next(&piter))) {
766 bool is_part0 = part == &disk->part0;
Tejun Heodef4e382008-09-03 08:57:12 +0200767
Will Drewryb5af9212010-08-31 15:47:07 -0500768 printk("%s%s %10llu %s %s", is_part0 ? "" : " ",
Tejun Heo1f014292008-08-25 19:47:23 +0900769 bdevt_str(part_devt(part), devt_buf),
Vivek Goyalc83f6bf2012-08-01 12:24:18 +0200770 (unsigned long long)part_nr_sects_read(part) >> 1
771 , disk_name(disk, part->partno, name_buf),
Stephen Warren1ad7e892012-11-08 16:12:25 -0800772 part->info ? part->info->uuid : "");
Tejun Heo074a7ac2008-08-25 19:56:14 +0900773 if (is_part0) {
774 if (disk->driverfs_dev != NULL &&
775 disk->driverfs_dev->driver != NULL)
776 printk(" driver: %s\n",
777 disk->driverfs_dev->driver->name);
778 else
779 printk(" (driver?)\n");
780 } else
781 printk("\n");
782 }
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200783 disk_part_iter_exit(&piter);
Tejun Heodef4e382008-09-03 08:57:12 +0200784 }
785 class_dev_iter_exit(&iter);
Dave Gilbertdd2a3452007-05-09 02:33:24 -0700786}
787
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788#ifdef CONFIG_PROC_FS
789/* iterator */
Tejun Heodef4e382008-09-03 08:57:12 +0200790static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
Greg Kroah-Hartman68c4d4a2008-05-22 17:21:08 -0400791{
Tejun Heodef4e382008-09-03 08:57:12 +0200792 loff_t skip = *pos;
793 struct class_dev_iter *iter;
794 struct device *dev;
Greg Kroah-Hartman68c4d4a2008-05-22 17:21:08 -0400795
Harvey Harrisonaeb3d3a2008-08-28 09:27:42 +0200796 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
Tejun Heodef4e382008-09-03 08:57:12 +0200797 if (!iter)
798 return ERR_PTR(-ENOMEM);
799
800 seqf->private = iter;
801 class_dev_iter_init(iter, &block_class, NULL, &disk_type);
802 do {
803 dev = class_dev_iter_next(iter);
804 if (!dev)
805 return NULL;
806 } while (skip--);
807
808 return dev_to_disk(dev);
Greg Kroah-Hartman68c4d4a2008-05-22 17:21:08 -0400809}
810
Tejun Heodef4e382008-09-03 08:57:12 +0200811static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200813 struct device *dev;
Greg Kroah-Hartman66c64af2008-05-22 17:21:08 -0400814
Tejun Heodef4e382008-09-03 08:57:12 +0200815 (*pos)++;
816 dev = class_dev_iter_next(seqf->private);
Tejun Heo2ac3cee2008-09-03 08:53:37 +0200817 if (dev)
Greg Kroah-Hartman68c4d4a2008-05-22 17:21:08 -0400818 return dev_to_disk(dev);
Tejun Heo2ac3cee2008-09-03 08:53:37 +0200819
Linus Torvalds1da177e2005-04-16 15:20:36 -0700820 return NULL;
821}
822
Tejun Heodef4e382008-09-03 08:57:12 +0200823static void disk_seqf_stop(struct seq_file *seqf, void *v)
Greg Kroah-Hartman27f30252008-05-22 17:21:08 -0400824{
Tejun Heodef4e382008-09-03 08:57:12 +0200825 struct class_dev_iter *iter = seqf->private;
Greg Kroah-Hartman27f30252008-05-22 17:21:08 -0400826
Tejun Heodef4e382008-09-03 08:57:12 +0200827 /* stop is called even after start failed :-( */
828 if (iter) {
829 class_dev_iter_exit(iter);
830 kfree(iter);
Kay Sievers5c0ef6d2008-08-16 14:30:30 +0200831 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700832}
833
Tejun Heodef4e382008-09-03 08:57:12 +0200834static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835{
Jianpeng Ma06768062012-08-03 10:42:00 +0200836 void *p;
Tejun Heodef4e382008-09-03 08:57:12 +0200837
838 p = disk_seqf_start(seqf, pos);
Yang Zhangb9f985b2010-12-17 08:58:36 +0100839 if (!IS_ERR_OR_NULL(p) && !*pos)
Tejun Heodef4e382008-09-03 08:57:12 +0200840 seq_puts(seqf, "major minor #blocks name\n\n");
841 return p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700842}
843
Tejun Heocf771cb2008-09-03 09:01:09 +0200844static int show_partition(struct seq_file *seqf, void *v)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845{
846 struct gendisk *sgp = v;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200847 struct disk_part_iter piter;
848 struct hd_struct *part;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849 char buf[BDEVNAME_SIZE];
850
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851 /* Don't show non-partitionable removeable devices or empty devices */
Tejun Heod27769e2011-08-23 20:01:04 +0200852 if (!get_capacity(sgp) || (!disk_max_parts(sgp) &&
Tejun Heof331c022008-09-03 09:01:48 +0200853 (sgp->flags & GENHD_FL_REMOVABLE)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854 return 0;
855 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
856 return 0;
857
858 /* show the full disk and all non-0 size partitions of it */
Tejun Heo074a7ac2008-08-25 19:56:14 +0900859 disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0);
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200860 while ((part = disk_part_iter_next(&piter)))
Tejun Heo1f014292008-08-25 19:47:23 +0900861 seq_printf(seqf, "%4d %7d %10llu %s\n",
Tejun Heof331c022008-09-03 09:01:48 +0200862 MAJOR(part_devt(part)), MINOR(part_devt(part)),
Vivek Goyalc83f6bf2012-08-01 12:24:18 +0200863 (unsigned long long)part_nr_sects_read(part) >> 1,
Tejun Heof331c022008-09-03 09:01:48 +0200864 disk_name(sgp, part->partno, buf));
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200865 disk_part_iter_exit(&piter);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866
867 return 0;
868}
869
Alexey Dobriyanf5009752008-10-04 23:53:21 +0400870static const struct seq_operations partitions_op = {
Tejun Heodef4e382008-09-03 08:57:12 +0200871 .start = show_partition_start,
872 .next = disk_seqf_next,
873 .stop = disk_seqf_stop,
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200874 .show = show_partition
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875};
Alexey Dobriyanf5009752008-10-04 23:53:21 +0400876
877static int partitions_open(struct inode *inode, struct file *file)
878{
879 return seq_open(file, &partitions_op);
880}
881
882static const struct file_operations proc_partitions_operations = {
883 .open = partitions_open,
884 .read = seq_read,
885 .llseek = seq_lseek,
886 .release = seq_release,
887};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888#endif
889
890
Tejun Heocf771cb2008-09-03 09:01:09 +0200891static struct kobject *base_probe(dev_t devt, int *partno, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700892{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200893 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894 /* Make old-style 2.4 aliases work */
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200895 request_module("block-major-%d", MAJOR(devt));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896 return NULL;
897}
898
899static int __init genhd_device_init(void)
900{
Dan Williamse105b8b2008-04-21 10:51:07 -0700901 int error;
902
903 block_class.dev_kobj = sysfs_dev_block_kobj;
904 error = class_register(&block_class);
Roland McGrathee27a552008-03-11 17:13:15 -0700905 if (unlikely(error))
906 return error;
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200907 bdev_map = kobj_map_init(base_probe, &block_class_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700908 blk_dev_init();
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200909
Zhang, Yanmin561ec682008-11-14 08:26:30 +0100910 register_blkdev(BLOCK_EXT_MAJOR, "blkext");
911
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200912 /* create top-level block dir */
Andi Kleene52eec12010-09-08 16:54:17 +0200913 if (!sysfs_deprecated)
914 block_depr = kobject_create_and_add("block", NULL);
Greg Kroah-Hartman830d3cf2007-11-06 10:36:58 -0800915 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916}
917
918subsys_initcall(genhd_device_init);
919
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200920static ssize_t disk_range_show(struct device *dev,
921 struct device_attribute *attr, char *buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200923 struct gendisk *disk = dev_to_disk(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200925 return sprintf(buf, "%d\n", disk->minors);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926}
927
Tejun Heo1f014292008-08-25 19:47:23 +0900928static ssize_t disk_ext_range_show(struct device *dev,
929 struct device_attribute *attr, char *buf)
930{
931 struct gendisk *disk = dev_to_disk(dev);
932
Tejun Heob5d0b9d2008-09-03 09:06:42 +0200933 return sprintf(buf, "%d\n", disk_max_parts(disk));
Tejun Heo1f014292008-08-25 19:47:23 +0900934}
935
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200936static ssize_t disk_removable_show(struct device *dev,
937 struct device_attribute *attr, char *buf)
Kay Sieversa7fd6702005-10-01 14:49:43 +0200938{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200939 struct gendisk *disk = dev_to_disk(dev);
Kay Sieversa7fd6702005-10-01 14:49:43 +0200940
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200941 return sprintf(buf, "%d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700942 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200943}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700944
Kay Sievers1c9ce522008-06-13 09:41:00 +0200945static ssize_t disk_ro_show(struct device *dev,
946 struct device_attribute *attr, char *buf)
947{
948 struct gendisk *disk = dev_to_disk(dev);
949
Tejun Heob7db9952008-08-25 19:56:10 +0900950 return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
Kay Sievers1c9ce522008-06-13 09:41:00 +0200951}
952
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200953static ssize_t disk_capability_show(struct device *dev,
954 struct device_attribute *attr, char *buf)
Kristen Carlson Accardi86ce18d2007-05-23 13:57:38 -0700955{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200956 struct gendisk *disk = dev_to_disk(dev);
957
958 return sprintf(buf, "%x\n", disk->flags);
Kristen Carlson Accardi86ce18d2007-05-23 13:57:38 -0700959}
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200960
Martin K. Petersenc72758f2009-05-22 17:17:53 -0400961static ssize_t disk_alignment_offset_show(struct device *dev,
962 struct device_attribute *attr,
963 char *buf)
964{
965 struct gendisk *disk = dev_to_disk(dev);
966
967 return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue));
968}
969
Martin K. Petersen86b37282009-11-10 11:50:21 +0100970static ssize_t disk_discard_alignment_show(struct device *dev,
971 struct device_attribute *attr,
972 char *buf)
973{
974 struct gendisk *disk = dev_to_disk(dev);
975
Martin K. Petersendd3d1452010-01-11 03:21:48 -0500976 return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue));
Martin K. Petersen86b37282009-11-10 11:50:21 +0100977}
978
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200979static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL);
Tejun Heo1f014292008-08-25 19:47:23 +0900980static DEVICE_ATTR(ext_range, S_IRUGO, disk_ext_range_show, NULL);
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200981static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL);
Kay Sievers1c9ce522008-06-13 09:41:00 +0200982static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL);
Tejun Heoe5610522008-08-25 19:56:09 +0900983static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL);
Martin K. Petersenc72758f2009-05-22 17:17:53 -0400984static DEVICE_ATTR(alignment_offset, S_IRUGO, disk_alignment_offset_show, NULL);
Martin K. Petersen86b37282009-11-10 11:50:21 +0100985static DEVICE_ATTR(discard_alignment, S_IRUGO, disk_discard_alignment_show,
986 NULL);
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200987static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL);
Tejun Heo074a7ac2008-08-25 19:56:14 +0900988static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL);
Nikanth Karthikesan316d3152009-10-06 20:16:55 +0200989static DEVICE_ATTR(inflight, S_IRUGO, part_inflight_show, NULL);
Akinobu Mitac17bb492006-12-08 02:39:46 -0800990#ifdef CONFIG_FAIL_MAKE_REQUEST
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200991static struct device_attribute dev_attr_fail =
Tejun Heoeddb2e22008-08-25 19:56:13 +0900992 __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store);
Akinobu Mitac17bb492006-12-08 02:39:46 -0800993#endif
Jens Axboe581d4e22008-09-14 05:56:33 -0700994#ifdef CONFIG_FAIL_IO_TIMEOUT
995static struct device_attribute dev_attr_fail_timeout =
996 __ATTR(io-timeout-fail, S_IRUGO|S_IWUSR, part_timeout_show,
997 part_timeout_store);
998#endif
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200999
1000static struct attribute *disk_attrs[] = {
1001 &dev_attr_range.attr,
Tejun Heo1f014292008-08-25 19:47:23 +09001002 &dev_attr_ext_range.attr,
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001003 &dev_attr_removable.attr,
Kay Sievers1c9ce522008-06-13 09:41:00 +02001004 &dev_attr_ro.attr,
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001005 &dev_attr_size.attr,
Martin K. Petersenc72758f2009-05-22 17:17:53 -04001006 &dev_attr_alignment_offset.attr,
Martin K. Petersen86b37282009-11-10 11:50:21 +01001007 &dev_attr_discard_alignment.attr,
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001008 &dev_attr_capability.attr,
1009 &dev_attr_stat.attr,
Nikanth Karthikesan316d3152009-10-06 20:16:55 +02001010 &dev_attr_inflight.attr,
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001011#ifdef CONFIG_FAIL_MAKE_REQUEST
1012 &dev_attr_fail.attr,
1013#endif
Jens Axboe581d4e22008-09-14 05:56:33 -07001014#ifdef CONFIG_FAIL_IO_TIMEOUT
1015 &dev_attr_fail_timeout.attr,
1016#endif
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001017 NULL
Linus Torvalds1da177e2005-04-16 15:20:36 -07001018};
1019
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001020static struct attribute_group disk_attr_group = {
1021 .attrs = disk_attrs,
1022};
1023
David Brownella4dbd672009-06-24 10:06:31 -07001024static const struct attribute_group *disk_attr_groups[] = {
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001025 &disk_attr_group,
1026 NULL
1027};
1028
Tejun Heo540eed52008-08-25 19:56:15 +09001029/**
1030 * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way
1031 * @disk: disk to replace part_tbl for
1032 * @new_ptbl: new part_tbl to install
1033 *
1034 * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The
1035 * original ptbl is freed using RCU callback.
1036 *
1037 * LOCKING:
1038 * Matching bd_mutx locked.
1039 */
1040static void disk_replace_part_tbl(struct gendisk *disk,
1041 struct disk_part_tbl *new_ptbl)
1042{
1043 struct disk_part_tbl *old_ptbl = disk->part_tbl;
1044
1045 rcu_assign_pointer(disk->part_tbl, new_ptbl);
Jens Axboea6f23652008-10-24 12:52:42 +02001046
1047 if (old_ptbl) {
1048 rcu_assign_pointer(old_ptbl->last_lookup, NULL);
Lai Jiangshan57bdfbf2011-03-18 11:42:58 +08001049 kfree_rcu(old_ptbl, rcu_head);
Jens Axboea6f23652008-10-24 12:52:42 +02001050 }
Tejun Heo540eed52008-08-25 19:56:15 +09001051}
1052
1053/**
1054 * disk_expand_part_tbl - expand disk->part_tbl
1055 * @disk: disk to expand part_tbl for
1056 * @partno: expand such that this partno can fit in
1057 *
1058 * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl
1059 * uses RCU to allow unlocked dereferencing for stats and other stuff.
1060 *
1061 * LOCKING:
1062 * Matching bd_mutex locked, might sleep.
1063 *
1064 * RETURNS:
1065 * 0 on success, -errno on failure.
1066 */
1067int disk_expand_part_tbl(struct gendisk *disk, int partno)
1068{
1069 struct disk_part_tbl *old_ptbl = disk->part_tbl;
1070 struct disk_part_tbl *new_ptbl;
1071 int len = old_ptbl ? old_ptbl->len : 0;
1072 int target = partno + 1;
1073 size_t size;
1074 int i;
1075
1076 /* disk_max_parts() is zero during initialization, ignore if so */
1077 if (disk_max_parts(disk) && target > disk_max_parts(disk))
1078 return -EINVAL;
1079
1080 if (target <= len)
1081 return 0;
1082
1083 size = sizeof(*new_ptbl) + target * sizeof(new_ptbl->part[0]);
1084 new_ptbl = kzalloc_node(size, GFP_KERNEL, disk->node_id);
1085 if (!new_ptbl)
1086 return -ENOMEM;
1087
Tejun Heo540eed52008-08-25 19:56:15 +09001088 new_ptbl->len = target;
1089
1090 for (i = 0; i < len; i++)
1091 rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]);
1092
1093 disk_replace_part_tbl(disk, new_ptbl);
1094 return 0;
1095}
1096
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001097static void disk_release(struct device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001098{
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001099 struct gendisk *disk = dev_to_disk(dev);
1100
Tejun Heo77ea8872010-12-08 20:57:37 +01001101 disk_release_events(disk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102 kfree(disk->random);
Tejun Heo540eed52008-08-25 19:56:15 +09001103 disk_replace_part_tbl(disk, NULL);
Tejun Heo074a7ac2008-08-25 19:56:14 +09001104 free_part_stats(&disk->part0);
Will Drewry6d1d8052010-08-31 15:47:05 -05001105 free_part_info(&disk->part0);
Tejun Heo523e1d32011-10-19 14:31:07 +02001106 if (disk->queue)
1107 blk_put_queue(disk->queue);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108 kfree(disk);
1109}
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001110struct class block_class = {
1111 .name = "block",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112};
1113
Kay Sievers3c2670e2013-04-06 09:56:00 -07001114static char *block_devnode(struct device *dev, umode_t *mode,
Greg Kroah-Hartman4e4098a2013-04-11 11:43:29 -07001115 kuid_t *uid, kgid_t *gid)
Kay Sieversb03f38b2009-04-30 15:23:42 +02001116{
1117 struct gendisk *disk = dev_to_disk(dev);
1118
Kay Sieverse454cea2009-09-18 23:01:12 +02001119 if (disk->devnode)
1120 return disk->devnode(disk, mode);
Kay Sieversb03f38b2009-04-30 15:23:42 +02001121 return NULL;
1122}
1123
Adrian Bunk1826ead2008-03-04 11:23:46 +01001124static struct device_type disk_type = {
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001125 .name = "disk",
1126 .groups = disk_attr_groups,
1127 .release = disk_release,
Kay Sieverse454cea2009-09-18 23:01:12 +02001128 .devnode = block_devnode,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001129};
1130
Randy Dunlapa6e2ba82008-05-23 09:44:11 -07001131#ifdef CONFIG_PROC_FS
Tejun Heocf771cb2008-09-03 09:01:09 +02001132/*
1133 * aggregate disk stat collector. Uses the same stats that the sysfs
1134 * entries do, above, but makes them available through one seq_file.
1135 *
1136 * The output looks suspiciously like /proc/partitions with a bunch of
1137 * extra fields.
1138 */
1139static int diskstats_show(struct seq_file *seqf, void *v)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001140{
1141 struct gendisk *gp = v;
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001142 struct disk_part_iter piter;
1143 struct hd_struct *hd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001144 char buf[BDEVNAME_SIZE];
Tejun Heoc9959052008-08-25 19:47:21 +09001145 int cpu;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001146
1147 /*
Tejun Heoed9e1982008-08-25 19:56:05 +09001148 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
Tejun Heocf771cb2008-09-03 09:01:09 +02001149 seq_puts(seqf, "major minor name"
Linus Torvalds1da177e2005-04-16 15:20:36 -07001150 " rio rmerge rsect ruse wio wmerge "
1151 "wsect wuse running use aveq"
1152 "\n\n");
1153 */
Wanlong Gao9f5e4862011-06-13 10:45:43 +02001154
Tejun Heo71982a42009-04-17 08:34:48 +02001155 disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0);
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001156 while ((hd = disk_part_iter_next(&piter))) {
Tejun Heo074a7ac2008-08-25 19:56:14 +09001157 cpu = part_stat_lock();
Tejun Heoc9959052008-08-25 19:47:21 +09001158 part_round_stats(cpu, hd);
Tejun Heo074a7ac2008-08-25 19:56:14 +09001159 part_stat_unlock();
Herbert Poetzlf95fe9c2011-08-02 12:43:50 +02001160 seq_printf(seqf, "%4d %7d %s %lu %lu %lu "
1161 "%u %lu %lu %lu %u %u %u %u\n",
Tejun Heof331c022008-09-03 09:01:48 +02001162 MAJOR(part_devt(hd)), MINOR(part_devt(hd)),
1163 disk_name(gp, hd->partno, buf),
Liu Yuan53f22952011-03-02 11:00:15 -05001164 part_stat_read(hd, ios[READ]),
1165 part_stat_read(hd, merges[READ]),
Herbert Poetzlf95fe9c2011-08-02 12:43:50 +02001166 part_stat_read(hd, sectors[READ]),
Liu Yuan53f22952011-03-02 11:00:15 -05001167 jiffies_to_msecs(part_stat_read(hd, ticks[READ])),
1168 part_stat_read(hd, ios[WRITE]),
1169 part_stat_read(hd, merges[WRITE]),
Herbert Poetzlf95fe9c2011-08-02 12:43:50 +02001170 part_stat_read(hd, sectors[WRITE]),
Liu Yuan53f22952011-03-02 11:00:15 -05001171 jiffies_to_msecs(part_stat_read(hd, ticks[WRITE])),
Nikanth Karthikesan316d3152009-10-06 20:16:55 +02001172 part_in_flight(hd),
Jerome Marchand28f39d52008-02-08 11:04:56 +01001173 jiffies_to_msecs(part_stat_read(hd, io_ticks)),
1174 jiffies_to_msecs(part_stat_read(hd, time_in_queue))
1175 );
Linus Torvalds1da177e2005-04-16 15:20:36 -07001176 }
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001177 disk_part_iter_exit(&piter);
Wanlong Gao9f5e4862011-06-13 10:45:43 +02001178
Linus Torvalds1da177e2005-04-16 15:20:36 -07001179 return 0;
1180}
1181
Alexey Dobriyan31d85ab2008-10-06 12:55:38 +04001182static const struct seq_operations diskstats_op = {
Tejun Heodef4e382008-09-03 08:57:12 +02001183 .start = disk_seqf_start,
1184 .next = disk_seqf_next,
1185 .stop = disk_seqf_stop,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001186 .show = diskstats_show
1187};
Alexey Dobriyanf5009752008-10-04 23:53:21 +04001188
Alexey Dobriyan31d85ab2008-10-06 12:55:38 +04001189static int diskstats_open(struct inode *inode, struct file *file)
1190{
1191 return seq_open(file, &diskstats_op);
1192}
1193
1194static const struct file_operations proc_diskstats_operations = {
1195 .open = diskstats_open,
1196 .read = seq_read,
1197 .llseek = seq_lseek,
1198 .release = seq_release,
1199};
1200
Alexey Dobriyanf5009752008-10-04 23:53:21 +04001201static int __init proc_genhd_init(void)
1202{
Alexey Dobriyan31d85ab2008-10-06 12:55:38 +04001203 proc_create("diskstats", 0, NULL, &proc_diskstats_operations);
Alexey Dobriyanf5009752008-10-04 23:53:21 +04001204 proc_create("partitions", 0, NULL, &proc_partitions_operations);
1205 return 0;
1206}
1207module_init(proc_genhd_init);
Randy Dunlapa6e2ba82008-05-23 09:44:11 -07001208#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001209
Tejun Heocf771cb2008-09-03 09:01:09 +02001210dev_t blk_lookup_devt(const char *name, int partno)
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001211{
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001212 dev_t devt = MKDEV(0, 0);
Tejun Heodef4e382008-09-03 08:57:12 +02001213 struct class_dev_iter iter;
1214 struct device *dev;
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001215
Tejun Heodef4e382008-09-03 08:57:12 +02001216 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1217 while ((dev = class_dev_iter_next(&iter))) {
1218 struct gendisk *disk = dev_to_disk(dev);
Tejun Heo548b10e2008-08-29 09:01:47 +02001219 struct hd_struct *part;
Tejun Heodef4e382008-09-03 08:57:12 +02001220
Kay Sievers3ada8b72009-01-06 10:44:43 -08001221 if (strcmp(dev_name(dev), name))
Tejun Heof331c022008-09-03 09:01:48 +02001222 continue;
Tejun Heof331c022008-09-03 09:01:48 +02001223
Neil Brown41b8c852009-02-18 10:33:59 +01001224 if (partno < disk->minors) {
1225 /* We need to return the right devno, even
1226 * if the partition doesn't exist yet.
1227 */
1228 devt = MKDEV(MAJOR(dev->devt),
1229 MINOR(dev->devt) + partno);
1230 break;
1231 }
Tejun Heo548b10e2008-08-29 09:01:47 +02001232 part = disk_get_part(disk, partno);
Tejun Heo2bbedcb2008-08-29 11:41:51 +02001233 if (part) {
Tejun Heof331c022008-09-03 09:01:48 +02001234 devt = part_devt(part);
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001235 disk_put_part(part);
Tejun Heo548b10e2008-08-29 09:01:47 +02001236 break;
Tejun Heodef4e382008-09-03 08:57:12 +02001237 }
Tejun Heo548b10e2008-08-29 09:01:47 +02001238 disk_put_part(part);
Kay Sievers5c0ef6d2008-08-16 14:30:30 +02001239 }
Tejun Heodef4e382008-09-03 08:57:12 +02001240 class_dev_iter_exit(&iter);
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001241 return devt;
1242}
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001243EXPORT_SYMBOL(blk_lookup_devt);
1244
Linus Torvalds1da177e2005-04-16 15:20:36 -07001245struct gendisk *alloc_disk(int minors)
1246{
Ezequiel Garciac304a512012-11-10 10:39:44 +01001247 return alloc_disk_node(minors, NUMA_NO_NODE);
Christoph Lameter19460892005-06-23 00:08:19 -07001248}
Tejun Heo689d6fa2008-08-25 19:56:16 +09001249EXPORT_SYMBOL(alloc_disk);
Christoph Lameter19460892005-06-23 00:08:19 -07001250
1251struct gendisk *alloc_disk_node(int minors, int node_id)
1252{
1253 struct gendisk *disk;
1254
Christoph Lameter94f60302007-07-17 04:03:29 -07001255 disk = kmalloc_node(sizeof(struct gendisk),
1256 GFP_KERNEL | __GFP_ZERO, node_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001257 if (disk) {
Tejun Heo074a7ac2008-08-25 19:56:14 +09001258 if (!init_part_stats(&disk->part0)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001259 kfree(disk);
1260 return NULL;
1261 }
Cheng Renquanbf91db12008-11-20 08:37:37 +01001262 disk->node_id = node_id;
Tejun Heo540eed52008-08-25 19:56:15 +09001263 if (disk_expand_part_tbl(disk, 0)) {
1264 free_part_stats(&disk->part0);
Tejun Heob5d0b9d2008-09-03 09:06:42 +02001265 kfree(disk);
1266 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001267 }
Tejun Heo540eed52008-08-25 19:56:15 +09001268 disk->part_tbl->part[0] = &disk->part0;
Jens Axboe6c23a962011-01-07 08:43:37 +01001269
Vivek Goyalc83f6bf2012-08-01 12:24:18 +02001270 /*
1271 * set_capacity() and get_capacity() currently don't use
1272 * seqcounter to read/update the part0->nr_sects. Still init
1273 * the counter as we can read the sectors in IO submission
1274 * patch using seqence counters.
1275 *
1276 * TODO: Ideally set_capacity() and get_capacity() should be
1277 * converted to make use of bd_mutex and sequence counters.
1278 */
1279 seqcount_init(&disk->part0.nr_sects_seq);
Jens Axboe6c23a962011-01-07 08:43:37 +01001280 hd_ref_init(&disk->part0);
Tejun Heob5d0b9d2008-09-03 09:06:42 +02001281
Linus Torvalds1da177e2005-04-16 15:20:36 -07001282 disk->minors = minors;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283 rand_initialize_disk(disk);
Tejun Heoed9e1982008-08-25 19:56:05 +09001284 disk_to_dev(disk)->class = &block_class;
1285 disk_to_dev(disk)->type = &disk_type;
1286 device_initialize(disk_to_dev(disk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001287 }
1288 return disk;
1289}
Christoph Lameter19460892005-06-23 00:08:19 -07001290EXPORT_SYMBOL(alloc_disk_node);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001291
1292struct kobject *get_disk(struct gendisk *disk)
1293{
1294 struct module *owner;
1295 struct kobject *kobj;
1296
1297 if (!disk->fops)
1298 return NULL;
1299 owner = disk->fops->owner;
1300 if (owner && !try_module_get(owner))
1301 return NULL;
Tejun Heoed9e1982008-08-25 19:56:05 +09001302 kobj = kobject_get(&disk_to_dev(disk)->kobj);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001303 if (kobj == NULL) {
1304 module_put(owner);
1305 return NULL;
1306 }
1307 return kobj;
1308
1309}
1310
1311EXPORT_SYMBOL(get_disk);
1312
1313void put_disk(struct gendisk *disk)
1314{
1315 if (disk)
Tejun Heoed9e1982008-08-25 19:56:05 +09001316 kobject_put(&disk_to_dev(disk)->kobj);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001317}
1318
1319EXPORT_SYMBOL(put_disk);
1320
Hannes Reineckee3264a42009-07-28 09:13:13 +02001321static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1322{
1323 char event[] = "DISK_RO=1";
1324 char *envp[] = { event, NULL };
1325
1326 if (!ro)
1327 event[8] = '0';
1328 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1329}
1330
Linus Torvalds1da177e2005-04-16 15:20:36 -07001331void set_device_ro(struct block_device *bdev, int flag)
1332{
Tejun Heob7db9952008-08-25 19:56:10 +09001333 bdev->bd_part->policy = flag;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001334}
1335
1336EXPORT_SYMBOL(set_device_ro);
1337
1338void set_disk_ro(struct gendisk *disk, int flag)
1339{
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001340 struct disk_part_iter piter;
1341 struct hd_struct *part;
1342
Hannes Reineckee3264a42009-07-28 09:13:13 +02001343 if (disk->part0.policy != flag) {
1344 set_disk_ro_uevent(disk, flag);
1345 disk->part0.policy = flag;
1346 }
1347
1348 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY);
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001349 while ((part = disk_part_iter_next(&piter)))
1350 part->policy = flag;
1351 disk_part_iter_exit(&piter);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352}
1353
1354EXPORT_SYMBOL(set_disk_ro);
1355
1356int bdev_read_only(struct block_device *bdev)
1357{
1358 if (!bdev)
1359 return 0;
Tejun Heob7db9952008-08-25 19:56:10 +09001360 return bdev->bd_part->policy;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001361}
1362
1363EXPORT_SYMBOL(bdev_read_only);
1364
Tejun Heocf771cb2008-09-03 09:01:09 +02001365int invalidate_partition(struct gendisk *disk, int partno)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001366{
1367 int res = 0;
Tejun Heocf771cb2008-09-03 09:01:09 +02001368 struct block_device *bdev = bdget_disk(disk, partno);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001369 if (bdev) {
Christoph Hellwig2ef41632005-05-05 16:15:59 -07001370 fsync_bdev(bdev);
NeilBrown93b270f2011-02-24 17:25:47 +11001371 res = __invalidate_device(bdev, true);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001372 bdput(bdev);
1373 }
1374 return res;
1375}
1376
1377EXPORT_SYMBOL(invalidate_partition);
Tejun Heo77ea8872010-12-08 20:57:37 +01001378
1379/*
1380 * Disk events - monitor disk events like media change and eject request.
1381 */
1382struct disk_events {
1383 struct list_head node; /* all disk_event's */
1384 struct gendisk *disk; /* the associated disk */
1385 spinlock_t lock;
1386
Tejun Heofdd514e2011-06-09 20:43:59 +02001387 struct mutex block_mutex; /* protects blocking */
Tejun Heo77ea8872010-12-08 20:57:37 +01001388 int block; /* event blocking depth */
1389 unsigned int pending; /* events already sent out */
1390 unsigned int clearing; /* events being cleared */
1391
1392 long poll_msecs; /* interval, -1 for default */
1393 struct delayed_work dwork;
1394};
1395
1396static const char *disk_events_strs[] = {
1397 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "media_change",
1398 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "eject_request",
1399};
1400
1401static char *disk_uevents[] = {
1402 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "DISK_MEDIA_CHANGE=1",
1403 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "DISK_EJECT_REQUEST=1",
1404};
1405
1406/* list of all disk_events */
1407static DEFINE_MUTEX(disk_events_mutex);
1408static LIST_HEAD(disk_events);
1409
1410/* disable in-kernel polling by default */
1411static unsigned long disk_events_dfl_poll_msecs = 0;
1412
1413static unsigned long disk_events_poll_jiffies(struct gendisk *disk)
1414{
1415 struct disk_events *ev = disk->ev;
1416 long intv_msecs = 0;
1417
1418 /*
1419 * If device-specific poll interval is set, always use it. If
1420 * the default is being used, poll iff there are events which
1421 * can't be monitored asynchronously.
1422 */
1423 if (ev->poll_msecs >= 0)
1424 intv_msecs = ev->poll_msecs;
1425 else if (disk->events & ~disk->async_events)
1426 intv_msecs = disk_events_dfl_poll_msecs;
1427
1428 return msecs_to_jiffies(intv_msecs);
1429}
1430
Tejun Heoc3af54a2011-06-09 20:43:55 +02001431/**
1432 * disk_block_events - block and flush disk event checking
1433 * @disk: disk to block events for
1434 *
1435 * On return from this function, it is guaranteed that event checking
1436 * isn't in progress and won't happen until unblocked by
1437 * disk_unblock_events(). Events blocking is counted and the actual
1438 * unblocking happens after the matching number of unblocks are done.
1439 *
1440 * Note that this intentionally does not block event checking from
1441 * disk_clear_events().
1442 *
1443 * CONTEXT:
1444 * Might sleep.
1445 */
1446void disk_block_events(struct gendisk *disk)
Tejun Heo77ea8872010-12-08 20:57:37 +01001447{
1448 struct disk_events *ev = disk->ev;
1449 unsigned long flags;
1450 bool cancel;
1451
Tejun Heoc3af54a2011-06-09 20:43:55 +02001452 if (!ev)
1453 return;
1454
Tejun Heofdd514e2011-06-09 20:43:59 +02001455 /*
1456 * Outer mutex ensures that the first blocker completes canceling
1457 * the event work before further blockers are allowed to finish.
1458 */
1459 mutex_lock(&ev->block_mutex);
1460
Tejun Heo77ea8872010-12-08 20:57:37 +01001461 spin_lock_irqsave(&ev->lock, flags);
1462 cancel = !ev->block++;
1463 spin_unlock_irqrestore(&ev->lock, flags);
1464
Tejun Heoc3af54a2011-06-09 20:43:55 +02001465 if (cancel)
1466 cancel_delayed_work_sync(&disk->ev->dwork);
Tejun Heofdd514e2011-06-09 20:43:59 +02001467
1468 mutex_unlock(&ev->block_mutex);
Tejun Heo77ea8872010-12-08 20:57:37 +01001469}
1470
1471static void __disk_unblock_events(struct gendisk *disk, bool check_now)
1472{
1473 struct disk_events *ev = disk->ev;
1474 unsigned long intv;
1475 unsigned long flags;
1476
1477 spin_lock_irqsave(&ev->lock, flags);
1478
1479 if (WARN_ON_ONCE(ev->block <= 0))
1480 goto out_unlock;
1481
1482 if (--ev->block)
1483 goto out_unlock;
1484
1485 /*
1486 * Not exactly a latency critical operation, set poll timer
1487 * slack to 25% and kick event check.
1488 */
1489 intv = disk_events_poll_jiffies(disk);
1490 set_timer_slack(&ev->dwork.timer, intv / 4);
1491 if (check_now)
Viresh Kumar695588f2013-04-24 17:12:56 +05301492 queue_delayed_work(system_freezable_power_efficient_wq,
1493 &ev->dwork, 0);
Tejun Heo77ea8872010-12-08 20:57:37 +01001494 else if (intv)
Viresh Kumar695588f2013-04-24 17:12:56 +05301495 queue_delayed_work(system_freezable_power_efficient_wq,
1496 &ev->dwork, intv);
Tejun Heo77ea8872010-12-08 20:57:37 +01001497out_unlock:
1498 spin_unlock_irqrestore(&ev->lock, flags);
1499}
1500
1501/**
Tejun Heo77ea8872010-12-08 20:57:37 +01001502 * disk_unblock_events - unblock disk event checking
1503 * @disk: disk to unblock events for
1504 *
1505 * Undo disk_block_events(). When the block count reaches zero, it
1506 * starts events polling if configured.
1507 *
1508 * CONTEXT:
1509 * Don't care. Safe to call from irq context.
1510 */
1511void disk_unblock_events(struct gendisk *disk)
1512{
1513 if (disk->ev)
Tejun Heofacc31d2011-03-09 19:54:27 +01001514 __disk_unblock_events(disk, false);
Tejun Heo77ea8872010-12-08 20:57:37 +01001515}
1516
1517/**
Tejun Heo85ef06d2011-07-01 16:17:47 +02001518 * disk_flush_events - schedule immediate event checking and flushing
1519 * @disk: disk to check and flush events for
1520 * @mask: events to flush
Tejun Heo77ea8872010-12-08 20:57:37 +01001521 *
Tejun Heo85ef06d2011-07-01 16:17:47 +02001522 * Schedule immediate event checking on @disk if not blocked. Events in
1523 * @mask are scheduled to be cleared from the driver. Note that this
1524 * doesn't clear the events from @disk->ev.
Tejun Heo77ea8872010-12-08 20:57:37 +01001525 *
1526 * CONTEXT:
Tejun Heo85ef06d2011-07-01 16:17:47 +02001527 * If @mask is non-zero must be called with bdev->bd_mutex held.
Tejun Heo77ea8872010-12-08 20:57:37 +01001528 */
Tejun Heo85ef06d2011-07-01 16:17:47 +02001529void disk_flush_events(struct gendisk *disk, unsigned int mask)
Tejun Heo77ea8872010-12-08 20:57:37 +01001530{
Tejun Heoa9dce2a2011-06-09 20:43:54 +02001531 struct disk_events *ev = disk->ev;
Tejun Heoa9dce2a2011-06-09 20:43:54 +02001532
1533 if (!ev)
1534 return;
1535
Tejun Heo85ef06d2011-07-01 16:17:47 +02001536 spin_lock_irq(&ev->lock);
1537 ev->clearing |= mask;
Tejun Heo41f63c52012-08-03 10:30:47 -07001538 if (!ev->block)
Viresh Kumar695588f2013-04-24 17:12:56 +05301539 mod_delayed_work(system_freezable_power_efficient_wq,
1540 &ev->dwork, 0);
Tejun Heo85ef06d2011-07-01 16:17:47 +02001541 spin_unlock_irq(&ev->lock);
Tejun Heo77ea8872010-12-08 20:57:37 +01001542}
Tejun Heo77ea8872010-12-08 20:57:37 +01001543
1544/**
1545 * disk_clear_events - synchronously check, clear and return pending events
1546 * @disk: disk to fetch and clear events from
1547 * @mask: mask of events to be fetched and clearted
1548 *
1549 * Disk events are synchronously checked and pending events in @mask
1550 * are cleared and returned. This ignores the block count.
1551 *
1552 * CONTEXT:
1553 * Might sleep.
1554 */
1555unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask)
1556{
1557 const struct block_device_operations *bdops = disk->fops;
1558 struct disk_events *ev = disk->ev;
1559 unsigned int pending;
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001560 unsigned int clearing = mask;
Tejun Heo77ea8872010-12-08 20:57:37 +01001561
1562 if (!ev) {
1563 /* for drivers still using the old ->media_changed method */
1564 if ((mask & DISK_EVENT_MEDIA_CHANGE) &&
1565 bdops->media_changed && bdops->media_changed(disk))
1566 return DISK_EVENT_MEDIA_CHANGE;
1567 return 0;
1568 }
1569
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001570 disk_block_events(disk);
1571
1572 /*
1573 * store the union of mask and ev->clearing on the stack so that the
1574 * race with disk_flush_events does not cause ambiguity (ev->clearing
1575 * can still be modified even if events are blocked).
1576 */
Tejun Heo77ea8872010-12-08 20:57:37 +01001577 spin_lock_irq(&ev->lock);
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001578 clearing |= ev->clearing;
1579 ev->clearing = 0;
Tejun Heo77ea8872010-12-08 20:57:37 +01001580 spin_unlock_irq(&ev->lock);
1581
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001582 disk_check_events(ev, &clearing);
Derek Basehoreaea24a82012-12-18 12:27:18 -08001583 /*
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001584 * if ev->clearing is not 0, the disk_flush_events got called in the
1585 * middle of this function, so we want to run the workfn without delay.
Derek Basehoreaea24a82012-12-18 12:27:18 -08001586 */
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001587 __disk_unblock_events(disk, ev->clearing ? true : false);
Tejun Heo77ea8872010-12-08 20:57:37 +01001588
1589 /* then, fetch and clear pending events */
1590 spin_lock_irq(&ev->lock);
Tejun Heo77ea8872010-12-08 20:57:37 +01001591 pending = ev->pending & mask;
1592 ev->pending &= ~mask;
1593 spin_unlock_irq(&ev->lock);
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001594 WARN_ON_ONCE(clearing & mask);
Tejun Heo77ea8872010-12-08 20:57:37 +01001595
1596 return pending;
1597}
1598
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001599/*
1600 * Separate this part out so that a different pointer for clearing_ptr can be
1601 * passed in for disk_clear_events.
1602 */
Tejun Heo77ea8872010-12-08 20:57:37 +01001603static void disk_events_workfn(struct work_struct *work)
1604{
1605 struct delayed_work *dwork = to_delayed_work(work);
1606 struct disk_events *ev = container_of(dwork, struct disk_events, dwork);
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001607
1608 disk_check_events(ev, &ev->clearing);
1609}
1610
1611static void disk_check_events(struct disk_events *ev,
1612 unsigned int *clearing_ptr)
1613{
Tejun Heo77ea8872010-12-08 20:57:37 +01001614 struct gendisk *disk = ev->disk;
1615 char *envp[ARRAY_SIZE(disk_uevents) + 1] = { };
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001616 unsigned int clearing = *clearing_ptr;
Tejun Heo77ea8872010-12-08 20:57:37 +01001617 unsigned int events;
1618 unsigned long intv;
1619 int nr_events = 0, i;
1620
1621 /* check events */
1622 events = disk->fops->check_events(disk, clearing);
1623
1624 /* accumulate pending events and schedule next poll if necessary */
1625 spin_lock_irq(&ev->lock);
1626
1627 events &= ~ev->pending;
1628 ev->pending |= events;
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001629 *clearing_ptr &= ~clearing;
Tejun Heo77ea8872010-12-08 20:57:37 +01001630
1631 intv = disk_events_poll_jiffies(disk);
1632 if (!ev->block && intv)
Viresh Kumar695588f2013-04-24 17:12:56 +05301633 queue_delayed_work(system_freezable_power_efficient_wq,
1634 &ev->dwork, intv);
Tejun Heo77ea8872010-12-08 20:57:37 +01001635
1636 spin_unlock_irq(&ev->lock);
1637
Tejun Heo7c88a162011-04-21 19:43:58 +02001638 /*
1639 * Tell userland about new events. Only the events listed in
1640 * @disk->events are reported. Unlisted events are processed the
1641 * same internally but never get reported to userland.
1642 */
Tejun Heo77ea8872010-12-08 20:57:37 +01001643 for (i = 0; i < ARRAY_SIZE(disk_uevents); i++)
Tejun Heo7c88a162011-04-21 19:43:58 +02001644 if (events & disk->events & (1 << i))
Tejun Heo77ea8872010-12-08 20:57:37 +01001645 envp[nr_events++] = disk_uevents[i];
1646
1647 if (nr_events)
1648 kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp);
1649}
1650
1651/*
1652 * A disk events enabled device has the following sysfs nodes under
1653 * its /sys/block/X/ directory.
1654 *
1655 * events : list of all supported events
1656 * events_async : list of events which can be detected w/o polling
1657 * events_poll_msecs : polling interval, 0: disable, -1: system default
1658 */
1659static ssize_t __disk_events_show(unsigned int events, char *buf)
1660{
1661 const char *delim = "";
1662 ssize_t pos = 0;
1663 int i;
1664
1665 for (i = 0; i < ARRAY_SIZE(disk_events_strs); i++)
1666 if (events & (1 << i)) {
1667 pos += sprintf(buf + pos, "%s%s",
1668 delim, disk_events_strs[i]);
1669 delim = " ";
1670 }
1671 if (pos)
1672 pos += sprintf(buf + pos, "\n");
1673 return pos;
1674}
1675
1676static ssize_t disk_events_show(struct device *dev,
1677 struct device_attribute *attr, char *buf)
1678{
1679 struct gendisk *disk = dev_to_disk(dev);
1680
1681 return __disk_events_show(disk->events, buf);
1682}
1683
1684static ssize_t disk_events_async_show(struct device *dev,
1685 struct device_attribute *attr, char *buf)
1686{
1687 struct gendisk *disk = dev_to_disk(dev);
1688
1689 return __disk_events_show(disk->async_events, buf);
1690}
1691
1692static ssize_t disk_events_poll_msecs_show(struct device *dev,
1693 struct device_attribute *attr,
1694 char *buf)
1695{
1696 struct gendisk *disk = dev_to_disk(dev);
1697
1698 return sprintf(buf, "%ld\n", disk->ev->poll_msecs);
1699}
1700
1701static ssize_t disk_events_poll_msecs_store(struct device *dev,
1702 struct device_attribute *attr,
1703 const char *buf, size_t count)
1704{
1705 struct gendisk *disk = dev_to_disk(dev);
1706 long intv;
1707
1708 if (!count || !sscanf(buf, "%ld", &intv))
1709 return -EINVAL;
1710
1711 if (intv < 0 && intv != -1)
1712 return -EINVAL;
1713
Tejun Heoc3af54a2011-06-09 20:43:55 +02001714 disk_block_events(disk);
Tejun Heo77ea8872010-12-08 20:57:37 +01001715 disk->ev->poll_msecs = intv;
1716 __disk_unblock_events(disk, true);
1717
1718 return count;
1719}
1720
1721static const DEVICE_ATTR(events, S_IRUGO, disk_events_show, NULL);
1722static const DEVICE_ATTR(events_async, S_IRUGO, disk_events_async_show, NULL);
1723static const DEVICE_ATTR(events_poll_msecs, S_IRUGO|S_IWUSR,
1724 disk_events_poll_msecs_show,
1725 disk_events_poll_msecs_store);
1726
1727static const struct attribute *disk_events_attrs[] = {
1728 &dev_attr_events.attr,
1729 &dev_attr_events_async.attr,
1730 &dev_attr_events_poll_msecs.attr,
1731 NULL,
1732};
1733
1734/*
1735 * The default polling interval can be specified by the kernel
1736 * parameter block.events_dfl_poll_msecs which defaults to 0
1737 * (disable). This can also be modified runtime by writing to
1738 * /sys/module/block/events_dfl_poll_msecs.
1739 */
1740static int disk_events_set_dfl_poll_msecs(const char *val,
1741 const struct kernel_param *kp)
1742{
1743 struct disk_events *ev;
1744 int ret;
1745
1746 ret = param_set_ulong(val, kp);
1747 if (ret < 0)
1748 return ret;
1749
1750 mutex_lock(&disk_events_mutex);
1751
1752 list_for_each_entry(ev, &disk_events, node)
Tejun Heo85ef06d2011-07-01 16:17:47 +02001753 disk_flush_events(ev->disk, 0);
Tejun Heo77ea8872010-12-08 20:57:37 +01001754
1755 mutex_unlock(&disk_events_mutex);
1756
1757 return 0;
1758}
1759
1760static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops = {
1761 .set = disk_events_set_dfl_poll_msecs,
1762 .get = param_get_ulong,
1763};
1764
1765#undef MODULE_PARAM_PREFIX
1766#define MODULE_PARAM_PREFIX "block."
1767
1768module_param_cb(events_dfl_poll_msecs, &disk_events_dfl_poll_msecs_param_ops,
1769 &disk_events_dfl_poll_msecs, 0644);
1770
1771/*
Stanislaw Gruszka9f53d2fe2012-03-02 10:43:28 +01001772 * disk_{alloc|add|del|release}_events - initialize and destroy disk_events.
Tejun Heo77ea8872010-12-08 20:57:37 +01001773 */
Stanislaw Gruszka9f53d2fe2012-03-02 10:43:28 +01001774static void disk_alloc_events(struct gendisk *disk)
Tejun Heo77ea8872010-12-08 20:57:37 +01001775{
1776 struct disk_events *ev;
1777
Tejun Heo75e3f3e2011-05-26 21:06:50 +02001778 if (!disk->fops->check_events)
Tejun Heo77ea8872010-12-08 20:57:37 +01001779 return;
1780
1781 ev = kzalloc(sizeof(*ev), GFP_KERNEL);
1782 if (!ev) {
1783 pr_warn("%s: failed to initialize events\n", disk->disk_name);
1784 return;
1785 }
1786
Tejun Heo77ea8872010-12-08 20:57:37 +01001787 INIT_LIST_HEAD(&ev->node);
1788 ev->disk = disk;
1789 spin_lock_init(&ev->lock);
Tejun Heofdd514e2011-06-09 20:43:59 +02001790 mutex_init(&ev->block_mutex);
Tejun Heo77ea8872010-12-08 20:57:37 +01001791 ev->block = 1;
1792 ev->poll_msecs = -1;
1793 INIT_DELAYED_WORK(&ev->dwork, disk_events_workfn);
1794
Stanislaw Gruszka9f53d2fe2012-03-02 10:43:28 +01001795 disk->ev = ev;
1796}
1797
1798static void disk_add_events(struct gendisk *disk)
1799{
1800 if (!disk->ev)
1801 return;
1802
1803 /* FIXME: error handling */
1804 if (sysfs_create_files(&disk_to_dev(disk)->kobj, disk_events_attrs) < 0)
1805 pr_warn("%s: failed to create sysfs files for events\n",
1806 disk->disk_name);
1807
Tejun Heo77ea8872010-12-08 20:57:37 +01001808 mutex_lock(&disk_events_mutex);
Stanislaw Gruszka9f53d2fe2012-03-02 10:43:28 +01001809 list_add_tail(&disk->ev->node, &disk_events);
Tejun Heo77ea8872010-12-08 20:57:37 +01001810 mutex_unlock(&disk_events_mutex);
1811
1812 /*
1813 * Block count is initialized to 1 and the following initial
1814 * unblock kicks it into action.
1815 */
1816 __disk_unblock_events(disk, true);
1817}
1818
1819static void disk_del_events(struct gendisk *disk)
1820{
1821 if (!disk->ev)
1822 return;
1823
Tejun Heoc3af54a2011-06-09 20:43:55 +02001824 disk_block_events(disk);
Tejun Heo77ea8872010-12-08 20:57:37 +01001825
1826 mutex_lock(&disk_events_mutex);
1827 list_del_init(&disk->ev->node);
1828 mutex_unlock(&disk_events_mutex);
1829
1830 sysfs_remove_files(&disk_to_dev(disk)->kobj, disk_events_attrs);
1831}
1832
1833static void disk_release_events(struct gendisk *disk)
1834{
1835 /* the block count should be 1 from disk_del_events() */
1836 WARN_ON_ONCE(disk->ev && disk->ev->block != 1);
1837 kfree(disk->ev);
1838}