blob: e6723bd4d7a16963da3367a1d4791c6402e78a82 [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
Keith Busch2da78092014-08-26 09:05:36 -060031/* For extended devt allocation. ext_devt_lock prevents look up
Tejun Heobcce3de2008-08-25 19:47:22 +090032 * results from going away underneath its user.
33 */
Keith Busch2da78092014-08-26 09:05:36 -060034static DEFINE_SPINLOCK(ext_devt_lock);
Tejun Heobcce3de2008-08-25 19:47:22 +090035static 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 */
Keith Busch2da78092014-08-26 09:05:36 -0600423 idr_preload(GFP_KERNEL);
424
425 spin_lock(&ext_devt_lock);
426 idx = idr_alloc(&ext_devt_idr, part, 0, NR_EXT_DEVT, GFP_NOWAIT);
427 spin_unlock(&ext_devt_lock);
428
429 idr_preload_end();
Tejun Heobab998d2013-02-27 17:03:57 -0800430 if (idx < 0)
431 return idx == -ENOSPC ? -EBUSY : idx;
Tejun Heobcce3de2008-08-25 19:47:22 +0900432
Tejun Heo870d6652008-08-25 19:47:25 +0900433 *devt = MKDEV(BLOCK_EXT_MAJOR, blk_mangle_minor(idx));
Tejun Heobcce3de2008-08-25 19:47:22 +0900434 return 0;
435}
436
437/**
438 * blk_free_devt - free a dev_t
439 * @devt: dev_t to free
440 *
441 * Free @devt which was allocated using blk_alloc_devt().
442 *
443 * CONTEXT:
444 * Might sleep.
445 */
446void blk_free_devt(dev_t devt)
447{
Tejun Heobcce3de2008-08-25 19:47:22 +0900448 if (devt == MKDEV(0, 0))
449 return;
450
451 if (MAJOR(devt) == BLOCK_EXT_MAJOR) {
Keith Busch2da78092014-08-26 09:05:36 -0600452 spin_lock(&ext_devt_lock);
Tejun Heo870d6652008-08-25 19:47:25 +0900453 idr_remove(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
Keith Busch2da78092014-08-26 09:05:36 -0600454 spin_unlock(&ext_devt_lock);
Tejun Heobcce3de2008-08-25 19:47:22 +0900455 }
456}
457
Tejun Heo1f014292008-08-25 19:47:23 +0900458static char *bdevt_str(dev_t devt, char *buf)
459{
460 if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
461 char tbuf[BDEVT_SIZE];
462 snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
463 snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
464 } else
465 snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
466
467 return buf;
468}
469
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470/*
471 * Register device numbers dev..(dev+range-1)
472 * range must be nonzero
473 * The hash chain is sorted on range, so that subranges can override.
474 */
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200475void blk_register_region(dev_t devt, unsigned long range, struct module *module,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476 struct kobject *(*probe)(dev_t, int *, void *),
477 int (*lock)(dev_t, void *), void *data)
478{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200479 kobj_map(bdev_map, devt, range, module, probe, lock, data);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480}
481
482EXPORT_SYMBOL(blk_register_region);
483
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200484void blk_unregister_region(dev_t devt, unsigned long range)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200486 kobj_unmap(bdev_map, devt, range);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487}
488
489EXPORT_SYMBOL(blk_unregister_region);
490
Tejun Heocf771cb2008-09-03 09:01:09 +0200491static struct kobject *exact_match(dev_t devt, int *partno, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492{
493 struct gendisk *p = data;
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200494
Tejun Heoed9e1982008-08-25 19:56:05 +0900495 return &disk_to_dev(p)->kobj;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496}
497
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200498static int exact_lock(dev_t devt, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499{
500 struct gendisk *p = data;
501
502 if (!get_disk(p))
503 return -1;
504 return 0;
505}
506
Al Viro4752bc32011-09-16 00:21:54 -0400507static void register_disk(struct gendisk *disk)
Tejun Heod2bf1b62010-12-08 20:57:36 +0100508{
509 struct device *ddev = disk_to_dev(disk);
510 struct block_device *bdev;
511 struct disk_part_iter piter;
512 struct hd_struct *part;
513 int err;
514
515 ddev->parent = disk->driverfs_dev;
516
Kees Cookffc8b302013-07-03 15:01:14 -0700517 dev_set_name(ddev, "%s", disk->disk_name);
Tejun Heod2bf1b62010-12-08 20:57:36 +0100518
519 /* delay uevents, until we scanned partition table */
520 dev_set_uevent_suppress(ddev, 1);
521
522 if (device_add(ddev))
523 return;
524 if (!sysfs_deprecated) {
525 err = sysfs_create_link(block_depr, &ddev->kobj,
526 kobject_name(&ddev->kobj));
527 if (err) {
528 device_del(ddev);
529 return;
530 }
531 }
Ming Lei25e823c2013-02-22 16:34:13 -0800532
533 /*
534 * avoid probable deadlock caused by allocating memory with
535 * GFP_KERNEL in runtime_resume callback of its all ancestor
536 * devices
537 */
538 pm_runtime_set_memalloc_noio(ddev, true);
539
Tejun Heod2bf1b62010-12-08 20:57:36 +0100540 disk->part0.holder_dir = kobject_create_and_add("holders", &ddev->kobj);
541 disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
542
543 /* No minors to use for partitions */
Tejun Heod27769e2011-08-23 20:01:04 +0200544 if (!disk_part_scan_enabled(disk))
Tejun Heod2bf1b62010-12-08 20:57:36 +0100545 goto exit;
546
547 /* No such device (e.g., media were just removed) */
548 if (!get_capacity(disk))
549 goto exit;
550
551 bdev = bdget_disk(disk, 0);
552 if (!bdev)
553 goto exit;
554
555 bdev->bd_invalidated = 1;
556 err = blkdev_get(bdev, FMODE_READ, NULL);
557 if (err < 0)
558 goto exit;
559 blkdev_put(bdev, FMODE_READ);
560
561exit:
562 /* announce disk after possible partitions are created */
563 dev_set_uevent_suppress(ddev, 0);
564 kobject_uevent(&ddev->kobj, KOBJ_ADD);
565
566 /* announce possible partitions */
567 disk_part_iter_init(&piter, disk, 0);
568 while ((part = disk_part_iter_next(&piter)))
569 kobject_uevent(&part_to_dev(part)->kobj, KOBJ_ADD);
570 disk_part_iter_exit(&piter);
571}
572
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573/**
574 * add_disk - add partitioning information to kernel list
575 * @disk: per-device partitioning information
576 *
577 * This function registers the partitioning information in @disk
578 * with the kernel.
Tejun Heo3e1a7ff2008-08-25 19:56:17 +0900579 *
580 * FIXME: error handling
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581 */
582void add_disk(struct gendisk *disk)
583{
Peter Zijlstracf0ca9f2008-04-30 00:54:32 -0700584 struct backing_dev_info *bdi;
Tejun Heo3e1a7ff2008-08-25 19:56:17 +0900585 dev_t devt;
Greg Kroah-Hartman6ffeea72008-05-22 17:21:08 -0400586 int retval;
Peter Zijlstracf0ca9f2008-04-30 00:54:32 -0700587
Tejun Heo3e1a7ff2008-08-25 19:56:17 +0900588 /* minors == 0 indicates to use ext devt from part0 and should
589 * be accompanied with EXT_DEVT flag. Make sure all
590 * parameters make sense.
591 */
592 WARN_ON(disk->minors && !(disk->major || disk->first_minor));
593 WARN_ON(!disk->minors && !(disk->flags & GENHD_FL_EXT_DEVT));
594
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595 disk->flags |= GENHD_FL_UP;
Tejun Heo3e1a7ff2008-08-25 19:56:17 +0900596
597 retval = blk_alloc_devt(&disk->part0, &devt);
598 if (retval) {
599 WARN_ON(1);
600 return;
601 }
602 disk_to_dev(disk)->devt = devt;
603
604 /* ->major and ->first_minor aren't supposed to be
605 * dereferenced from here on, but set them just in case.
606 */
607 disk->major = MAJOR(devt);
608 disk->first_minor = MINOR(devt);
609
Stanislaw Gruszka9f53d2fe2012-03-02 10:43:28 +0100610 disk_alloc_events(disk);
611
Wanlong Gao9f5e4862011-06-13 10:45:43 +0200612 /* Register BDI before referencing it from bdev */
Signed-off-by: Jan Kara01ea5062010-09-16 20:36:36 +0200613 bdi = &disk->queue->backing_dev_info;
614 bdi_register_dev(bdi, disk_devt(disk));
615
Tejun Heof331c022008-09-03 09:01:48 +0200616 blk_register_region(disk_devt(disk), disk->minors, NULL,
617 exact_match, exact_lock, disk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618 register_disk(disk);
619 blk_register_queue(disk);
Peter Zijlstracf0ca9f2008-04-30 00:54:32 -0700620
Tejun Heo523e1d32011-10-19 14:31:07 +0200621 /*
622 * Take an extra ref on queue which will be put on disk_release()
623 * so that it sticks around as long as @disk is there.
624 */
Tejun Heo09ac46c2011-12-14 00:33:38 +0100625 WARN_ON_ONCE(!blk_get_queue(disk->queue));
Tejun Heo523e1d32011-10-19 14:31:07 +0200626
Tejun Heoed9e1982008-08-25 19:56:05 +0900627 retval = sysfs_create_link(&disk_to_dev(disk)->kobj, &bdi->dev->kobj,
628 "bdi");
Greg Kroah-Hartman6ffeea72008-05-22 17:21:08 -0400629 WARN_ON(retval);
Tejun Heo77ea8872010-12-08 20:57:37 +0100630
631 disk_add_events(disk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633EXPORT_SYMBOL(add_disk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634
Tejun Heod2bf1b62010-12-08 20:57:36 +0100635void del_gendisk(struct gendisk *disk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636{
Tejun Heod2bf1b62010-12-08 20:57:36 +0100637 struct disk_part_iter piter;
638 struct hd_struct *part;
639
Tejun Heo77ea8872010-12-08 20:57:37 +0100640 disk_del_events(disk);
641
Tejun Heod2bf1b62010-12-08 20:57:36 +0100642 /* invalidate stuff */
643 disk_part_iter_init(&piter, disk,
644 DISK_PITER_INCL_EMPTY | DISK_PITER_REVERSE);
645 while ((part = disk_part_iter_next(&piter))) {
646 invalidate_partition(disk, part->partno);
647 delete_partition(disk, part->partno);
648 }
649 disk_part_iter_exit(&piter);
650
651 invalidate_partition(disk, 0);
Tejun Heod2bf1b62010-12-08 20:57:36 +0100652 set_capacity(disk, 0);
653 disk->flags &= ~GENHD_FL_UP;
654
Tejun Heoed9e1982008-08-25 19:56:05 +0900655 sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
Peter Zijlstracf0ca9f2008-04-30 00:54:32 -0700656 bdi_unregister(&disk->queue->backing_dev_info);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657 blk_unregister_queue(disk);
Tejun Heof331c022008-09-03 09:01:48 +0200658 blk_unregister_region(disk_devt(disk), disk->minors);
Tejun Heod2bf1b62010-12-08 20:57:36 +0100659
660 part_stat_set_all(&disk->part0, 0);
661 disk->part0.stamp = 0;
662
663 kobject_put(disk->part0.holder_dir);
664 kobject_put(disk->slave_dir);
665 disk->driverfs_dev = NULL;
666 if (!sysfs_deprecated)
667 sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
Ming Lei25e823c2013-02-22 16:34:13 -0800668 pm_runtime_set_memalloc_noio(disk_to_dev(disk), false);
Tejun Heod2bf1b62010-12-08 20:57:36 +0100669 device_del(disk_to_dev(disk));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670}
Tejun Heod2bf1b62010-12-08 20:57:36 +0100671EXPORT_SYMBOL(del_gendisk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700672
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673/**
674 * get_gendisk - get partitioning information for a given device
Randy Dunlap710027a2008-08-19 20:13:11 +0200675 * @devt: device to get partitioning information for
Randy Dunlap496aa8a2008-10-16 07:46:23 +0200676 * @partno: returned partition index
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677 *
678 * This function gets the structure containing partitioning
Randy Dunlap710027a2008-08-19 20:13:11 +0200679 * information for the given device @devt.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680 */
Tejun Heocf771cb2008-09-03 09:01:09 +0200681struct gendisk *get_gendisk(dev_t devt, int *partno)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682{
Tejun Heobcce3de2008-08-25 19:47:22 +0900683 struct gendisk *disk = NULL;
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200684
Tejun Heobcce3de2008-08-25 19:47:22 +0900685 if (MAJOR(devt) != BLOCK_EXT_MAJOR) {
686 struct kobject *kobj;
687
688 kobj = kobj_lookup(bdev_map, devt, partno);
689 if (kobj)
690 disk = dev_to_disk(kobj_to_dev(kobj));
691 } else {
692 struct hd_struct *part;
693
Keith Busch2da78092014-08-26 09:05:36 -0600694 spin_lock(&ext_devt_lock);
Tejun Heo870d6652008-08-25 19:47:25 +0900695 part = idr_find(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
Tejun Heobcce3de2008-08-25 19:47:22 +0900696 if (part && get_disk(part_to_disk(part))) {
697 *partno = part->partno;
698 disk = part_to_disk(part);
699 }
Keith Busch2da78092014-08-26 09:05:36 -0600700 spin_unlock(&ext_devt_lock);
Tejun Heobcce3de2008-08-25 19:47:22 +0900701 }
702
703 return disk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700704}
Divyesh Shahb6ac23af2010-04-15 08:54:59 +0200705EXPORT_SYMBOL(get_gendisk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706
Tejun Heof331c022008-09-03 09:01:48 +0200707/**
708 * bdget_disk - do bdget() by gendisk and partition number
709 * @disk: gendisk of interest
710 * @partno: partition number
711 *
712 * Find partition @partno from @disk, do bdget() on it.
713 *
714 * CONTEXT:
715 * Don't care.
716 *
717 * RETURNS:
718 * Resulting block_device on success, NULL on failure.
719 */
Harvey Harrisonaeb3d3a2008-08-28 09:27:42 +0200720struct block_device *bdget_disk(struct gendisk *disk, int partno)
Tejun Heof331c022008-09-03 09:01:48 +0200721{
Tejun Heo548b10e2008-08-29 09:01:47 +0200722 struct hd_struct *part;
723 struct block_device *bdev = NULL;
Tejun Heof331c022008-09-03 09:01:48 +0200724
Tejun Heo548b10e2008-08-29 09:01:47 +0200725 part = disk_get_part(disk, partno);
Tejun Heo2bbedcb2008-08-29 11:41:51 +0200726 if (part)
Tejun Heo548b10e2008-08-29 09:01:47 +0200727 bdev = bdget(part_devt(part));
728 disk_put_part(part);
Tejun Heof331c022008-09-03 09:01:48 +0200729
Tejun Heo548b10e2008-08-29 09:01:47 +0200730 return bdev;
Tejun Heof331c022008-09-03 09:01:48 +0200731}
732EXPORT_SYMBOL(bdget_disk);
733
Dave Gilbertdd2a3452007-05-09 02:33:24 -0700734/*
735 * print a full list of all partitions - intended for places where the root
736 * filesystem can't be mounted and thus to give the victim some idea of what
737 * went wrong
738 */
739void __init printk_all_partitions(void)
740{
Tejun Heodef4e382008-09-03 08:57:12 +0200741 struct class_dev_iter iter;
742 struct device *dev;
743
744 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
745 while ((dev = class_dev_iter_next(&iter))) {
746 struct gendisk *disk = dev_to_disk(dev);
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200747 struct disk_part_iter piter;
748 struct hd_struct *part;
Tejun Heo1f014292008-08-25 19:47:23 +0900749 char name_buf[BDEVNAME_SIZE];
750 char devt_buf[BDEVT_SIZE];
Tejun Heodef4e382008-09-03 08:57:12 +0200751
752 /*
753 * Don't show empty devices or things that have been
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300754 * suppressed
Tejun Heodef4e382008-09-03 08:57:12 +0200755 */
756 if (get_capacity(disk) == 0 ||
757 (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
758 continue;
759
760 /*
761 * Note, unlike /proc/partitions, I am showing the
762 * numbers in hex - the same format as the root=
763 * option takes.
764 */
Tejun Heo074a7ac2008-08-25 19:56:14 +0900765 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0);
766 while ((part = disk_part_iter_next(&piter))) {
767 bool is_part0 = part == &disk->part0;
Tejun Heodef4e382008-09-03 08:57:12 +0200768
Will Drewryb5af9212010-08-31 15:47:07 -0500769 printk("%s%s %10llu %s %s", is_part0 ? "" : " ",
Tejun Heo1f014292008-08-25 19:47:23 +0900770 bdevt_str(part_devt(part), devt_buf),
Vivek Goyalc83f6bf2012-08-01 12:24:18 +0200771 (unsigned long long)part_nr_sects_read(part) >> 1
772 , disk_name(disk, part->partno, name_buf),
Stephen Warren1ad7e892012-11-08 16:12:25 -0800773 part->info ? part->info->uuid : "");
Tejun Heo074a7ac2008-08-25 19:56:14 +0900774 if (is_part0) {
775 if (disk->driverfs_dev != NULL &&
776 disk->driverfs_dev->driver != NULL)
777 printk(" driver: %s\n",
778 disk->driverfs_dev->driver->name);
779 else
780 printk(" (driver?)\n");
781 } else
782 printk("\n");
783 }
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200784 disk_part_iter_exit(&piter);
Tejun Heodef4e382008-09-03 08:57:12 +0200785 }
786 class_dev_iter_exit(&iter);
Dave Gilbertdd2a3452007-05-09 02:33:24 -0700787}
788
Linus Torvalds1da177e2005-04-16 15:20:36 -0700789#ifdef CONFIG_PROC_FS
790/* iterator */
Tejun Heodef4e382008-09-03 08:57:12 +0200791static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
Greg Kroah-Hartman68c4d4a2008-05-22 17:21:08 -0400792{
Tejun Heodef4e382008-09-03 08:57:12 +0200793 loff_t skip = *pos;
794 struct class_dev_iter *iter;
795 struct device *dev;
Greg Kroah-Hartman68c4d4a2008-05-22 17:21:08 -0400796
Harvey Harrisonaeb3d3a2008-08-28 09:27:42 +0200797 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
Tejun Heodef4e382008-09-03 08:57:12 +0200798 if (!iter)
799 return ERR_PTR(-ENOMEM);
800
801 seqf->private = iter;
802 class_dev_iter_init(iter, &block_class, NULL, &disk_type);
803 do {
804 dev = class_dev_iter_next(iter);
805 if (!dev)
806 return NULL;
807 } while (skip--);
808
809 return dev_to_disk(dev);
Greg Kroah-Hartman68c4d4a2008-05-22 17:21:08 -0400810}
811
Tejun Heodef4e382008-09-03 08:57:12 +0200812static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200814 struct device *dev;
Greg Kroah-Hartman66c64af2008-05-22 17:21:08 -0400815
Tejun Heodef4e382008-09-03 08:57:12 +0200816 (*pos)++;
817 dev = class_dev_iter_next(seqf->private);
Tejun Heo2ac3cee2008-09-03 08:53:37 +0200818 if (dev)
Greg Kroah-Hartman68c4d4a2008-05-22 17:21:08 -0400819 return dev_to_disk(dev);
Tejun Heo2ac3cee2008-09-03 08:53:37 +0200820
Linus Torvalds1da177e2005-04-16 15:20:36 -0700821 return NULL;
822}
823
Tejun Heodef4e382008-09-03 08:57:12 +0200824static void disk_seqf_stop(struct seq_file *seqf, void *v)
Greg Kroah-Hartman27f30252008-05-22 17:21:08 -0400825{
Tejun Heodef4e382008-09-03 08:57:12 +0200826 struct class_dev_iter *iter = seqf->private;
Greg Kroah-Hartman27f30252008-05-22 17:21:08 -0400827
Tejun Heodef4e382008-09-03 08:57:12 +0200828 /* stop is called even after start failed :-( */
829 if (iter) {
830 class_dev_iter_exit(iter);
831 kfree(iter);
Kay Sievers5c0ef6d2008-08-16 14:30:30 +0200832 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833}
834
Tejun Heodef4e382008-09-03 08:57:12 +0200835static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836{
Jianpeng Ma06768062012-08-03 10:42:00 +0200837 void *p;
Tejun Heodef4e382008-09-03 08:57:12 +0200838
839 p = disk_seqf_start(seqf, pos);
Yang Zhangb9f985b2010-12-17 08:58:36 +0100840 if (!IS_ERR_OR_NULL(p) && !*pos)
Tejun Heodef4e382008-09-03 08:57:12 +0200841 seq_puts(seqf, "major minor #blocks name\n\n");
842 return p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843}
844
Tejun Heocf771cb2008-09-03 09:01:09 +0200845static int show_partition(struct seq_file *seqf, void *v)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700846{
847 struct gendisk *sgp = v;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200848 struct disk_part_iter piter;
849 struct hd_struct *part;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850 char buf[BDEVNAME_SIZE];
851
Linus Torvalds1da177e2005-04-16 15:20:36 -0700852 /* Don't show non-partitionable removeable devices or empty devices */
Tejun Heod27769e2011-08-23 20:01:04 +0200853 if (!get_capacity(sgp) || (!disk_max_parts(sgp) &&
Tejun Heof331c022008-09-03 09:01:48 +0200854 (sgp->flags & GENHD_FL_REMOVABLE)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700855 return 0;
856 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
857 return 0;
858
859 /* show the full disk and all non-0 size partitions of it */
Tejun Heo074a7ac2008-08-25 19:56:14 +0900860 disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0);
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200861 while ((part = disk_part_iter_next(&piter)))
Tejun Heo1f014292008-08-25 19:47:23 +0900862 seq_printf(seqf, "%4d %7d %10llu %s\n",
Tejun Heof331c022008-09-03 09:01:48 +0200863 MAJOR(part_devt(part)), MINOR(part_devt(part)),
Vivek Goyalc83f6bf2012-08-01 12:24:18 +0200864 (unsigned long long)part_nr_sects_read(part) >> 1,
Tejun Heof331c022008-09-03 09:01:48 +0200865 disk_name(sgp, part->partno, buf));
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200866 disk_part_iter_exit(&piter);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700867
868 return 0;
869}
870
Alexey Dobriyanf5009752008-10-04 23:53:21 +0400871static const struct seq_operations partitions_op = {
Tejun Heodef4e382008-09-03 08:57:12 +0200872 .start = show_partition_start,
873 .next = disk_seqf_next,
874 .stop = disk_seqf_stop,
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200875 .show = show_partition
Linus Torvalds1da177e2005-04-16 15:20:36 -0700876};
Alexey Dobriyanf5009752008-10-04 23:53:21 +0400877
878static int partitions_open(struct inode *inode, struct file *file)
879{
880 return seq_open(file, &partitions_op);
881}
882
883static const struct file_operations proc_partitions_operations = {
884 .open = partitions_open,
885 .read = seq_read,
886 .llseek = seq_lseek,
887 .release = seq_release,
888};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889#endif
890
891
Tejun Heocf771cb2008-09-03 09:01:09 +0200892static struct kobject *base_probe(dev_t devt, int *partno, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200894 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895 /* Make old-style 2.4 aliases work */
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200896 request_module("block-major-%d", MAJOR(devt));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700897 return NULL;
898}
899
900static int __init genhd_device_init(void)
901{
Dan Williamse105b8b2008-04-21 10:51:07 -0700902 int error;
903
904 block_class.dev_kobj = sysfs_dev_block_kobj;
905 error = class_register(&block_class);
Roland McGrathee27a552008-03-11 17:13:15 -0700906 if (unlikely(error))
907 return error;
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200908 bdev_map = kobj_map_init(base_probe, &block_class_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700909 blk_dev_init();
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200910
Zhang, Yanmin561ec682008-11-14 08:26:30 +0100911 register_blkdev(BLOCK_EXT_MAJOR, "blkext");
912
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200913 /* create top-level block dir */
Andi Kleene52eec12010-09-08 16:54:17 +0200914 if (!sysfs_deprecated)
915 block_depr = kobject_create_and_add("block", NULL);
Greg Kroah-Hartman830d3cf2007-11-06 10:36:58 -0800916 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700917}
918
919subsys_initcall(genhd_device_init);
920
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200921static ssize_t disk_range_show(struct device *dev,
922 struct device_attribute *attr, char *buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700923{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200924 struct gendisk *disk = dev_to_disk(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700925
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200926 return sprintf(buf, "%d\n", disk->minors);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927}
928
Tejun Heo1f014292008-08-25 19:47:23 +0900929static ssize_t disk_ext_range_show(struct device *dev,
930 struct device_attribute *attr, char *buf)
931{
932 struct gendisk *disk = dev_to_disk(dev);
933
Tejun Heob5d0b9d2008-09-03 09:06:42 +0200934 return sprintf(buf, "%d\n", disk_max_parts(disk));
Tejun Heo1f014292008-08-25 19:47:23 +0900935}
936
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200937static ssize_t disk_removable_show(struct device *dev,
938 struct device_attribute *attr, char *buf)
Kay Sieversa7fd6702005-10-01 14:49:43 +0200939{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200940 struct gendisk *disk = dev_to_disk(dev);
Kay Sieversa7fd6702005-10-01 14:49:43 +0200941
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200942 return sprintf(buf, "%d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700943 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200944}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700945
Kay Sievers1c9ce522008-06-13 09:41:00 +0200946static ssize_t disk_ro_show(struct device *dev,
947 struct device_attribute *attr, char *buf)
948{
949 struct gendisk *disk = dev_to_disk(dev);
950
Tejun Heob7db9952008-08-25 19:56:10 +0900951 return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
Kay Sievers1c9ce522008-06-13 09:41:00 +0200952}
953
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200954static ssize_t disk_capability_show(struct device *dev,
955 struct device_attribute *attr, char *buf)
Kristen Carlson Accardi86ce18d2007-05-23 13:57:38 -0700956{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200957 struct gendisk *disk = dev_to_disk(dev);
958
959 return sprintf(buf, "%x\n", disk->flags);
Kristen Carlson Accardi86ce18d2007-05-23 13:57:38 -0700960}
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200961
Martin K. Petersenc72758f2009-05-22 17:17:53 -0400962static ssize_t disk_alignment_offset_show(struct device *dev,
963 struct device_attribute *attr,
964 char *buf)
965{
966 struct gendisk *disk = dev_to_disk(dev);
967
968 return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue));
969}
970
Martin K. Petersen86b37282009-11-10 11:50:21 +0100971static ssize_t disk_discard_alignment_show(struct device *dev,
972 struct device_attribute *attr,
973 char *buf)
974{
975 struct gendisk *disk = dev_to_disk(dev);
976
Martin K. Petersendd3d1452010-01-11 03:21:48 -0500977 return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue));
Martin K. Petersen86b37282009-11-10 11:50:21 +0100978}
979
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200980static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL);
Tejun Heo1f014292008-08-25 19:47:23 +0900981static DEVICE_ATTR(ext_range, S_IRUGO, disk_ext_range_show, NULL);
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200982static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL);
Kay Sievers1c9ce522008-06-13 09:41:00 +0200983static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL);
Tejun Heoe5610522008-08-25 19:56:09 +0900984static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL);
Martin K. Petersenc72758f2009-05-22 17:17:53 -0400985static DEVICE_ATTR(alignment_offset, S_IRUGO, disk_alignment_offset_show, NULL);
Martin K. Petersen86b37282009-11-10 11:50:21 +0100986static DEVICE_ATTR(discard_alignment, S_IRUGO, disk_discard_alignment_show,
987 NULL);
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200988static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL);
Tejun Heo074a7ac2008-08-25 19:56:14 +0900989static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL);
Nikanth Karthikesan316d3152009-10-06 20:16:55 +0200990static DEVICE_ATTR(inflight, S_IRUGO, part_inflight_show, NULL);
Akinobu Mitac17bb492006-12-08 02:39:46 -0800991#ifdef CONFIG_FAIL_MAKE_REQUEST
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200992static struct device_attribute dev_attr_fail =
Tejun Heoeddb2e22008-08-25 19:56:13 +0900993 __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store);
Akinobu Mitac17bb492006-12-08 02:39:46 -0800994#endif
Jens Axboe581d4e22008-09-14 05:56:33 -0700995#ifdef CONFIG_FAIL_IO_TIMEOUT
996static struct device_attribute dev_attr_fail_timeout =
997 __ATTR(io-timeout-fail, S_IRUGO|S_IWUSR, part_timeout_show,
998 part_timeout_store);
999#endif
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001000
1001static struct attribute *disk_attrs[] = {
1002 &dev_attr_range.attr,
Tejun Heo1f014292008-08-25 19:47:23 +09001003 &dev_attr_ext_range.attr,
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001004 &dev_attr_removable.attr,
Kay Sievers1c9ce522008-06-13 09:41:00 +02001005 &dev_attr_ro.attr,
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001006 &dev_attr_size.attr,
Martin K. Petersenc72758f2009-05-22 17:17:53 -04001007 &dev_attr_alignment_offset.attr,
Martin K. Petersen86b37282009-11-10 11:50:21 +01001008 &dev_attr_discard_alignment.attr,
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001009 &dev_attr_capability.attr,
1010 &dev_attr_stat.attr,
Nikanth Karthikesan316d3152009-10-06 20:16:55 +02001011 &dev_attr_inflight.attr,
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001012#ifdef CONFIG_FAIL_MAKE_REQUEST
1013 &dev_attr_fail.attr,
1014#endif
Jens Axboe581d4e22008-09-14 05:56:33 -07001015#ifdef CONFIG_FAIL_IO_TIMEOUT
1016 &dev_attr_fail_timeout.attr,
1017#endif
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001018 NULL
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019};
1020
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001021static struct attribute_group disk_attr_group = {
1022 .attrs = disk_attrs,
1023};
1024
David Brownella4dbd672009-06-24 10:06:31 -07001025static const struct attribute_group *disk_attr_groups[] = {
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001026 &disk_attr_group,
1027 NULL
1028};
1029
Tejun Heo540eed52008-08-25 19:56:15 +09001030/**
1031 * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way
1032 * @disk: disk to replace part_tbl for
1033 * @new_ptbl: new part_tbl to install
1034 *
1035 * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The
1036 * original ptbl is freed using RCU callback.
1037 *
1038 * LOCKING:
1039 * Matching bd_mutx locked.
1040 */
1041static void disk_replace_part_tbl(struct gendisk *disk,
1042 struct disk_part_tbl *new_ptbl)
1043{
1044 struct disk_part_tbl *old_ptbl = disk->part_tbl;
1045
1046 rcu_assign_pointer(disk->part_tbl, new_ptbl);
Jens Axboea6f23652008-10-24 12:52:42 +02001047
1048 if (old_ptbl) {
1049 rcu_assign_pointer(old_ptbl->last_lookup, NULL);
Lai Jiangshan57bdfbf2011-03-18 11:42:58 +08001050 kfree_rcu(old_ptbl, rcu_head);
Jens Axboea6f23652008-10-24 12:52:42 +02001051 }
Tejun Heo540eed52008-08-25 19:56:15 +09001052}
1053
1054/**
1055 * disk_expand_part_tbl - expand disk->part_tbl
1056 * @disk: disk to expand part_tbl for
1057 * @partno: expand such that this partno can fit in
1058 *
1059 * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl
1060 * uses RCU to allow unlocked dereferencing for stats and other stuff.
1061 *
1062 * LOCKING:
1063 * Matching bd_mutex locked, might sleep.
1064 *
1065 * RETURNS:
1066 * 0 on success, -errno on failure.
1067 */
1068int disk_expand_part_tbl(struct gendisk *disk, int partno)
1069{
1070 struct disk_part_tbl *old_ptbl = disk->part_tbl;
1071 struct disk_part_tbl *new_ptbl;
1072 int len = old_ptbl ? old_ptbl->len : 0;
1073 int target = partno + 1;
1074 size_t size;
1075 int i;
1076
1077 /* disk_max_parts() is zero during initialization, ignore if so */
1078 if (disk_max_parts(disk) && target > disk_max_parts(disk))
1079 return -EINVAL;
1080
1081 if (target <= len)
1082 return 0;
1083
1084 size = sizeof(*new_ptbl) + target * sizeof(new_ptbl->part[0]);
1085 new_ptbl = kzalloc_node(size, GFP_KERNEL, disk->node_id);
1086 if (!new_ptbl)
1087 return -ENOMEM;
1088
Tejun Heo540eed52008-08-25 19:56:15 +09001089 new_ptbl->len = target;
1090
1091 for (i = 0; i < len; i++)
1092 rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]);
1093
1094 disk_replace_part_tbl(disk, new_ptbl);
1095 return 0;
1096}
1097
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001098static void disk_release(struct device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001099{
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001100 struct gendisk *disk = dev_to_disk(dev);
1101
Keith Busch2da78092014-08-26 09:05:36 -06001102 blk_free_devt(dev->devt);
Tejun Heo77ea8872010-12-08 20:57:37 +01001103 disk_release_events(disk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001104 kfree(disk->random);
Tejun Heo540eed52008-08-25 19:56:15 +09001105 disk_replace_part_tbl(disk, NULL);
Tejun Heo074a7ac2008-08-25 19:56:14 +09001106 free_part_stats(&disk->part0);
Will Drewry6d1d8052010-08-31 15:47:05 -05001107 free_part_info(&disk->part0);
Tejun Heo523e1d32011-10-19 14:31:07 +02001108 if (disk->queue)
1109 blk_put_queue(disk->queue);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110 kfree(disk);
1111}
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001112struct class block_class = {
1113 .name = "block",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001114};
1115
Kay Sievers3c2670e2013-04-06 09:56:00 -07001116static char *block_devnode(struct device *dev, umode_t *mode,
Greg Kroah-Hartman4e4098a2013-04-11 11:43:29 -07001117 kuid_t *uid, kgid_t *gid)
Kay Sieversb03f38b2009-04-30 15:23:42 +02001118{
1119 struct gendisk *disk = dev_to_disk(dev);
1120
Kay Sieverse454cea2009-09-18 23:01:12 +02001121 if (disk->devnode)
1122 return disk->devnode(disk, mode);
Kay Sieversb03f38b2009-04-30 15:23:42 +02001123 return NULL;
1124}
1125
Adrian Bunk1826ead2008-03-04 11:23:46 +01001126static struct device_type disk_type = {
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001127 .name = "disk",
1128 .groups = disk_attr_groups,
1129 .release = disk_release,
Kay Sieverse454cea2009-09-18 23:01:12 +02001130 .devnode = block_devnode,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131};
1132
Randy Dunlapa6e2ba82008-05-23 09:44:11 -07001133#ifdef CONFIG_PROC_FS
Tejun Heocf771cb2008-09-03 09:01:09 +02001134/*
1135 * aggregate disk stat collector. Uses the same stats that the sysfs
1136 * entries do, above, but makes them available through one seq_file.
1137 *
1138 * The output looks suspiciously like /proc/partitions with a bunch of
1139 * extra fields.
1140 */
1141static int diskstats_show(struct seq_file *seqf, void *v)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001142{
1143 struct gendisk *gp = v;
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001144 struct disk_part_iter piter;
1145 struct hd_struct *hd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001146 char buf[BDEVNAME_SIZE];
Tejun Heoc9959052008-08-25 19:47:21 +09001147 int cpu;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001148
1149 /*
Tejun Heoed9e1982008-08-25 19:56:05 +09001150 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
Tejun Heocf771cb2008-09-03 09:01:09 +02001151 seq_puts(seqf, "major minor name"
Linus Torvalds1da177e2005-04-16 15:20:36 -07001152 " rio rmerge rsect ruse wio wmerge "
1153 "wsect wuse running use aveq"
1154 "\n\n");
1155 */
Wanlong Gao9f5e4862011-06-13 10:45:43 +02001156
Tejun Heo71982a42009-04-17 08:34:48 +02001157 disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0);
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001158 while ((hd = disk_part_iter_next(&piter))) {
Tejun Heo074a7ac2008-08-25 19:56:14 +09001159 cpu = part_stat_lock();
Tejun Heoc9959052008-08-25 19:47:21 +09001160 part_round_stats(cpu, hd);
Tejun Heo074a7ac2008-08-25 19:56:14 +09001161 part_stat_unlock();
Herbert Poetzlf95fe9c2011-08-02 12:43:50 +02001162 seq_printf(seqf, "%4d %7d %s %lu %lu %lu "
1163 "%u %lu %lu %lu %u %u %u %u\n",
Tejun Heof331c022008-09-03 09:01:48 +02001164 MAJOR(part_devt(hd)), MINOR(part_devt(hd)),
1165 disk_name(gp, hd->partno, buf),
Liu Yuan53f22952011-03-02 11:00:15 -05001166 part_stat_read(hd, ios[READ]),
1167 part_stat_read(hd, merges[READ]),
Herbert Poetzlf95fe9c2011-08-02 12:43:50 +02001168 part_stat_read(hd, sectors[READ]),
Liu Yuan53f22952011-03-02 11:00:15 -05001169 jiffies_to_msecs(part_stat_read(hd, ticks[READ])),
1170 part_stat_read(hd, ios[WRITE]),
1171 part_stat_read(hd, merges[WRITE]),
Herbert Poetzlf95fe9c2011-08-02 12:43:50 +02001172 part_stat_read(hd, sectors[WRITE]),
Liu Yuan53f22952011-03-02 11:00:15 -05001173 jiffies_to_msecs(part_stat_read(hd, ticks[WRITE])),
Nikanth Karthikesan316d3152009-10-06 20:16:55 +02001174 part_in_flight(hd),
Jerome Marchand28f39d52008-02-08 11:04:56 +01001175 jiffies_to_msecs(part_stat_read(hd, io_ticks)),
1176 jiffies_to_msecs(part_stat_read(hd, time_in_queue))
1177 );
Linus Torvalds1da177e2005-04-16 15:20:36 -07001178 }
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001179 disk_part_iter_exit(&piter);
Wanlong Gao9f5e4862011-06-13 10:45:43 +02001180
Linus Torvalds1da177e2005-04-16 15:20:36 -07001181 return 0;
1182}
1183
Alexey Dobriyan31d85ab2008-10-06 12:55:38 +04001184static const struct seq_operations diskstats_op = {
Tejun Heodef4e382008-09-03 08:57:12 +02001185 .start = disk_seqf_start,
1186 .next = disk_seqf_next,
1187 .stop = disk_seqf_stop,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001188 .show = diskstats_show
1189};
Alexey Dobriyanf5009752008-10-04 23:53:21 +04001190
Alexey Dobriyan31d85ab2008-10-06 12:55:38 +04001191static int diskstats_open(struct inode *inode, struct file *file)
1192{
1193 return seq_open(file, &diskstats_op);
1194}
1195
1196static const struct file_operations proc_diskstats_operations = {
1197 .open = diskstats_open,
1198 .read = seq_read,
1199 .llseek = seq_lseek,
1200 .release = seq_release,
1201};
1202
Alexey Dobriyanf5009752008-10-04 23:53:21 +04001203static int __init proc_genhd_init(void)
1204{
Alexey Dobriyan31d85ab2008-10-06 12:55:38 +04001205 proc_create("diskstats", 0, NULL, &proc_diskstats_operations);
Alexey Dobriyanf5009752008-10-04 23:53:21 +04001206 proc_create("partitions", 0, NULL, &proc_partitions_operations);
1207 return 0;
1208}
1209module_init(proc_genhd_init);
Randy Dunlapa6e2ba82008-05-23 09:44:11 -07001210#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001211
Tejun Heocf771cb2008-09-03 09:01:09 +02001212dev_t blk_lookup_devt(const char *name, int partno)
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001213{
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001214 dev_t devt = MKDEV(0, 0);
Tejun Heodef4e382008-09-03 08:57:12 +02001215 struct class_dev_iter iter;
1216 struct device *dev;
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001217
Tejun Heodef4e382008-09-03 08:57:12 +02001218 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1219 while ((dev = class_dev_iter_next(&iter))) {
1220 struct gendisk *disk = dev_to_disk(dev);
Tejun Heo548b10e2008-08-29 09:01:47 +02001221 struct hd_struct *part;
Tejun Heodef4e382008-09-03 08:57:12 +02001222
Kay Sievers3ada8b72009-01-06 10:44:43 -08001223 if (strcmp(dev_name(dev), name))
Tejun Heof331c022008-09-03 09:01:48 +02001224 continue;
Tejun Heof331c022008-09-03 09:01:48 +02001225
Neil Brown41b8c852009-02-18 10:33:59 +01001226 if (partno < disk->minors) {
1227 /* We need to return the right devno, even
1228 * if the partition doesn't exist yet.
1229 */
1230 devt = MKDEV(MAJOR(dev->devt),
1231 MINOR(dev->devt) + partno);
1232 break;
1233 }
Tejun Heo548b10e2008-08-29 09:01:47 +02001234 part = disk_get_part(disk, partno);
Tejun Heo2bbedcb2008-08-29 11:41:51 +02001235 if (part) {
Tejun Heof331c022008-09-03 09:01:48 +02001236 devt = part_devt(part);
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001237 disk_put_part(part);
Tejun Heo548b10e2008-08-29 09:01:47 +02001238 break;
Tejun Heodef4e382008-09-03 08:57:12 +02001239 }
Tejun Heo548b10e2008-08-29 09:01:47 +02001240 disk_put_part(part);
Kay Sievers5c0ef6d2008-08-16 14:30:30 +02001241 }
Tejun Heodef4e382008-09-03 08:57:12 +02001242 class_dev_iter_exit(&iter);
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001243 return devt;
1244}
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001245EXPORT_SYMBOL(blk_lookup_devt);
1246
Linus Torvalds1da177e2005-04-16 15:20:36 -07001247struct gendisk *alloc_disk(int minors)
1248{
Ezequiel Garciac304a512012-11-10 10:39:44 +01001249 return alloc_disk_node(minors, NUMA_NO_NODE);
Christoph Lameter19460892005-06-23 00:08:19 -07001250}
Tejun Heo689d6fa2008-08-25 19:56:16 +09001251EXPORT_SYMBOL(alloc_disk);
Christoph Lameter19460892005-06-23 00:08:19 -07001252
1253struct gendisk *alloc_disk_node(int minors, int node_id)
1254{
1255 struct gendisk *disk;
1256
Joe Perchesc1b511e2013-08-29 15:21:42 -07001257 disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001258 if (disk) {
Tejun Heo074a7ac2008-08-25 19:56:14 +09001259 if (!init_part_stats(&disk->part0)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001260 kfree(disk);
1261 return NULL;
1262 }
Cheng Renquanbf91db12008-11-20 08:37:37 +01001263 disk->node_id = node_id;
Tejun Heo540eed52008-08-25 19:56:15 +09001264 if (disk_expand_part_tbl(disk, 0)) {
1265 free_part_stats(&disk->part0);
Tejun Heob5d0b9d2008-09-03 09:06:42 +02001266 kfree(disk);
1267 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001268 }
Tejun Heo540eed52008-08-25 19:56:15 +09001269 disk->part_tbl->part[0] = &disk->part0;
Jens Axboe6c23a962011-01-07 08:43:37 +01001270
Vivek Goyalc83f6bf2012-08-01 12:24:18 +02001271 /*
1272 * set_capacity() and get_capacity() currently don't use
1273 * seqcounter to read/update the part0->nr_sects. Still init
1274 * the counter as we can read the sectors in IO submission
1275 * patch using seqence counters.
1276 *
1277 * TODO: Ideally set_capacity() and get_capacity() should be
1278 * converted to make use of bd_mutex and sequence counters.
1279 */
1280 seqcount_init(&disk->part0.nr_sects_seq);
Jens Axboe6c23a962011-01-07 08:43:37 +01001281 hd_ref_init(&disk->part0);
Tejun Heob5d0b9d2008-09-03 09:06:42 +02001282
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283 disk->minors = minors;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284 rand_initialize_disk(disk);
Tejun Heoed9e1982008-08-25 19:56:05 +09001285 disk_to_dev(disk)->class = &block_class;
1286 disk_to_dev(disk)->type = &disk_type;
1287 device_initialize(disk_to_dev(disk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001288 }
1289 return disk;
1290}
Christoph Lameter19460892005-06-23 00:08:19 -07001291EXPORT_SYMBOL(alloc_disk_node);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292
1293struct kobject *get_disk(struct gendisk *disk)
1294{
1295 struct module *owner;
1296 struct kobject *kobj;
1297
1298 if (!disk->fops)
1299 return NULL;
1300 owner = disk->fops->owner;
1301 if (owner && !try_module_get(owner))
1302 return NULL;
Tejun Heoed9e1982008-08-25 19:56:05 +09001303 kobj = kobject_get(&disk_to_dev(disk)->kobj);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001304 if (kobj == NULL) {
1305 module_put(owner);
1306 return NULL;
1307 }
1308 return kobj;
1309
1310}
1311
1312EXPORT_SYMBOL(get_disk);
1313
1314void put_disk(struct gendisk *disk)
1315{
1316 if (disk)
Tejun Heoed9e1982008-08-25 19:56:05 +09001317 kobject_put(&disk_to_dev(disk)->kobj);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001318}
1319
1320EXPORT_SYMBOL(put_disk);
1321
Hannes Reineckee3264a42009-07-28 09:13:13 +02001322static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1323{
1324 char event[] = "DISK_RO=1";
1325 char *envp[] = { event, NULL };
1326
1327 if (!ro)
1328 event[8] = '0';
1329 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1330}
1331
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332void set_device_ro(struct block_device *bdev, int flag)
1333{
Tejun Heob7db9952008-08-25 19:56:10 +09001334 bdev->bd_part->policy = flag;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001335}
1336
1337EXPORT_SYMBOL(set_device_ro);
1338
1339void set_disk_ro(struct gendisk *disk, int flag)
1340{
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001341 struct disk_part_iter piter;
1342 struct hd_struct *part;
1343
Hannes Reineckee3264a42009-07-28 09:13:13 +02001344 if (disk->part0.policy != flag) {
1345 set_disk_ro_uevent(disk, flag);
1346 disk->part0.policy = flag;
1347 }
1348
1349 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY);
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001350 while ((part = disk_part_iter_next(&piter)))
1351 part->policy = flag;
1352 disk_part_iter_exit(&piter);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001353}
1354
1355EXPORT_SYMBOL(set_disk_ro);
1356
1357int bdev_read_only(struct block_device *bdev)
1358{
1359 if (!bdev)
1360 return 0;
Tejun Heob7db9952008-08-25 19:56:10 +09001361 return bdev->bd_part->policy;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362}
1363
1364EXPORT_SYMBOL(bdev_read_only);
1365
Tejun Heocf771cb2008-09-03 09:01:09 +02001366int invalidate_partition(struct gendisk *disk, int partno)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001367{
1368 int res = 0;
Tejun Heocf771cb2008-09-03 09:01:09 +02001369 struct block_device *bdev = bdget_disk(disk, partno);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001370 if (bdev) {
Christoph Hellwig2ef41632005-05-05 16:15:59 -07001371 fsync_bdev(bdev);
NeilBrown93b270f2011-02-24 17:25:47 +11001372 res = __invalidate_device(bdev, true);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001373 bdput(bdev);
1374 }
1375 return res;
1376}
1377
1378EXPORT_SYMBOL(invalidate_partition);
Tejun Heo77ea8872010-12-08 20:57:37 +01001379
1380/*
1381 * Disk events - monitor disk events like media change and eject request.
1382 */
1383struct disk_events {
1384 struct list_head node; /* all disk_event's */
1385 struct gendisk *disk; /* the associated disk */
1386 spinlock_t lock;
1387
Tejun Heofdd514e2011-06-09 20:43:59 +02001388 struct mutex block_mutex; /* protects blocking */
Tejun Heo77ea8872010-12-08 20:57:37 +01001389 int block; /* event blocking depth */
1390 unsigned int pending; /* events already sent out */
1391 unsigned int clearing; /* events being cleared */
1392
1393 long poll_msecs; /* interval, -1 for default */
1394 struct delayed_work dwork;
1395};
1396
1397static const char *disk_events_strs[] = {
1398 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "media_change",
1399 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "eject_request",
1400};
1401
1402static char *disk_uevents[] = {
1403 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "DISK_MEDIA_CHANGE=1",
1404 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "DISK_EJECT_REQUEST=1",
1405};
1406
1407/* list of all disk_events */
1408static DEFINE_MUTEX(disk_events_mutex);
1409static LIST_HEAD(disk_events);
1410
1411/* disable in-kernel polling by default */
1412static unsigned long disk_events_dfl_poll_msecs = 0;
1413
1414static unsigned long disk_events_poll_jiffies(struct gendisk *disk)
1415{
1416 struct disk_events *ev = disk->ev;
1417 long intv_msecs = 0;
1418
1419 /*
1420 * If device-specific poll interval is set, always use it. If
1421 * the default is being used, poll iff there are events which
1422 * can't be monitored asynchronously.
1423 */
1424 if (ev->poll_msecs >= 0)
1425 intv_msecs = ev->poll_msecs;
1426 else if (disk->events & ~disk->async_events)
1427 intv_msecs = disk_events_dfl_poll_msecs;
1428
1429 return msecs_to_jiffies(intv_msecs);
1430}
1431
Tejun Heoc3af54a2011-06-09 20:43:55 +02001432/**
1433 * disk_block_events - block and flush disk event checking
1434 * @disk: disk to block events for
1435 *
1436 * On return from this function, it is guaranteed that event checking
1437 * isn't in progress and won't happen until unblocked by
1438 * disk_unblock_events(). Events blocking is counted and the actual
1439 * unblocking happens after the matching number of unblocks are done.
1440 *
1441 * Note that this intentionally does not block event checking from
1442 * disk_clear_events().
1443 *
1444 * CONTEXT:
1445 * Might sleep.
1446 */
1447void disk_block_events(struct gendisk *disk)
Tejun Heo77ea8872010-12-08 20:57:37 +01001448{
1449 struct disk_events *ev = disk->ev;
1450 unsigned long flags;
1451 bool cancel;
1452
Tejun Heoc3af54a2011-06-09 20:43:55 +02001453 if (!ev)
1454 return;
1455
Tejun Heofdd514e2011-06-09 20:43:59 +02001456 /*
1457 * Outer mutex ensures that the first blocker completes canceling
1458 * the event work before further blockers are allowed to finish.
1459 */
1460 mutex_lock(&ev->block_mutex);
1461
Tejun Heo77ea8872010-12-08 20:57:37 +01001462 spin_lock_irqsave(&ev->lock, flags);
1463 cancel = !ev->block++;
1464 spin_unlock_irqrestore(&ev->lock, flags);
1465
Tejun Heoc3af54a2011-06-09 20:43:55 +02001466 if (cancel)
1467 cancel_delayed_work_sync(&disk->ev->dwork);
Tejun Heofdd514e2011-06-09 20:43:59 +02001468
1469 mutex_unlock(&ev->block_mutex);
Tejun Heo77ea8872010-12-08 20:57:37 +01001470}
1471
1472static void __disk_unblock_events(struct gendisk *disk, bool check_now)
1473{
1474 struct disk_events *ev = disk->ev;
1475 unsigned long intv;
1476 unsigned long flags;
1477
1478 spin_lock_irqsave(&ev->lock, flags);
1479
1480 if (WARN_ON_ONCE(ev->block <= 0))
1481 goto out_unlock;
1482
1483 if (--ev->block)
1484 goto out_unlock;
1485
1486 /*
1487 * Not exactly a latency critical operation, set poll timer
1488 * slack to 25% and kick event check.
1489 */
1490 intv = disk_events_poll_jiffies(disk);
1491 set_timer_slack(&ev->dwork.timer, intv / 4);
1492 if (check_now)
Viresh Kumar695588f2013-04-24 17:12:56 +05301493 queue_delayed_work(system_freezable_power_efficient_wq,
1494 &ev->dwork, 0);
Tejun Heo77ea8872010-12-08 20:57:37 +01001495 else if (intv)
Viresh Kumar695588f2013-04-24 17:12:56 +05301496 queue_delayed_work(system_freezable_power_efficient_wq,
1497 &ev->dwork, intv);
Tejun Heo77ea8872010-12-08 20:57:37 +01001498out_unlock:
1499 spin_unlock_irqrestore(&ev->lock, flags);
1500}
1501
1502/**
Tejun Heo77ea8872010-12-08 20:57:37 +01001503 * disk_unblock_events - unblock disk event checking
1504 * @disk: disk to unblock events for
1505 *
1506 * Undo disk_block_events(). When the block count reaches zero, it
1507 * starts events polling if configured.
1508 *
1509 * CONTEXT:
1510 * Don't care. Safe to call from irq context.
1511 */
1512void disk_unblock_events(struct gendisk *disk)
1513{
1514 if (disk->ev)
Tejun Heofacc31d2011-03-09 19:54:27 +01001515 __disk_unblock_events(disk, false);
Tejun Heo77ea8872010-12-08 20:57:37 +01001516}
1517
1518/**
Tejun Heo85ef06d2011-07-01 16:17:47 +02001519 * disk_flush_events - schedule immediate event checking and flushing
1520 * @disk: disk to check and flush events for
1521 * @mask: events to flush
Tejun Heo77ea8872010-12-08 20:57:37 +01001522 *
Tejun Heo85ef06d2011-07-01 16:17:47 +02001523 * Schedule immediate event checking on @disk if not blocked. Events in
1524 * @mask are scheduled to be cleared from the driver. Note that this
1525 * doesn't clear the events from @disk->ev.
Tejun Heo77ea8872010-12-08 20:57:37 +01001526 *
1527 * CONTEXT:
Tejun Heo85ef06d2011-07-01 16:17:47 +02001528 * If @mask is non-zero must be called with bdev->bd_mutex held.
Tejun Heo77ea8872010-12-08 20:57:37 +01001529 */
Tejun Heo85ef06d2011-07-01 16:17:47 +02001530void disk_flush_events(struct gendisk *disk, unsigned int mask)
Tejun Heo77ea8872010-12-08 20:57:37 +01001531{
Tejun Heoa9dce2a2011-06-09 20:43:54 +02001532 struct disk_events *ev = disk->ev;
Tejun Heoa9dce2a2011-06-09 20:43:54 +02001533
1534 if (!ev)
1535 return;
1536
Tejun Heo85ef06d2011-07-01 16:17:47 +02001537 spin_lock_irq(&ev->lock);
1538 ev->clearing |= mask;
Tejun Heo41f63c52012-08-03 10:30:47 -07001539 if (!ev->block)
Viresh Kumar695588f2013-04-24 17:12:56 +05301540 mod_delayed_work(system_freezable_power_efficient_wq,
1541 &ev->dwork, 0);
Tejun Heo85ef06d2011-07-01 16:17:47 +02001542 spin_unlock_irq(&ev->lock);
Tejun Heo77ea8872010-12-08 20:57:37 +01001543}
Tejun Heo77ea8872010-12-08 20:57:37 +01001544
1545/**
1546 * disk_clear_events - synchronously check, clear and return pending events
1547 * @disk: disk to fetch and clear events from
1548 * @mask: mask of events to be fetched and clearted
1549 *
1550 * Disk events are synchronously checked and pending events in @mask
1551 * are cleared and returned. This ignores the block count.
1552 *
1553 * CONTEXT:
1554 * Might sleep.
1555 */
1556unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask)
1557{
1558 const struct block_device_operations *bdops = disk->fops;
1559 struct disk_events *ev = disk->ev;
1560 unsigned int pending;
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001561 unsigned int clearing = mask;
Tejun Heo77ea8872010-12-08 20:57:37 +01001562
1563 if (!ev) {
1564 /* for drivers still using the old ->media_changed method */
1565 if ((mask & DISK_EVENT_MEDIA_CHANGE) &&
1566 bdops->media_changed && bdops->media_changed(disk))
1567 return DISK_EVENT_MEDIA_CHANGE;
1568 return 0;
1569 }
1570
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001571 disk_block_events(disk);
1572
1573 /*
1574 * store the union of mask and ev->clearing on the stack so that the
1575 * race with disk_flush_events does not cause ambiguity (ev->clearing
1576 * can still be modified even if events are blocked).
1577 */
Tejun Heo77ea8872010-12-08 20:57:37 +01001578 spin_lock_irq(&ev->lock);
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001579 clearing |= ev->clearing;
1580 ev->clearing = 0;
Tejun Heo77ea8872010-12-08 20:57:37 +01001581 spin_unlock_irq(&ev->lock);
1582
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001583 disk_check_events(ev, &clearing);
Derek Basehoreaea24a82012-12-18 12:27:18 -08001584 /*
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001585 * if ev->clearing is not 0, the disk_flush_events got called in the
1586 * middle of this function, so we want to run the workfn without delay.
Derek Basehoreaea24a82012-12-18 12:27:18 -08001587 */
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001588 __disk_unblock_events(disk, ev->clearing ? true : false);
Tejun Heo77ea8872010-12-08 20:57:37 +01001589
1590 /* then, fetch and clear pending events */
1591 spin_lock_irq(&ev->lock);
Tejun Heo77ea8872010-12-08 20:57:37 +01001592 pending = ev->pending & mask;
1593 ev->pending &= ~mask;
1594 spin_unlock_irq(&ev->lock);
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001595 WARN_ON_ONCE(clearing & mask);
Tejun Heo77ea8872010-12-08 20:57:37 +01001596
1597 return pending;
1598}
1599
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001600/*
1601 * Separate this part out so that a different pointer for clearing_ptr can be
1602 * passed in for disk_clear_events.
1603 */
Tejun Heo77ea8872010-12-08 20:57:37 +01001604static void disk_events_workfn(struct work_struct *work)
1605{
1606 struct delayed_work *dwork = to_delayed_work(work);
1607 struct disk_events *ev = container_of(dwork, struct disk_events, dwork);
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001608
1609 disk_check_events(ev, &ev->clearing);
1610}
1611
1612static void disk_check_events(struct disk_events *ev,
1613 unsigned int *clearing_ptr)
1614{
Tejun Heo77ea8872010-12-08 20:57:37 +01001615 struct gendisk *disk = ev->disk;
1616 char *envp[ARRAY_SIZE(disk_uevents) + 1] = { };
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001617 unsigned int clearing = *clearing_ptr;
Tejun Heo77ea8872010-12-08 20:57:37 +01001618 unsigned int events;
1619 unsigned long intv;
1620 int nr_events = 0, i;
1621
1622 /* check events */
1623 events = disk->fops->check_events(disk, clearing);
1624
1625 /* accumulate pending events and schedule next poll if necessary */
1626 spin_lock_irq(&ev->lock);
1627
1628 events &= ~ev->pending;
1629 ev->pending |= events;
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001630 *clearing_ptr &= ~clearing;
Tejun Heo77ea8872010-12-08 20:57:37 +01001631
1632 intv = disk_events_poll_jiffies(disk);
1633 if (!ev->block && intv)
Viresh Kumar695588f2013-04-24 17:12:56 +05301634 queue_delayed_work(system_freezable_power_efficient_wq,
1635 &ev->dwork, intv);
Tejun Heo77ea8872010-12-08 20:57:37 +01001636
1637 spin_unlock_irq(&ev->lock);
1638
Tejun Heo7c88a162011-04-21 19:43:58 +02001639 /*
1640 * Tell userland about new events. Only the events listed in
1641 * @disk->events are reported. Unlisted events are processed the
1642 * same internally but never get reported to userland.
1643 */
Tejun Heo77ea8872010-12-08 20:57:37 +01001644 for (i = 0; i < ARRAY_SIZE(disk_uevents); i++)
Tejun Heo7c88a162011-04-21 19:43:58 +02001645 if (events & disk->events & (1 << i))
Tejun Heo77ea8872010-12-08 20:57:37 +01001646 envp[nr_events++] = disk_uevents[i];
1647
1648 if (nr_events)
1649 kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp);
1650}
1651
1652/*
1653 * A disk events enabled device has the following sysfs nodes under
1654 * its /sys/block/X/ directory.
1655 *
1656 * events : list of all supported events
1657 * events_async : list of events which can be detected w/o polling
1658 * events_poll_msecs : polling interval, 0: disable, -1: system default
1659 */
1660static ssize_t __disk_events_show(unsigned int events, char *buf)
1661{
1662 const char *delim = "";
1663 ssize_t pos = 0;
1664 int i;
1665
1666 for (i = 0; i < ARRAY_SIZE(disk_events_strs); i++)
1667 if (events & (1 << i)) {
1668 pos += sprintf(buf + pos, "%s%s",
1669 delim, disk_events_strs[i]);
1670 delim = " ";
1671 }
1672 if (pos)
1673 pos += sprintf(buf + pos, "\n");
1674 return pos;
1675}
1676
1677static ssize_t disk_events_show(struct device *dev,
1678 struct device_attribute *attr, char *buf)
1679{
1680 struct gendisk *disk = dev_to_disk(dev);
1681
1682 return __disk_events_show(disk->events, buf);
1683}
1684
1685static ssize_t disk_events_async_show(struct device *dev,
1686 struct device_attribute *attr, char *buf)
1687{
1688 struct gendisk *disk = dev_to_disk(dev);
1689
1690 return __disk_events_show(disk->async_events, buf);
1691}
1692
1693static ssize_t disk_events_poll_msecs_show(struct device *dev,
1694 struct device_attribute *attr,
1695 char *buf)
1696{
1697 struct gendisk *disk = dev_to_disk(dev);
1698
1699 return sprintf(buf, "%ld\n", disk->ev->poll_msecs);
1700}
1701
1702static ssize_t disk_events_poll_msecs_store(struct device *dev,
1703 struct device_attribute *attr,
1704 const char *buf, size_t count)
1705{
1706 struct gendisk *disk = dev_to_disk(dev);
1707 long intv;
1708
1709 if (!count || !sscanf(buf, "%ld", &intv))
1710 return -EINVAL;
1711
1712 if (intv < 0 && intv != -1)
1713 return -EINVAL;
1714
Tejun Heoc3af54a2011-06-09 20:43:55 +02001715 disk_block_events(disk);
Tejun Heo77ea8872010-12-08 20:57:37 +01001716 disk->ev->poll_msecs = intv;
1717 __disk_unblock_events(disk, true);
1718
1719 return count;
1720}
1721
1722static const DEVICE_ATTR(events, S_IRUGO, disk_events_show, NULL);
1723static const DEVICE_ATTR(events_async, S_IRUGO, disk_events_async_show, NULL);
1724static const DEVICE_ATTR(events_poll_msecs, S_IRUGO|S_IWUSR,
1725 disk_events_poll_msecs_show,
1726 disk_events_poll_msecs_store);
1727
1728static const struct attribute *disk_events_attrs[] = {
1729 &dev_attr_events.attr,
1730 &dev_attr_events_async.attr,
1731 &dev_attr_events_poll_msecs.attr,
1732 NULL,
1733};
1734
1735/*
1736 * The default polling interval can be specified by the kernel
1737 * parameter block.events_dfl_poll_msecs which defaults to 0
1738 * (disable). This can also be modified runtime by writing to
1739 * /sys/module/block/events_dfl_poll_msecs.
1740 */
1741static int disk_events_set_dfl_poll_msecs(const char *val,
1742 const struct kernel_param *kp)
1743{
1744 struct disk_events *ev;
1745 int ret;
1746
1747 ret = param_set_ulong(val, kp);
1748 if (ret < 0)
1749 return ret;
1750
1751 mutex_lock(&disk_events_mutex);
1752
1753 list_for_each_entry(ev, &disk_events, node)
Tejun Heo85ef06d2011-07-01 16:17:47 +02001754 disk_flush_events(ev->disk, 0);
Tejun Heo77ea8872010-12-08 20:57:37 +01001755
1756 mutex_unlock(&disk_events_mutex);
1757
1758 return 0;
1759}
1760
1761static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops = {
1762 .set = disk_events_set_dfl_poll_msecs,
1763 .get = param_get_ulong,
1764};
1765
1766#undef MODULE_PARAM_PREFIX
1767#define MODULE_PARAM_PREFIX "block."
1768
1769module_param_cb(events_dfl_poll_msecs, &disk_events_dfl_poll_msecs_param_ops,
1770 &disk_events_dfl_poll_msecs, 0644);
1771
1772/*
Stanislaw Gruszka9f53d2fe2012-03-02 10:43:28 +01001773 * disk_{alloc|add|del|release}_events - initialize and destroy disk_events.
Tejun Heo77ea8872010-12-08 20:57:37 +01001774 */
Stanislaw Gruszka9f53d2fe2012-03-02 10:43:28 +01001775static void disk_alloc_events(struct gendisk *disk)
Tejun Heo77ea8872010-12-08 20:57:37 +01001776{
1777 struct disk_events *ev;
1778
Tejun Heo75e3f3e2011-05-26 21:06:50 +02001779 if (!disk->fops->check_events)
Tejun Heo77ea8872010-12-08 20:57:37 +01001780 return;
1781
1782 ev = kzalloc(sizeof(*ev), GFP_KERNEL);
1783 if (!ev) {
1784 pr_warn("%s: failed to initialize events\n", disk->disk_name);
1785 return;
1786 }
1787
Tejun Heo77ea8872010-12-08 20:57:37 +01001788 INIT_LIST_HEAD(&ev->node);
1789 ev->disk = disk;
1790 spin_lock_init(&ev->lock);
Tejun Heofdd514e2011-06-09 20:43:59 +02001791 mutex_init(&ev->block_mutex);
Tejun Heo77ea8872010-12-08 20:57:37 +01001792 ev->block = 1;
1793 ev->poll_msecs = -1;
1794 INIT_DELAYED_WORK(&ev->dwork, disk_events_workfn);
1795
Stanislaw Gruszka9f53d2fe2012-03-02 10:43:28 +01001796 disk->ev = ev;
1797}
1798
1799static void disk_add_events(struct gendisk *disk)
1800{
1801 if (!disk->ev)
1802 return;
1803
1804 /* FIXME: error handling */
1805 if (sysfs_create_files(&disk_to_dev(disk)->kobj, disk_events_attrs) < 0)
1806 pr_warn("%s: failed to create sysfs files for events\n",
1807 disk->disk_name);
1808
Tejun Heo77ea8872010-12-08 20:57:37 +01001809 mutex_lock(&disk_events_mutex);
Stanislaw Gruszka9f53d2fe2012-03-02 10:43:28 +01001810 list_add_tail(&disk->ev->node, &disk_events);
Tejun Heo77ea8872010-12-08 20:57:37 +01001811 mutex_unlock(&disk_events_mutex);
1812
1813 /*
1814 * Block count is initialized to 1 and the following initial
1815 * unblock kicks it into action.
1816 */
1817 __disk_unblock_events(disk, true);
1818}
1819
1820static void disk_del_events(struct gendisk *disk)
1821{
1822 if (!disk->ev)
1823 return;
1824
Tejun Heoc3af54a2011-06-09 20:43:55 +02001825 disk_block_events(disk);
Tejun Heo77ea8872010-12-08 20:57:37 +01001826
1827 mutex_lock(&disk_events_mutex);
1828 list_del_init(&disk->ev->node);
1829 mutex_unlock(&disk_events_mutex);
1830
1831 sysfs_remove_files(&disk_to_dev(disk)->kobj, disk_events_attrs);
1832}
1833
1834static void disk_release_events(struct gendisk *disk)
1835{
1836 /* the block count should be 1 from disk_del_events() */
1837 WARN_ON_ONCE(disk->ev && disk->ev->block != 1);
1838 kfree(disk->ev);
1839}