blob: e5cafa51567c9d589147523c8ab7b43504f9d725 [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>
Tejun Heo66114ca2015-05-22 17:13:32 -040011#include <linux/backing-dev.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070012#include <linux/init.h>
13#include <linux/spinlock.h>
Alexey Dobriyanf5009752008-10-04 23:53:21 +040014#include <linux/proc_fs.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070015#include <linux/seq_file.h>
16#include <linux/slab.h>
17#include <linux/kmod.h>
18#include <linux/kobj_map.h>
Jes Sorensen58383af2006-02-06 14:12:43 -080019#include <linux/mutex.h>
Tejun Heobcce3de2008-08-25 19:47:22 +090020#include <linux/idr.h>
Tejun Heo77ea8872010-12-08 20:57:37 +010021#include <linux/log2.h>
Ming Lei25e823c2013-02-22 16:34:13 -080022#include <linux/pm_runtime.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070023
Adrian Bunkff889722008-03-04 11:23:45 +010024#include "blk.h"
25
Kay Sieversedfaa7c2007-05-21 22:08:01 +020026static DEFINE_MUTEX(block_class_lock);
Kay Sieversedfaa7c2007-05-21 22:08:01 +020027struct kobject *block_depr;
Linus Torvalds1da177e2005-04-16 15:20:36 -070028
Tejun Heobcce3de2008-08-25 19:47:22 +090029/* for extended dynamic devt allocation, currently only one major is used */
Tejun Heoce23bba2013-02-27 17:03:56 -080030#define NR_EXT_DEVT (1 << MINORBITS)
Tejun Heobcce3de2008-08-25 19:47:22 +090031
Keith Busch2da78092014-08-26 09:05:36 -060032/* For extended devt allocation. ext_devt_lock prevents look up
Tejun Heobcce3de2008-08-25 19:47:22 +090033 * results from going away underneath its user.
34 */
Keith Busch2da78092014-08-26 09:05:36 -060035static DEFINE_SPINLOCK(ext_devt_lock);
Tejun Heobcce3de2008-08-25 19:47:22 +090036static DEFINE_IDR(ext_devt_idr);
37
Adrian Bunk1826ead2008-03-04 11:23:46 +010038static struct device_type disk_type;
39
Derek Basehore12c2bdb2012-12-18 12:27:20 -080040static void disk_check_events(struct disk_events *ev,
41 unsigned int *clearing_ptr);
Stanislaw Gruszka9f53d2fe2012-03-02 10:43:28 +010042static void disk_alloc_events(struct gendisk *disk);
Tejun Heo77ea8872010-12-08 20:57:37 +010043static void disk_add_events(struct gendisk *disk);
44static void disk_del_events(struct gendisk *disk);
45static void disk_release_events(struct gendisk *disk);
46
Tejun Heoe71bf0d2008-09-03 09:03:02 +020047/**
48 * disk_get_part - get partition
49 * @disk: disk to look partition from
50 * @partno: partition number
51 *
52 * Look for partition @partno from @disk. If found, increment
53 * reference count and return it.
54 *
55 * CONTEXT:
56 * Don't care.
57 *
58 * RETURNS:
59 * Pointer to the found partition on success, NULL if not found.
60 */
61struct hd_struct *disk_get_part(struct gendisk *disk, int partno)
62{
Tejun Heo540eed52008-08-25 19:56:15 +090063 struct hd_struct *part = NULL;
64 struct disk_part_tbl *ptbl;
Tejun Heoe71bf0d2008-09-03 09:03:02 +020065
Tejun Heo540eed52008-08-25 19:56:15 +090066 if (unlikely(partno < 0))
Tejun Heoe71bf0d2008-09-03 09:03:02 +020067 return NULL;
Tejun Heo540eed52008-08-25 19:56:15 +090068
Tejun Heoe71bf0d2008-09-03 09:03:02 +020069 rcu_read_lock();
Tejun Heo540eed52008-08-25 19:56:15 +090070
71 ptbl = rcu_dereference(disk->part_tbl);
72 if (likely(partno < ptbl->len)) {
73 part = rcu_dereference(ptbl->part[partno]);
74 if (part)
75 get_device(part_to_dev(part));
76 }
77
Tejun Heoe71bf0d2008-09-03 09:03:02 +020078 rcu_read_unlock();
79
80 return part;
81}
82EXPORT_SYMBOL_GPL(disk_get_part);
83
84/**
85 * disk_part_iter_init - initialize partition iterator
86 * @piter: iterator to initialize
87 * @disk: disk to iterate over
88 * @flags: DISK_PITER_* flags
89 *
90 * Initialize @piter so that it iterates over partitions of @disk.
91 *
92 * CONTEXT:
93 * Don't care.
94 */
95void disk_part_iter_init(struct disk_part_iter *piter, struct gendisk *disk,
96 unsigned int flags)
97{
Tejun Heo540eed52008-08-25 19:56:15 +090098 struct disk_part_tbl *ptbl;
99
100 rcu_read_lock();
101 ptbl = rcu_dereference(disk->part_tbl);
102
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200103 piter->disk = disk;
104 piter->part = NULL;
105
106 if (flags & DISK_PITER_REVERSE)
Tejun Heo540eed52008-08-25 19:56:15 +0900107 piter->idx = ptbl->len - 1;
Tejun Heo71982a42009-04-17 08:34:48 +0200108 else if (flags & (DISK_PITER_INCL_PART0 | DISK_PITER_INCL_EMPTY_PART0))
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200109 piter->idx = 0;
Tejun Heob5d0b9d2008-09-03 09:06:42 +0200110 else
111 piter->idx = 1;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200112
113 piter->flags = flags;
Tejun Heo540eed52008-08-25 19:56:15 +0900114
115 rcu_read_unlock();
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200116}
117EXPORT_SYMBOL_GPL(disk_part_iter_init);
118
119/**
120 * disk_part_iter_next - proceed iterator to the next partition and return it
121 * @piter: iterator of interest
122 *
123 * Proceed @piter to the next partition and return it.
124 *
125 * CONTEXT:
126 * Don't care.
127 */
128struct hd_struct *disk_part_iter_next(struct disk_part_iter *piter)
129{
Tejun Heo540eed52008-08-25 19:56:15 +0900130 struct disk_part_tbl *ptbl;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200131 int inc, end;
132
133 /* put the last partition */
134 disk_put_part(piter->part);
135 piter->part = NULL;
136
Tejun Heo540eed52008-08-25 19:56:15 +0900137 /* get part_tbl */
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200138 rcu_read_lock();
Tejun Heo540eed52008-08-25 19:56:15 +0900139 ptbl = rcu_dereference(piter->disk->part_tbl);
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200140
141 /* determine iteration parameters */
142 if (piter->flags & DISK_PITER_REVERSE) {
143 inc = -1;
Tejun Heo71982a42009-04-17 08:34:48 +0200144 if (piter->flags & (DISK_PITER_INCL_PART0 |
145 DISK_PITER_INCL_EMPTY_PART0))
Tejun Heob5d0b9d2008-09-03 09:06:42 +0200146 end = -1;
147 else
148 end = 0;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200149 } else {
150 inc = 1;
Tejun Heo540eed52008-08-25 19:56:15 +0900151 end = ptbl->len;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200152 }
153
154 /* iterate to the next partition */
155 for (; piter->idx != end; piter->idx += inc) {
156 struct hd_struct *part;
157
Tejun Heo540eed52008-08-25 19:56:15 +0900158 part = rcu_dereference(ptbl->part[piter->idx]);
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200159 if (!part)
160 continue;
Vivek Goyalc83f6bf2012-08-01 12:24:18 +0200161 if (!part_nr_sects_read(part) &&
Tejun Heo71982a42009-04-17 08:34:48 +0200162 !(piter->flags & DISK_PITER_INCL_EMPTY) &&
163 !(piter->flags & DISK_PITER_INCL_EMPTY_PART0 &&
164 piter->idx == 0))
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200165 continue;
166
Tejun Heoed9e1982008-08-25 19:56:05 +0900167 get_device(part_to_dev(part));
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200168 piter->part = part;
169 piter->idx += inc;
170 break;
171 }
172
173 rcu_read_unlock();
174
175 return piter->part;
176}
177EXPORT_SYMBOL_GPL(disk_part_iter_next);
178
179/**
180 * disk_part_iter_exit - finish up partition iteration
181 * @piter: iter of interest
182 *
183 * Called when iteration is over. Cleans up @piter.
184 *
185 * CONTEXT:
186 * Don't care.
187 */
188void disk_part_iter_exit(struct disk_part_iter *piter)
189{
190 disk_put_part(piter->part);
191 piter->part = NULL;
192}
193EXPORT_SYMBOL_GPL(disk_part_iter_exit);
194
Jens Axboea6f23652008-10-24 12:52:42 +0200195static inline int sector_in_part(struct hd_struct *part, sector_t sector)
196{
197 return part->start_sect <= sector &&
Vivek Goyalc83f6bf2012-08-01 12:24:18 +0200198 sector < part->start_sect + part_nr_sects_read(part);
Jens Axboea6f23652008-10-24 12:52:42 +0200199}
200
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200201/**
202 * disk_map_sector_rcu - map sector to partition
203 * @disk: gendisk of interest
204 * @sector: sector to map
205 *
206 * Find out which partition @sector maps to on @disk. This is
207 * primarily used for stats accounting.
208 *
209 * CONTEXT:
210 * RCU read locked. The returned partition pointer is valid only
211 * while preemption is disabled.
212 *
213 * RETURNS:
Tejun Heo074a7ac2008-08-25 19:56:14 +0900214 * Found partition on success, part0 is returned if no partition matches
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200215 */
216struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector)
217{
Tejun Heo540eed52008-08-25 19:56:15 +0900218 struct disk_part_tbl *ptbl;
Jens Axboea6f23652008-10-24 12:52:42 +0200219 struct hd_struct *part;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200220 int i;
221
Tejun Heo540eed52008-08-25 19:56:15 +0900222 ptbl = rcu_dereference(disk->part_tbl);
223
Jens Axboea6f23652008-10-24 12:52:42 +0200224 part = rcu_dereference(ptbl->last_lookup);
225 if (part && sector_in_part(part, sector))
226 return part;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200227
Jens Axboea6f23652008-10-24 12:52:42 +0200228 for (i = 1; i < ptbl->len; i++) {
229 part = rcu_dereference(ptbl->part[i]);
230
231 if (part && sector_in_part(part, sector)) {
232 rcu_assign_pointer(ptbl->last_lookup, part);
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200233 return part;
Jens Axboea6f23652008-10-24 12:52:42 +0200234 }
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200235 }
Tejun Heo074a7ac2008-08-25 19:56:14 +0900236 return &disk->part0;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200237}
238EXPORT_SYMBOL_GPL(disk_map_sector_rcu);
239
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240/*
241 * Can be deleted altogether. Later.
242 *
243 */
244static struct blk_major_name {
245 struct blk_major_name *next;
246 int major;
247 char name[16];
Joe Korty68eef3b2006-03-31 02:30:32 -0800248} *major_names[BLKDEV_MAJOR_HASH_SIZE];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249
250/* index in the above - for now: assume no multimajor ranges */
Yang Zhange61eb2e2010-12-17 09:00:18 +0100251static inline int major_to_index(unsigned major)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252{
Joe Korty68eef3b2006-03-31 02:30:32 -0800253 return major % BLKDEV_MAJOR_HASH_SIZE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254}
255
Joe Korty68eef3b2006-03-31 02:30:32 -0800256#ifdef CONFIG_PROC_FS
Tejun Heocf771cb2008-09-03 09:01:09 +0200257void blkdev_show(struct seq_file *seqf, off_t offset)
Neil Horman7170be52006-01-14 13:20:38 -0800258{
Joe Korty68eef3b2006-03-31 02:30:32 -0800259 struct blk_major_name *dp;
Neil Horman7170be52006-01-14 13:20:38 -0800260
Joe Korty68eef3b2006-03-31 02:30:32 -0800261 if (offset < BLKDEV_MAJOR_HASH_SIZE) {
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200262 mutex_lock(&block_class_lock);
Joe Korty68eef3b2006-03-31 02:30:32 -0800263 for (dp = major_names[offset]; dp; dp = dp->next)
Tejun Heocf771cb2008-09-03 09:01:09 +0200264 seq_printf(seqf, "%3d %s\n", dp->major, dp->name);
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200265 mutex_unlock(&block_class_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700267}
Joe Korty68eef3b2006-03-31 02:30:32 -0800268#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700269
Márton Németh9e8c0bc2009-02-20 08:12:51 +0100270/**
271 * register_blkdev - register a new block device
272 *
273 * @major: the requested major device number [1..255]. If @major=0, try to
274 * allocate any unused major number.
275 * @name: the name of the new block device as a zero terminated string
276 *
277 * The @name must be unique within the system.
278 *
279 * The return value depends on the @major input parameter.
280 * - if a major device number was requested in range [1..255] then the
281 * function returns zero on success, or a negative error code
282 * - if any unused major number was requested with @major=0 parameter
283 * then the return value is the allocated major number in range
284 * [1..255] or a negative error code otherwise
285 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286int register_blkdev(unsigned int major, const char *name)
287{
288 struct blk_major_name **n, *p;
289 int index, ret = 0;
290
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200291 mutex_lock(&block_class_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700292
293 /* temporary */
294 if (major == 0) {
295 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
296 if (major_names[index] == NULL)
297 break;
298 }
299
300 if (index == 0) {
301 printk("register_blkdev: failed to get major for %s\n",
302 name);
303 ret = -EBUSY;
304 goto out;
305 }
306 major = index;
307 ret = major;
308 }
309
310 p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
311 if (p == NULL) {
312 ret = -ENOMEM;
313 goto out;
314 }
315
316 p->major = major;
317 strlcpy(p->name, name, sizeof(p->name));
318 p->next = NULL;
319 index = major_to_index(major);
320
321 for (n = &major_names[index]; *n; n = &(*n)->next) {
322 if ((*n)->major == major)
323 break;
324 }
325 if (!*n)
326 *n = p;
327 else
328 ret = -EBUSY;
329
330 if (ret < 0) {
331 printk("register_blkdev: cannot get major %d for %s\n",
332 major, name);
333 kfree(p);
334 }
335out:
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200336 mutex_unlock(&block_class_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337 return ret;
338}
339
340EXPORT_SYMBOL(register_blkdev);
341
Akinobu Mitaf4480242007-07-17 04:03:47 -0700342void unregister_blkdev(unsigned int major, const char *name)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700343{
344 struct blk_major_name **n;
345 struct blk_major_name *p = NULL;
346 int index = major_to_index(major);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200348 mutex_lock(&block_class_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349 for (n = &major_names[index]; *n; n = &(*n)->next)
350 if ((*n)->major == major)
351 break;
Akinobu Mita294462a2007-07-17 04:03:45 -0700352 if (!*n || strcmp((*n)->name, name)) {
353 WARN_ON(1);
Akinobu Mita294462a2007-07-17 04:03:45 -0700354 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355 p = *n;
356 *n = p->next;
357 }
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200358 mutex_unlock(&block_class_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359 kfree(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360}
361
362EXPORT_SYMBOL(unregister_blkdev);
363
364static struct kobj_map *bdev_map;
365
Tejun Heobcce3de2008-08-25 19:47:22 +0900366/**
Tejun Heo870d6652008-08-25 19:47:25 +0900367 * blk_mangle_minor - scatter minor numbers apart
368 * @minor: minor number to mangle
369 *
370 * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
371 * is enabled. Mangling twice gives the original value.
372 *
373 * RETURNS:
374 * Mangled value.
375 *
376 * CONTEXT:
377 * Don't care.
378 */
379static int blk_mangle_minor(int minor)
380{
381#ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
382 int i;
383
384 for (i = 0; i < MINORBITS / 2; i++) {
385 int low = minor & (1 << i);
386 int high = minor & (1 << (MINORBITS - 1 - i));
387 int distance = MINORBITS - 1 - 2 * i;
388
389 minor ^= low | high; /* clear both bits */
390 low <<= distance; /* swap the positions */
391 high >>= distance;
392 minor |= low | high; /* and set */
393 }
394#endif
395 return minor;
396}
397
398/**
Tejun Heobcce3de2008-08-25 19:47:22 +0900399 * blk_alloc_devt - allocate a dev_t for a partition
400 * @part: partition to allocate dev_t for
Tejun Heobcce3de2008-08-25 19:47:22 +0900401 * @devt: out parameter for resulting dev_t
402 *
403 * Allocate a dev_t for block device.
404 *
405 * RETURNS:
406 * 0 on success, allocated dev_t is returned in *@devt. -errno on
407 * failure.
408 *
409 * CONTEXT:
410 * Might sleep.
411 */
412int blk_alloc_devt(struct hd_struct *part, dev_t *devt)
413{
414 struct gendisk *disk = part_to_disk(part);
Tejun Heobab998d2013-02-27 17:03:57 -0800415 int idx;
Tejun Heobcce3de2008-08-25 19:47:22 +0900416
417 /* in consecutive minor range? */
418 if (part->partno < disk->minors) {
419 *devt = MKDEV(disk->major, disk->first_minor + part->partno);
420 return 0;
421 }
422
423 /* allocate ext devt */
Keith Busch2da78092014-08-26 09:05:36 -0600424 idr_preload(GFP_KERNEL);
425
Dan Williams4d66e5e2015-06-10 23:47:14 -0400426 spin_lock_bh(&ext_devt_lock);
Keith Busch2da78092014-08-26 09:05:36 -0600427 idx = idr_alloc(&ext_devt_idr, part, 0, NR_EXT_DEVT, GFP_NOWAIT);
Dan Williams4d66e5e2015-06-10 23:47:14 -0400428 spin_unlock_bh(&ext_devt_lock);
Keith Busch2da78092014-08-26 09:05:36 -0600429
430 idr_preload_end();
Tejun Heobab998d2013-02-27 17:03:57 -0800431 if (idx < 0)
432 return idx == -ENOSPC ? -EBUSY : idx;
Tejun Heobcce3de2008-08-25 19:47:22 +0900433
Tejun Heo870d6652008-08-25 19:47:25 +0900434 *devt = MKDEV(BLOCK_EXT_MAJOR, blk_mangle_minor(idx));
Tejun Heobcce3de2008-08-25 19:47:22 +0900435 return 0;
436}
437
438/**
439 * blk_free_devt - free a dev_t
440 * @devt: dev_t to free
441 *
442 * Free @devt which was allocated using blk_alloc_devt().
443 *
444 * CONTEXT:
445 * Might sleep.
446 */
447void blk_free_devt(dev_t devt)
448{
Tejun Heobcce3de2008-08-25 19:47:22 +0900449 if (devt == MKDEV(0, 0))
450 return;
451
452 if (MAJOR(devt) == BLOCK_EXT_MAJOR) {
Dan Williams4d66e5e2015-06-10 23:47:14 -0400453 spin_lock_bh(&ext_devt_lock);
Tejun Heo870d6652008-08-25 19:47:25 +0900454 idr_remove(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
Dan Williams4d66e5e2015-06-10 23:47:14 -0400455 spin_unlock_bh(&ext_devt_lock);
Tejun Heobcce3de2008-08-25 19:47:22 +0900456 }
457}
458
Tejun Heo1f014292008-08-25 19:47:23 +0900459static char *bdevt_str(dev_t devt, char *buf)
460{
461 if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
462 char tbuf[BDEVT_SIZE];
463 snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
464 snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
465 } else
466 snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
467
468 return buf;
469}
470
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471/*
472 * Register device numbers dev..(dev+range-1)
473 * range must be nonzero
474 * The hash chain is sorted on range, so that subranges can override.
475 */
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200476void blk_register_region(dev_t devt, unsigned long range, struct module *module,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477 struct kobject *(*probe)(dev_t, int *, void *),
478 int (*lock)(dev_t, void *), void *data)
479{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200480 kobj_map(bdev_map, devt, range, module, probe, lock, data);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481}
482
483EXPORT_SYMBOL(blk_register_region);
484
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200485void blk_unregister_region(dev_t devt, unsigned long range)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200487 kobj_unmap(bdev_map, devt, range);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488}
489
490EXPORT_SYMBOL(blk_unregister_region);
491
Tejun Heocf771cb2008-09-03 09:01:09 +0200492static struct kobject *exact_match(dev_t devt, int *partno, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493{
494 struct gendisk *p = data;
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200495
Tejun Heoed9e1982008-08-25 19:56:05 +0900496 return &disk_to_dev(p)->kobj;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497}
498
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200499static int exact_lock(dev_t devt, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700500{
501 struct gendisk *p = data;
502
503 if (!get_disk(p))
504 return -1;
505 return 0;
506}
507
Al Viro4752bc32011-09-16 00:21:54 -0400508static void register_disk(struct gendisk *disk)
Tejun Heod2bf1b62010-12-08 20:57:36 +0100509{
510 struct device *ddev = disk_to_dev(disk);
511 struct block_device *bdev;
512 struct disk_part_iter piter;
513 struct hd_struct *part;
514 int err;
515
516 ddev->parent = disk->driverfs_dev;
517
Kees Cookffc8b302013-07-03 15:01:14 -0700518 dev_set_name(ddev, "%s", disk->disk_name);
Tejun Heod2bf1b62010-12-08 20:57:36 +0100519
520 /* delay uevents, until we scanned partition table */
521 dev_set_uevent_suppress(ddev, 1);
522
523 if (device_add(ddev))
524 return;
525 if (!sysfs_deprecated) {
526 err = sysfs_create_link(block_depr, &ddev->kobj,
527 kobject_name(&ddev->kobj));
528 if (err) {
529 device_del(ddev);
530 return;
531 }
532 }
Ming Lei25e823c2013-02-22 16:34:13 -0800533
534 /*
535 * avoid probable deadlock caused by allocating memory with
536 * GFP_KERNEL in runtime_resume callback of its all ancestor
537 * devices
538 */
539 pm_runtime_set_memalloc_noio(ddev, true);
540
Tejun Heod2bf1b62010-12-08 20:57:36 +0100541 disk->part0.holder_dir = kobject_create_and_add("holders", &ddev->kobj);
542 disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
543
544 /* No minors to use for partitions */
Tejun Heod27769e2011-08-23 20:01:04 +0200545 if (!disk_part_scan_enabled(disk))
Tejun Heod2bf1b62010-12-08 20:57:36 +0100546 goto exit;
547
548 /* No such device (e.g., media were just removed) */
549 if (!get_capacity(disk))
550 goto exit;
551
552 bdev = bdget_disk(disk, 0);
553 if (!bdev)
554 goto exit;
555
556 bdev->bd_invalidated = 1;
557 err = blkdev_get(bdev, FMODE_READ, NULL);
558 if (err < 0)
559 goto exit;
560 blkdev_put(bdev, FMODE_READ);
561
562exit:
563 /* announce disk after possible partitions are created */
564 dev_set_uevent_suppress(ddev, 0);
565 kobject_uevent(&ddev->kobj, KOBJ_ADD);
566
567 /* announce possible partitions */
568 disk_part_iter_init(&piter, disk, 0);
569 while ((part = disk_part_iter_next(&piter)))
570 kobject_uevent(&part_to_dev(part)->kobj, KOBJ_ADD);
571 disk_part_iter_exit(&piter);
572}
573
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574/**
575 * add_disk - add partitioning information to kernel list
576 * @disk: per-device partitioning information
577 *
578 * This function registers the partitioning information in @disk
579 * with the kernel.
Tejun Heo3e1a7ff2008-08-25 19:56:17 +0900580 *
581 * FIXME: error handling
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582 */
583void add_disk(struct gendisk *disk)
584{
Peter Zijlstracf0ca9f2008-04-30 00:54:32 -0700585 struct backing_dev_info *bdi;
Tejun Heo3e1a7ff2008-08-25 19:56:17 +0900586 dev_t devt;
Greg Kroah-Hartman6ffeea72008-05-22 17:21:08 -0400587 int retval;
Peter Zijlstracf0ca9f2008-04-30 00:54:32 -0700588
Tejun Heo3e1a7ff2008-08-25 19:56:17 +0900589 /* minors == 0 indicates to use ext devt from part0 and should
590 * be accompanied with EXT_DEVT flag. Make sure all
591 * parameters make sense.
592 */
593 WARN_ON(disk->minors && !(disk->major || disk->first_minor));
594 WARN_ON(!disk->minors && !(disk->flags & GENHD_FL_EXT_DEVT));
595
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596 disk->flags |= GENHD_FL_UP;
Tejun Heo3e1a7ff2008-08-25 19:56:17 +0900597
598 retval = blk_alloc_devt(&disk->part0, &devt);
599 if (retval) {
600 WARN_ON(1);
601 return;
602 }
603 disk_to_dev(disk)->devt = devt;
604
605 /* ->major and ->first_minor aren't supposed to be
606 * dereferenced from here on, but set them just in case.
607 */
608 disk->major = MAJOR(devt);
609 disk->first_minor = MINOR(devt);
610
Stanislaw Gruszka9f53d2fe2012-03-02 10:43:28 +0100611 disk_alloc_events(disk);
612
Wanlong Gao9f5e4862011-06-13 10:45:43 +0200613 /* Register BDI before referencing it from bdev */
Signed-off-by: Jan Kara01ea5062010-09-16 20:36:36 +0200614 bdi = &disk->queue->backing_dev_info;
615 bdi_register_dev(bdi, disk_devt(disk));
616
Tejun Heof331c022008-09-03 09:01:48 +0200617 blk_register_region(disk_devt(disk), disk->minors, NULL,
618 exact_match, exact_lock, disk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700619 register_disk(disk);
620 blk_register_queue(disk);
Peter Zijlstracf0ca9f2008-04-30 00:54:32 -0700621
Tejun Heo523e1d32011-10-19 14:31:07 +0200622 /*
623 * Take an extra ref on queue which will be put on disk_release()
624 * so that it sticks around as long as @disk is there.
625 */
Tejun Heo09ac46c2011-12-14 00:33:38 +0100626 WARN_ON_ONCE(!blk_get_queue(disk->queue));
Tejun Heo523e1d32011-10-19 14:31:07 +0200627
Tejun Heoed9e1982008-08-25 19:56:05 +0900628 retval = sysfs_create_link(&disk_to_dev(disk)->kobj, &bdi->dev->kobj,
629 "bdi");
Greg Kroah-Hartman6ffeea72008-05-22 17:21:08 -0400630 WARN_ON(retval);
Tejun Heo77ea8872010-12-08 20:57:37 +0100631
632 disk_add_events(disk);
Martin K. Petersen25520d52015-10-21 13:19:49 -0400633 blk_integrity_add(disk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700635EXPORT_SYMBOL(add_disk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636
Tejun Heod2bf1b62010-12-08 20:57:36 +0100637void del_gendisk(struct gendisk *disk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638{
Tejun Heod2bf1b62010-12-08 20:57:36 +0100639 struct disk_part_iter piter;
640 struct hd_struct *part;
641
Martin K. Petersen25520d52015-10-21 13:19:49 -0400642 blk_integrity_del(disk);
Tejun Heo77ea8872010-12-08 20:57:37 +0100643 disk_del_events(disk);
644
Tejun Heod2bf1b62010-12-08 20:57:36 +0100645 /* invalidate stuff */
646 disk_part_iter_init(&piter, disk,
647 DISK_PITER_INCL_EMPTY | DISK_PITER_REVERSE);
648 while ((part = disk_part_iter_next(&piter))) {
649 invalidate_partition(disk, part->partno);
650 delete_partition(disk, part->partno);
651 }
652 disk_part_iter_exit(&piter);
653
654 invalidate_partition(disk, 0);
Tejun Heod2bf1b62010-12-08 20:57:36 +0100655 set_capacity(disk, 0);
656 disk->flags &= ~GENHD_FL_UP;
657
Tejun Heoed9e1982008-08-25 19:56:05 +0900658 sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659 blk_unregister_queue(disk);
Tejun Heof331c022008-09-03 09:01:48 +0200660 blk_unregister_region(disk_devt(disk), disk->minors);
Tejun Heod2bf1b62010-12-08 20:57:36 +0100661
662 part_stat_set_all(&disk->part0, 0);
663 disk->part0.stamp = 0;
664
665 kobject_put(disk->part0.holder_dir);
666 kobject_put(disk->slave_dir);
667 disk->driverfs_dev = NULL;
668 if (!sysfs_deprecated)
669 sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
Ming Lei25e823c2013-02-22 16:34:13 -0800670 pm_runtime_set_memalloc_noio(disk_to_dev(disk), false);
Tejun Heod2bf1b62010-12-08 20:57:36 +0100671 device_del(disk_to_dev(disk));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700672}
Tejun Heod2bf1b62010-12-08 20:57:36 +0100673EXPORT_SYMBOL(del_gendisk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675/**
676 * get_gendisk - get partitioning information for a given device
Randy Dunlap710027a2008-08-19 20:13:11 +0200677 * @devt: device to get partitioning information for
Randy Dunlap496aa8a2008-10-16 07:46:23 +0200678 * @partno: returned partition index
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679 *
680 * This function gets the structure containing partitioning
Randy Dunlap710027a2008-08-19 20:13:11 +0200681 * information for the given device @devt.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682 */
Tejun Heocf771cb2008-09-03 09:01:09 +0200683struct gendisk *get_gendisk(dev_t devt, int *partno)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684{
Tejun Heobcce3de2008-08-25 19:47:22 +0900685 struct gendisk *disk = NULL;
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200686
Tejun Heobcce3de2008-08-25 19:47:22 +0900687 if (MAJOR(devt) != BLOCK_EXT_MAJOR) {
688 struct kobject *kobj;
689
690 kobj = kobj_lookup(bdev_map, devt, partno);
691 if (kobj)
692 disk = dev_to_disk(kobj_to_dev(kobj));
693 } else {
694 struct hd_struct *part;
695
Dan Williams4d66e5e2015-06-10 23:47:14 -0400696 spin_lock_bh(&ext_devt_lock);
Tejun Heo870d6652008-08-25 19:47:25 +0900697 part = idr_find(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
Tejun Heobcce3de2008-08-25 19:47:22 +0900698 if (part && get_disk(part_to_disk(part))) {
699 *partno = part->partno;
700 disk = part_to_disk(part);
701 }
Dan Williams4d66e5e2015-06-10 23:47:14 -0400702 spin_unlock_bh(&ext_devt_lock);
Tejun Heobcce3de2008-08-25 19:47:22 +0900703 }
704
705 return disk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706}
Divyesh Shahb6ac23af2010-04-15 08:54:59 +0200707EXPORT_SYMBOL(get_gendisk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708
Tejun Heof331c022008-09-03 09:01:48 +0200709/**
710 * bdget_disk - do bdget() by gendisk and partition number
711 * @disk: gendisk of interest
712 * @partno: partition number
713 *
714 * Find partition @partno from @disk, do bdget() on it.
715 *
716 * CONTEXT:
717 * Don't care.
718 *
719 * RETURNS:
720 * Resulting block_device on success, NULL on failure.
721 */
Harvey Harrisonaeb3d3a2008-08-28 09:27:42 +0200722struct block_device *bdget_disk(struct gendisk *disk, int partno)
Tejun Heof331c022008-09-03 09:01:48 +0200723{
Tejun Heo548b10e2008-08-29 09:01:47 +0200724 struct hd_struct *part;
725 struct block_device *bdev = NULL;
Tejun Heof331c022008-09-03 09:01:48 +0200726
Tejun Heo548b10e2008-08-29 09:01:47 +0200727 part = disk_get_part(disk, partno);
Tejun Heo2bbedcb2008-08-29 11:41:51 +0200728 if (part)
Tejun Heo548b10e2008-08-29 09:01:47 +0200729 bdev = bdget(part_devt(part));
730 disk_put_part(part);
Tejun Heof331c022008-09-03 09:01:48 +0200731
Tejun Heo548b10e2008-08-29 09:01:47 +0200732 return bdev;
Tejun Heof331c022008-09-03 09:01:48 +0200733}
734EXPORT_SYMBOL(bdget_disk);
735
Dave Gilbertdd2a3452007-05-09 02:33:24 -0700736/*
737 * print a full list of all partitions - intended for places where the root
738 * filesystem can't be mounted and thus to give the victim some idea of what
739 * went wrong
740 */
741void __init printk_all_partitions(void)
742{
Tejun Heodef4e382008-09-03 08:57:12 +0200743 struct class_dev_iter iter;
744 struct device *dev;
745
746 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
747 while ((dev = class_dev_iter_next(&iter))) {
748 struct gendisk *disk = dev_to_disk(dev);
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200749 struct disk_part_iter piter;
750 struct hd_struct *part;
Tejun Heo1f014292008-08-25 19:47:23 +0900751 char name_buf[BDEVNAME_SIZE];
752 char devt_buf[BDEVT_SIZE];
Tejun Heodef4e382008-09-03 08:57:12 +0200753
754 /*
755 * Don't show empty devices or things that have been
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300756 * suppressed
Tejun Heodef4e382008-09-03 08:57:12 +0200757 */
758 if (get_capacity(disk) == 0 ||
759 (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
760 continue;
761
762 /*
763 * Note, unlike /proc/partitions, I am showing the
764 * numbers in hex - the same format as the root=
765 * option takes.
766 */
Tejun Heo074a7ac2008-08-25 19:56:14 +0900767 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0);
768 while ((part = disk_part_iter_next(&piter))) {
769 bool is_part0 = part == &disk->part0;
Tejun Heodef4e382008-09-03 08:57:12 +0200770
Will Drewryb5af9212010-08-31 15:47:07 -0500771 printk("%s%s %10llu %s %s", is_part0 ? "" : " ",
Tejun Heo1f014292008-08-25 19:47:23 +0900772 bdevt_str(part_devt(part), devt_buf),
Vivek Goyalc83f6bf2012-08-01 12:24:18 +0200773 (unsigned long long)part_nr_sects_read(part) >> 1
774 , disk_name(disk, part->partno, name_buf),
Stephen Warren1ad7e892012-11-08 16:12:25 -0800775 part->info ? part->info->uuid : "");
Tejun Heo074a7ac2008-08-25 19:56:14 +0900776 if (is_part0) {
777 if (disk->driverfs_dev != NULL &&
778 disk->driverfs_dev->driver != NULL)
779 printk(" driver: %s\n",
780 disk->driverfs_dev->driver->name);
781 else
782 printk(" (driver?)\n");
783 } else
784 printk("\n");
785 }
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200786 disk_part_iter_exit(&piter);
Tejun Heodef4e382008-09-03 08:57:12 +0200787 }
788 class_dev_iter_exit(&iter);
Dave Gilbertdd2a3452007-05-09 02:33:24 -0700789}
790
Linus Torvalds1da177e2005-04-16 15:20:36 -0700791#ifdef CONFIG_PROC_FS
792/* iterator */
Tejun Heodef4e382008-09-03 08:57:12 +0200793static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
Greg Kroah-Hartman68c4d4a2008-05-22 17:21:08 -0400794{
Tejun Heodef4e382008-09-03 08:57:12 +0200795 loff_t skip = *pos;
796 struct class_dev_iter *iter;
797 struct device *dev;
Greg Kroah-Hartman68c4d4a2008-05-22 17:21:08 -0400798
Harvey Harrisonaeb3d3a2008-08-28 09:27:42 +0200799 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
Tejun Heodef4e382008-09-03 08:57:12 +0200800 if (!iter)
801 return ERR_PTR(-ENOMEM);
802
803 seqf->private = iter;
804 class_dev_iter_init(iter, &block_class, NULL, &disk_type);
805 do {
806 dev = class_dev_iter_next(iter);
807 if (!dev)
808 return NULL;
809 } while (skip--);
810
811 return dev_to_disk(dev);
Greg Kroah-Hartman68c4d4a2008-05-22 17:21:08 -0400812}
813
Tejun Heodef4e382008-09-03 08:57:12 +0200814static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200816 struct device *dev;
Greg Kroah-Hartman66c64af2008-05-22 17:21:08 -0400817
Tejun Heodef4e382008-09-03 08:57:12 +0200818 (*pos)++;
819 dev = class_dev_iter_next(seqf->private);
Tejun Heo2ac3cee2008-09-03 08:53:37 +0200820 if (dev)
Greg Kroah-Hartman68c4d4a2008-05-22 17:21:08 -0400821 return dev_to_disk(dev);
Tejun Heo2ac3cee2008-09-03 08:53:37 +0200822
Linus Torvalds1da177e2005-04-16 15:20:36 -0700823 return NULL;
824}
825
Tejun Heodef4e382008-09-03 08:57:12 +0200826static void disk_seqf_stop(struct seq_file *seqf, void *v)
Greg Kroah-Hartman27f30252008-05-22 17:21:08 -0400827{
Tejun Heodef4e382008-09-03 08:57:12 +0200828 struct class_dev_iter *iter = seqf->private;
Greg Kroah-Hartman27f30252008-05-22 17:21:08 -0400829
Tejun Heodef4e382008-09-03 08:57:12 +0200830 /* stop is called even after start failed :-( */
831 if (iter) {
832 class_dev_iter_exit(iter);
833 kfree(iter);
Kay Sievers5c0ef6d2008-08-16 14:30:30 +0200834 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835}
836
Tejun Heodef4e382008-09-03 08:57:12 +0200837static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838{
Jianpeng Ma06768062012-08-03 10:42:00 +0200839 void *p;
Tejun Heodef4e382008-09-03 08:57:12 +0200840
841 p = disk_seqf_start(seqf, pos);
Yang Zhangb9f985b2010-12-17 08:58:36 +0100842 if (!IS_ERR_OR_NULL(p) && !*pos)
Tejun Heodef4e382008-09-03 08:57:12 +0200843 seq_puts(seqf, "major minor #blocks name\n\n");
844 return p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845}
846
Tejun Heocf771cb2008-09-03 09:01:09 +0200847static int show_partition(struct seq_file *seqf, void *v)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700848{
849 struct gendisk *sgp = v;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200850 struct disk_part_iter piter;
851 struct hd_struct *part;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700852 char buf[BDEVNAME_SIZE];
853
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854 /* Don't show non-partitionable removeable devices or empty devices */
Tejun Heod27769e2011-08-23 20:01:04 +0200855 if (!get_capacity(sgp) || (!disk_max_parts(sgp) &&
Tejun Heof331c022008-09-03 09:01:48 +0200856 (sgp->flags & GENHD_FL_REMOVABLE)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857 return 0;
858 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
859 return 0;
860
861 /* show the full disk and all non-0 size partitions of it */
Tejun Heo074a7ac2008-08-25 19:56:14 +0900862 disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0);
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200863 while ((part = disk_part_iter_next(&piter)))
Tejun Heo1f014292008-08-25 19:47:23 +0900864 seq_printf(seqf, "%4d %7d %10llu %s\n",
Tejun Heof331c022008-09-03 09:01:48 +0200865 MAJOR(part_devt(part)), MINOR(part_devt(part)),
Vivek Goyalc83f6bf2012-08-01 12:24:18 +0200866 (unsigned long long)part_nr_sects_read(part) >> 1,
Tejun Heof331c022008-09-03 09:01:48 +0200867 disk_name(sgp, part->partno, buf));
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200868 disk_part_iter_exit(&piter);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869
870 return 0;
871}
872
Alexey Dobriyanf5009752008-10-04 23:53:21 +0400873static const struct seq_operations partitions_op = {
Tejun Heodef4e382008-09-03 08:57:12 +0200874 .start = show_partition_start,
875 .next = disk_seqf_next,
876 .stop = disk_seqf_stop,
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200877 .show = show_partition
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878};
Alexey Dobriyanf5009752008-10-04 23:53:21 +0400879
880static int partitions_open(struct inode *inode, struct file *file)
881{
882 return seq_open(file, &partitions_op);
883}
884
885static const struct file_operations proc_partitions_operations = {
886 .open = partitions_open,
887 .read = seq_read,
888 .llseek = seq_lseek,
889 .release = seq_release,
890};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891#endif
892
893
Tejun Heocf771cb2008-09-03 09:01:09 +0200894static struct kobject *base_probe(dev_t devt, int *partno, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200896 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700897 /* Make old-style 2.4 aliases work */
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200898 request_module("block-major-%d", MAJOR(devt));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899 return NULL;
900}
901
902static int __init genhd_device_init(void)
903{
Dan Williamse105b8b2008-04-21 10:51:07 -0700904 int error;
905
906 block_class.dev_kobj = sysfs_dev_block_kobj;
907 error = class_register(&block_class);
Roland McGrathee27a552008-03-11 17:13:15 -0700908 if (unlikely(error))
909 return error;
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200910 bdev_map = kobj_map_init(base_probe, &block_class_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700911 blk_dev_init();
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200912
Zhang, Yanmin561ec682008-11-14 08:26:30 +0100913 register_blkdev(BLOCK_EXT_MAJOR, "blkext");
914
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200915 /* create top-level block dir */
Andi Kleene52eec12010-09-08 16:54:17 +0200916 if (!sysfs_deprecated)
917 block_depr = kobject_create_and_add("block", NULL);
Greg Kroah-Hartman830d3cf2007-11-06 10:36:58 -0800918 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700919}
920
921subsys_initcall(genhd_device_init);
922
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200923static ssize_t disk_range_show(struct device *dev,
924 struct device_attribute *attr, char *buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700925{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200926 struct gendisk *disk = dev_to_disk(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200928 return sprintf(buf, "%d\n", disk->minors);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929}
930
Tejun Heo1f014292008-08-25 19:47:23 +0900931static ssize_t disk_ext_range_show(struct device *dev,
932 struct device_attribute *attr, char *buf)
933{
934 struct gendisk *disk = dev_to_disk(dev);
935
Tejun Heob5d0b9d2008-09-03 09:06:42 +0200936 return sprintf(buf, "%d\n", disk_max_parts(disk));
Tejun Heo1f014292008-08-25 19:47:23 +0900937}
938
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200939static ssize_t disk_removable_show(struct device *dev,
940 struct device_attribute *attr, char *buf)
Kay Sieversa7fd6702005-10-01 14:49:43 +0200941{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200942 struct gendisk *disk = dev_to_disk(dev);
Kay Sieversa7fd6702005-10-01 14:49:43 +0200943
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200944 return sprintf(buf, "%d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700945 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200946}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700947
Kay Sievers1c9ce522008-06-13 09:41:00 +0200948static ssize_t disk_ro_show(struct device *dev,
949 struct device_attribute *attr, char *buf)
950{
951 struct gendisk *disk = dev_to_disk(dev);
952
Tejun Heob7db9952008-08-25 19:56:10 +0900953 return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
Kay Sievers1c9ce522008-06-13 09:41:00 +0200954}
955
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200956static ssize_t disk_capability_show(struct device *dev,
957 struct device_attribute *attr, char *buf)
Kristen Carlson Accardi86ce18d2007-05-23 13:57:38 -0700958{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200959 struct gendisk *disk = dev_to_disk(dev);
960
961 return sprintf(buf, "%x\n", disk->flags);
Kristen Carlson Accardi86ce18d2007-05-23 13:57:38 -0700962}
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200963
Martin K. Petersenc72758f2009-05-22 17:17:53 -0400964static ssize_t disk_alignment_offset_show(struct device *dev,
965 struct device_attribute *attr,
966 char *buf)
967{
968 struct gendisk *disk = dev_to_disk(dev);
969
970 return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue));
971}
972
Martin K. Petersen86b37282009-11-10 11:50:21 +0100973static ssize_t disk_discard_alignment_show(struct device *dev,
974 struct device_attribute *attr,
975 char *buf)
976{
977 struct gendisk *disk = dev_to_disk(dev);
978
Martin K. Petersendd3d1452010-01-11 03:21:48 -0500979 return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue));
Martin K. Petersen86b37282009-11-10 11:50:21 +0100980}
981
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200982static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL);
Tejun Heo1f014292008-08-25 19:47:23 +0900983static DEVICE_ATTR(ext_range, S_IRUGO, disk_ext_range_show, NULL);
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200984static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL);
Kay Sievers1c9ce522008-06-13 09:41:00 +0200985static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL);
Tejun Heoe5610522008-08-25 19:56:09 +0900986static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL);
Martin K. Petersenc72758f2009-05-22 17:17:53 -0400987static DEVICE_ATTR(alignment_offset, S_IRUGO, disk_alignment_offset_show, NULL);
Martin K. Petersen86b37282009-11-10 11:50:21 +0100988static DEVICE_ATTR(discard_alignment, S_IRUGO, disk_discard_alignment_show,
989 NULL);
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200990static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL);
Tejun Heo074a7ac2008-08-25 19:56:14 +0900991static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL);
Nikanth Karthikesan316d3152009-10-06 20:16:55 +0200992static DEVICE_ATTR(inflight, S_IRUGO, part_inflight_show, NULL);
Akinobu Mitac17bb492006-12-08 02:39:46 -0800993#ifdef CONFIG_FAIL_MAKE_REQUEST
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200994static struct device_attribute dev_attr_fail =
Tejun Heoeddb2e22008-08-25 19:56:13 +0900995 __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store);
Akinobu Mitac17bb492006-12-08 02:39:46 -0800996#endif
Jens Axboe581d4e22008-09-14 05:56:33 -0700997#ifdef CONFIG_FAIL_IO_TIMEOUT
998static struct device_attribute dev_attr_fail_timeout =
999 __ATTR(io-timeout-fail, S_IRUGO|S_IWUSR, part_timeout_show,
1000 part_timeout_store);
1001#endif
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001002
1003static struct attribute *disk_attrs[] = {
1004 &dev_attr_range.attr,
Tejun Heo1f014292008-08-25 19:47:23 +09001005 &dev_attr_ext_range.attr,
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001006 &dev_attr_removable.attr,
Kay Sievers1c9ce522008-06-13 09:41:00 +02001007 &dev_attr_ro.attr,
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001008 &dev_attr_size.attr,
Martin K. Petersenc72758f2009-05-22 17:17:53 -04001009 &dev_attr_alignment_offset.attr,
Martin K. Petersen86b37282009-11-10 11:50:21 +01001010 &dev_attr_discard_alignment.attr,
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001011 &dev_attr_capability.attr,
1012 &dev_attr_stat.attr,
Nikanth Karthikesan316d3152009-10-06 20:16:55 +02001013 &dev_attr_inflight.attr,
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001014#ifdef CONFIG_FAIL_MAKE_REQUEST
1015 &dev_attr_fail.attr,
1016#endif
Jens Axboe581d4e22008-09-14 05:56:33 -07001017#ifdef CONFIG_FAIL_IO_TIMEOUT
1018 &dev_attr_fail_timeout.attr,
1019#endif
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001020 NULL
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021};
1022
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001023static struct attribute_group disk_attr_group = {
1024 .attrs = disk_attrs,
1025};
1026
David Brownella4dbd672009-06-24 10:06:31 -07001027static const struct attribute_group *disk_attr_groups[] = {
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001028 &disk_attr_group,
1029 NULL
1030};
1031
Tejun Heo540eed52008-08-25 19:56:15 +09001032/**
1033 * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way
1034 * @disk: disk to replace part_tbl for
1035 * @new_ptbl: new part_tbl to install
1036 *
1037 * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The
1038 * original ptbl is freed using RCU callback.
1039 *
1040 * LOCKING:
1041 * Matching bd_mutx locked.
1042 */
1043static void disk_replace_part_tbl(struct gendisk *disk,
1044 struct disk_part_tbl *new_ptbl)
1045{
1046 struct disk_part_tbl *old_ptbl = disk->part_tbl;
1047
1048 rcu_assign_pointer(disk->part_tbl, new_ptbl);
Jens Axboea6f23652008-10-24 12:52:42 +02001049
1050 if (old_ptbl) {
1051 rcu_assign_pointer(old_ptbl->last_lookup, NULL);
Lai Jiangshan57bdfbf2011-03-18 11:42:58 +08001052 kfree_rcu(old_ptbl, rcu_head);
Jens Axboea6f23652008-10-24 12:52:42 +02001053 }
Tejun Heo540eed52008-08-25 19:56:15 +09001054}
1055
1056/**
1057 * disk_expand_part_tbl - expand disk->part_tbl
1058 * @disk: disk to expand part_tbl for
1059 * @partno: expand such that this partno can fit in
1060 *
1061 * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl
1062 * uses RCU to allow unlocked dereferencing for stats and other stuff.
1063 *
1064 * LOCKING:
1065 * Matching bd_mutex locked, might sleep.
1066 *
1067 * RETURNS:
1068 * 0 on success, -errno on failure.
1069 */
1070int disk_expand_part_tbl(struct gendisk *disk, int partno)
1071{
1072 struct disk_part_tbl *old_ptbl = disk->part_tbl;
1073 struct disk_part_tbl *new_ptbl;
1074 int len = old_ptbl ? old_ptbl->len : 0;
Jens Axboe5fabcb42014-11-19 13:06:22 -07001075 int i, target;
Tejun Heo540eed52008-08-25 19:56:15 +09001076 size_t size;
Jens Axboe5fabcb42014-11-19 13:06:22 -07001077
1078 /*
1079 * check for int overflow, since we can get here from blkpg_ioctl()
1080 * with a user passed 'partno'.
1081 */
1082 target = partno + 1;
1083 if (target < 0)
1084 return -EINVAL;
Tejun Heo540eed52008-08-25 19:56:15 +09001085
1086 /* disk_max_parts() is zero during initialization, ignore if so */
1087 if (disk_max_parts(disk) && target > disk_max_parts(disk))
1088 return -EINVAL;
1089
1090 if (target <= len)
1091 return 0;
1092
1093 size = sizeof(*new_ptbl) + target * sizeof(new_ptbl->part[0]);
1094 new_ptbl = kzalloc_node(size, GFP_KERNEL, disk->node_id);
1095 if (!new_ptbl)
1096 return -ENOMEM;
1097
Tejun Heo540eed52008-08-25 19:56:15 +09001098 new_ptbl->len = target;
1099
1100 for (i = 0; i < len; i++)
1101 rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]);
1102
1103 disk_replace_part_tbl(disk, new_ptbl);
1104 return 0;
1105}
1106
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001107static void disk_release(struct device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108{
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001109 struct gendisk *disk = dev_to_disk(dev);
1110
Keith Busch2da78092014-08-26 09:05:36 -06001111 blk_free_devt(dev->devt);
Tejun Heo77ea8872010-12-08 20:57:37 +01001112 disk_release_events(disk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001113 kfree(disk->random);
Tejun Heo540eed52008-08-25 19:56:15 +09001114 disk_replace_part_tbl(disk, NULL);
Ming Leib54e5ed2015-07-16 11:16:44 +08001115 hd_free_part(&disk->part0);
Tejun Heo523e1d32011-10-19 14:31:07 +02001116 if (disk->queue)
1117 blk_put_queue(disk->queue);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001118 kfree(disk);
1119}
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001120struct class block_class = {
1121 .name = "block",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001122};
1123
Kay Sievers3c2670e2013-04-06 09:56:00 -07001124static char *block_devnode(struct device *dev, umode_t *mode,
Greg Kroah-Hartman4e4098a2013-04-11 11:43:29 -07001125 kuid_t *uid, kgid_t *gid)
Kay Sieversb03f38b2009-04-30 15:23:42 +02001126{
1127 struct gendisk *disk = dev_to_disk(dev);
1128
Kay Sieverse454cea2009-09-18 23:01:12 +02001129 if (disk->devnode)
1130 return disk->devnode(disk, mode);
Kay Sieversb03f38b2009-04-30 15:23:42 +02001131 return NULL;
1132}
1133
Adrian Bunk1826ead2008-03-04 11:23:46 +01001134static struct device_type disk_type = {
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001135 .name = "disk",
1136 .groups = disk_attr_groups,
1137 .release = disk_release,
Kay Sieverse454cea2009-09-18 23:01:12 +02001138 .devnode = block_devnode,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139};
1140
Randy Dunlapa6e2ba82008-05-23 09:44:11 -07001141#ifdef CONFIG_PROC_FS
Tejun Heocf771cb2008-09-03 09:01:09 +02001142/*
1143 * aggregate disk stat collector. Uses the same stats that the sysfs
1144 * entries do, above, but makes them available through one seq_file.
1145 *
1146 * The output looks suspiciously like /proc/partitions with a bunch of
1147 * extra fields.
1148 */
1149static int diskstats_show(struct seq_file *seqf, void *v)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001150{
1151 struct gendisk *gp = v;
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001152 struct disk_part_iter piter;
1153 struct hd_struct *hd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001154 char buf[BDEVNAME_SIZE];
Tejun Heoc9959052008-08-25 19:47:21 +09001155 int cpu;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001156
1157 /*
Tejun Heoed9e1982008-08-25 19:56:05 +09001158 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
Tejun Heocf771cb2008-09-03 09:01:09 +02001159 seq_puts(seqf, "major minor name"
Linus Torvalds1da177e2005-04-16 15:20:36 -07001160 " rio rmerge rsect ruse wio wmerge "
1161 "wsect wuse running use aveq"
1162 "\n\n");
1163 */
Wanlong Gao9f5e4862011-06-13 10:45:43 +02001164
Tejun Heo71982a42009-04-17 08:34:48 +02001165 disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0);
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001166 while ((hd = disk_part_iter_next(&piter))) {
Tejun Heo074a7ac2008-08-25 19:56:14 +09001167 cpu = part_stat_lock();
Tejun Heoc9959052008-08-25 19:47:21 +09001168 part_round_stats(cpu, hd);
Tejun Heo074a7ac2008-08-25 19:56:14 +09001169 part_stat_unlock();
Herbert Poetzlf95fe9c2011-08-02 12:43:50 +02001170 seq_printf(seqf, "%4d %7d %s %lu %lu %lu "
1171 "%u %lu %lu %lu %u %u %u %u\n",
Tejun Heof331c022008-09-03 09:01:48 +02001172 MAJOR(part_devt(hd)), MINOR(part_devt(hd)),
1173 disk_name(gp, hd->partno, buf),
Liu Yuan53f22952011-03-02 11:00:15 -05001174 part_stat_read(hd, ios[READ]),
1175 part_stat_read(hd, merges[READ]),
Herbert Poetzlf95fe9c2011-08-02 12:43:50 +02001176 part_stat_read(hd, sectors[READ]),
Liu Yuan53f22952011-03-02 11:00:15 -05001177 jiffies_to_msecs(part_stat_read(hd, ticks[READ])),
1178 part_stat_read(hd, ios[WRITE]),
1179 part_stat_read(hd, merges[WRITE]),
Herbert Poetzlf95fe9c2011-08-02 12:43:50 +02001180 part_stat_read(hd, sectors[WRITE]),
Liu Yuan53f22952011-03-02 11:00:15 -05001181 jiffies_to_msecs(part_stat_read(hd, ticks[WRITE])),
Nikanth Karthikesan316d3152009-10-06 20:16:55 +02001182 part_in_flight(hd),
Jerome Marchand28f39d52008-02-08 11:04:56 +01001183 jiffies_to_msecs(part_stat_read(hd, io_ticks)),
1184 jiffies_to_msecs(part_stat_read(hd, time_in_queue))
1185 );
Linus Torvalds1da177e2005-04-16 15:20:36 -07001186 }
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001187 disk_part_iter_exit(&piter);
Wanlong Gao9f5e4862011-06-13 10:45:43 +02001188
Linus Torvalds1da177e2005-04-16 15:20:36 -07001189 return 0;
1190}
1191
Alexey Dobriyan31d85ab2008-10-06 12:55:38 +04001192static const struct seq_operations diskstats_op = {
Tejun Heodef4e382008-09-03 08:57:12 +02001193 .start = disk_seqf_start,
1194 .next = disk_seqf_next,
1195 .stop = disk_seqf_stop,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196 .show = diskstats_show
1197};
Alexey Dobriyanf5009752008-10-04 23:53:21 +04001198
Alexey Dobriyan31d85ab2008-10-06 12:55:38 +04001199static int diskstats_open(struct inode *inode, struct file *file)
1200{
1201 return seq_open(file, &diskstats_op);
1202}
1203
1204static const struct file_operations proc_diskstats_operations = {
1205 .open = diskstats_open,
1206 .read = seq_read,
1207 .llseek = seq_lseek,
1208 .release = seq_release,
1209};
1210
Alexey Dobriyanf5009752008-10-04 23:53:21 +04001211static int __init proc_genhd_init(void)
1212{
Alexey Dobriyan31d85ab2008-10-06 12:55:38 +04001213 proc_create("diskstats", 0, NULL, &proc_diskstats_operations);
Alexey Dobriyanf5009752008-10-04 23:53:21 +04001214 proc_create("partitions", 0, NULL, &proc_partitions_operations);
1215 return 0;
1216}
1217module_init(proc_genhd_init);
Randy Dunlapa6e2ba82008-05-23 09:44:11 -07001218#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001219
Tejun Heocf771cb2008-09-03 09:01:09 +02001220dev_t blk_lookup_devt(const char *name, int partno)
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001221{
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001222 dev_t devt = MKDEV(0, 0);
Tejun Heodef4e382008-09-03 08:57:12 +02001223 struct class_dev_iter iter;
1224 struct device *dev;
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001225
Tejun Heodef4e382008-09-03 08:57:12 +02001226 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1227 while ((dev = class_dev_iter_next(&iter))) {
1228 struct gendisk *disk = dev_to_disk(dev);
Tejun Heo548b10e2008-08-29 09:01:47 +02001229 struct hd_struct *part;
Tejun Heodef4e382008-09-03 08:57:12 +02001230
Kay Sievers3ada8b72009-01-06 10:44:43 -08001231 if (strcmp(dev_name(dev), name))
Tejun Heof331c022008-09-03 09:01:48 +02001232 continue;
Tejun Heof331c022008-09-03 09:01:48 +02001233
Neil Brown41b8c852009-02-18 10:33:59 +01001234 if (partno < disk->minors) {
1235 /* We need to return the right devno, even
1236 * if the partition doesn't exist yet.
1237 */
1238 devt = MKDEV(MAJOR(dev->devt),
1239 MINOR(dev->devt) + partno);
1240 break;
1241 }
Tejun Heo548b10e2008-08-29 09:01:47 +02001242 part = disk_get_part(disk, partno);
Tejun Heo2bbedcb2008-08-29 11:41:51 +02001243 if (part) {
Tejun Heof331c022008-09-03 09:01:48 +02001244 devt = part_devt(part);
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001245 disk_put_part(part);
Tejun Heo548b10e2008-08-29 09:01:47 +02001246 break;
Tejun Heodef4e382008-09-03 08:57:12 +02001247 }
Tejun Heo548b10e2008-08-29 09:01:47 +02001248 disk_put_part(part);
Kay Sievers5c0ef6d2008-08-16 14:30:30 +02001249 }
Tejun Heodef4e382008-09-03 08:57:12 +02001250 class_dev_iter_exit(&iter);
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001251 return devt;
1252}
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001253EXPORT_SYMBOL(blk_lookup_devt);
1254
Linus Torvalds1da177e2005-04-16 15:20:36 -07001255struct gendisk *alloc_disk(int minors)
1256{
Ezequiel Garciac304a512012-11-10 10:39:44 +01001257 return alloc_disk_node(minors, NUMA_NO_NODE);
Christoph Lameter19460892005-06-23 00:08:19 -07001258}
Tejun Heo689d6fa2008-08-25 19:56:16 +09001259EXPORT_SYMBOL(alloc_disk);
Christoph Lameter19460892005-06-23 00:08:19 -07001260
1261struct gendisk *alloc_disk_node(int minors, int node_id)
1262{
1263 struct gendisk *disk;
1264
Joe Perchesc1b511e2013-08-29 15:21:42 -07001265 disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001266 if (disk) {
Tejun Heo074a7ac2008-08-25 19:56:14 +09001267 if (!init_part_stats(&disk->part0)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001268 kfree(disk);
1269 return NULL;
1270 }
Cheng Renquanbf91db12008-11-20 08:37:37 +01001271 disk->node_id = node_id;
Tejun Heo540eed52008-08-25 19:56:15 +09001272 if (disk_expand_part_tbl(disk, 0)) {
1273 free_part_stats(&disk->part0);
Tejun Heob5d0b9d2008-09-03 09:06:42 +02001274 kfree(disk);
1275 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276 }
Tejun Heo540eed52008-08-25 19:56:15 +09001277 disk->part_tbl->part[0] = &disk->part0;
Jens Axboe6c23a962011-01-07 08:43:37 +01001278
Vivek Goyalc83f6bf2012-08-01 12:24:18 +02001279 /*
1280 * set_capacity() and get_capacity() currently don't use
1281 * seqcounter to read/update the part0->nr_sects. Still init
1282 * the counter as we can read the sectors in IO submission
1283 * patch using seqence counters.
1284 *
1285 * TODO: Ideally set_capacity() and get_capacity() should be
1286 * converted to make use of bd_mutex and sequence counters.
1287 */
1288 seqcount_init(&disk->part0.nr_sects_seq);
Ming Lei6c710132015-07-16 11:16:45 +08001289 if (hd_ref_init(&disk->part0)) {
1290 hd_free_part(&disk->part0);
1291 kfree(disk);
1292 return NULL;
1293 }
Tejun Heob5d0b9d2008-09-03 09:06:42 +02001294
Linus Torvalds1da177e2005-04-16 15:20:36 -07001295 disk->minors = minors;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001296 rand_initialize_disk(disk);
Tejun Heoed9e1982008-08-25 19:56:05 +09001297 disk_to_dev(disk)->class = &block_class;
1298 disk_to_dev(disk)->type = &disk_type;
1299 device_initialize(disk_to_dev(disk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001300 }
1301 return disk;
1302}
Christoph Lameter19460892005-06-23 00:08:19 -07001303EXPORT_SYMBOL(alloc_disk_node);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001304
1305struct kobject *get_disk(struct gendisk *disk)
1306{
1307 struct module *owner;
1308 struct kobject *kobj;
1309
1310 if (!disk->fops)
1311 return NULL;
1312 owner = disk->fops->owner;
1313 if (owner && !try_module_get(owner))
1314 return NULL;
Tejun Heoed9e1982008-08-25 19:56:05 +09001315 kobj = kobject_get(&disk_to_dev(disk)->kobj);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001316 if (kobj == NULL) {
1317 module_put(owner);
1318 return NULL;
1319 }
1320 return kobj;
1321
1322}
1323
1324EXPORT_SYMBOL(get_disk);
1325
1326void put_disk(struct gendisk *disk)
1327{
1328 if (disk)
Tejun Heoed9e1982008-08-25 19:56:05 +09001329 kobject_put(&disk_to_dev(disk)->kobj);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001330}
1331
1332EXPORT_SYMBOL(put_disk);
1333
Hannes Reineckee3264a42009-07-28 09:13:13 +02001334static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1335{
1336 char event[] = "DISK_RO=1";
1337 char *envp[] = { event, NULL };
1338
1339 if (!ro)
1340 event[8] = '0';
1341 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1342}
1343
Linus Torvalds1da177e2005-04-16 15:20:36 -07001344void set_device_ro(struct block_device *bdev, int flag)
1345{
Tejun Heob7db9952008-08-25 19:56:10 +09001346 bdev->bd_part->policy = flag;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001347}
1348
1349EXPORT_SYMBOL(set_device_ro);
1350
1351void set_disk_ro(struct gendisk *disk, int flag)
1352{
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001353 struct disk_part_iter piter;
1354 struct hd_struct *part;
1355
Hannes Reineckee3264a42009-07-28 09:13:13 +02001356 if (disk->part0.policy != flag) {
1357 set_disk_ro_uevent(disk, flag);
1358 disk->part0.policy = flag;
1359 }
1360
1361 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY);
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001362 while ((part = disk_part_iter_next(&piter)))
1363 part->policy = flag;
1364 disk_part_iter_exit(&piter);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365}
1366
1367EXPORT_SYMBOL(set_disk_ro);
1368
1369int bdev_read_only(struct block_device *bdev)
1370{
1371 if (!bdev)
1372 return 0;
Tejun Heob7db9952008-08-25 19:56:10 +09001373 return bdev->bd_part->policy;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001374}
1375
1376EXPORT_SYMBOL(bdev_read_only);
1377
Tejun Heocf771cb2008-09-03 09:01:09 +02001378int invalidate_partition(struct gendisk *disk, int partno)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001379{
1380 int res = 0;
Tejun Heocf771cb2008-09-03 09:01:09 +02001381 struct block_device *bdev = bdget_disk(disk, partno);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001382 if (bdev) {
Christoph Hellwig2ef41632005-05-05 16:15:59 -07001383 fsync_bdev(bdev);
NeilBrown93b270f2011-02-24 17:25:47 +11001384 res = __invalidate_device(bdev, true);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001385 bdput(bdev);
1386 }
1387 return res;
1388}
1389
1390EXPORT_SYMBOL(invalidate_partition);
Tejun Heo77ea8872010-12-08 20:57:37 +01001391
1392/*
1393 * Disk events - monitor disk events like media change and eject request.
1394 */
1395struct disk_events {
1396 struct list_head node; /* all disk_event's */
1397 struct gendisk *disk; /* the associated disk */
1398 spinlock_t lock;
1399
Tejun Heofdd514e2011-06-09 20:43:59 +02001400 struct mutex block_mutex; /* protects blocking */
Tejun Heo77ea8872010-12-08 20:57:37 +01001401 int block; /* event blocking depth */
1402 unsigned int pending; /* events already sent out */
1403 unsigned int clearing; /* events being cleared */
1404
1405 long poll_msecs; /* interval, -1 for default */
1406 struct delayed_work dwork;
1407};
1408
1409static const char *disk_events_strs[] = {
1410 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "media_change",
1411 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "eject_request",
1412};
1413
1414static char *disk_uevents[] = {
1415 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "DISK_MEDIA_CHANGE=1",
1416 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "DISK_EJECT_REQUEST=1",
1417};
1418
1419/* list of all disk_events */
1420static DEFINE_MUTEX(disk_events_mutex);
1421static LIST_HEAD(disk_events);
1422
1423/* disable in-kernel polling by default */
1424static unsigned long disk_events_dfl_poll_msecs = 0;
1425
1426static unsigned long disk_events_poll_jiffies(struct gendisk *disk)
1427{
1428 struct disk_events *ev = disk->ev;
1429 long intv_msecs = 0;
1430
1431 /*
1432 * If device-specific poll interval is set, always use it. If
1433 * the default is being used, poll iff there are events which
1434 * can't be monitored asynchronously.
1435 */
1436 if (ev->poll_msecs >= 0)
1437 intv_msecs = ev->poll_msecs;
1438 else if (disk->events & ~disk->async_events)
1439 intv_msecs = disk_events_dfl_poll_msecs;
1440
1441 return msecs_to_jiffies(intv_msecs);
1442}
1443
Tejun Heoc3af54a2011-06-09 20:43:55 +02001444/**
1445 * disk_block_events - block and flush disk event checking
1446 * @disk: disk to block events for
1447 *
1448 * On return from this function, it is guaranteed that event checking
1449 * isn't in progress and won't happen until unblocked by
1450 * disk_unblock_events(). Events blocking is counted and the actual
1451 * unblocking happens after the matching number of unblocks are done.
1452 *
1453 * Note that this intentionally does not block event checking from
1454 * disk_clear_events().
1455 *
1456 * CONTEXT:
1457 * Might sleep.
1458 */
1459void disk_block_events(struct gendisk *disk)
Tejun Heo77ea8872010-12-08 20:57:37 +01001460{
1461 struct disk_events *ev = disk->ev;
1462 unsigned long flags;
1463 bool cancel;
1464
Tejun Heoc3af54a2011-06-09 20:43:55 +02001465 if (!ev)
1466 return;
1467
Tejun Heofdd514e2011-06-09 20:43:59 +02001468 /*
1469 * Outer mutex ensures that the first blocker completes canceling
1470 * the event work before further blockers are allowed to finish.
1471 */
1472 mutex_lock(&ev->block_mutex);
1473
Tejun Heo77ea8872010-12-08 20:57:37 +01001474 spin_lock_irqsave(&ev->lock, flags);
1475 cancel = !ev->block++;
1476 spin_unlock_irqrestore(&ev->lock, flags);
1477
Tejun Heoc3af54a2011-06-09 20:43:55 +02001478 if (cancel)
1479 cancel_delayed_work_sync(&disk->ev->dwork);
Tejun Heofdd514e2011-06-09 20:43:59 +02001480
1481 mutex_unlock(&ev->block_mutex);
Tejun Heo77ea8872010-12-08 20:57:37 +01001482}
1483
1484static void __disk_unblock_events(struct gendisk *disk, bool check_now)
1485{
1486 struct disk_events *ev = disk->ev;
1487 unsigned long intv;
1488 unsigned long flags;
1489
1490 spin_lock_irqsave(&ev->lock, flags);
1491
1492 if (WARN_ON_ONCE(ev->block <= 0))
1493 goto out_unlock;
1494
1495 if (--ev->block)
1496 goto out_unlock;
1497
1498 /*
1499 * Not exactly a latency critical operation, set poll timer
1500 * slack to 25% and kick event check.
1501 */
1502 intv = disk_events_poll_jiffies(disk);
1503 set_timer_slack(&ev->dwork.timer, intv / 4);
1504 if (check_now)
Viresh Kumar695588f2013-04-24 17:12:56 +05301505 queue_delayed_work(system_freezable_power_efficient_wq,
1506 &ev->dwork, 0);
Tejun Heo77ea8872010-12-08 20:57:37 +01001507 else if (intv)
Viresh Kumar695588f2013-04-24 17:12:56 +05301508 queue_delayed_work(system_freezable_power_efficient_wq,
1509 &ev->dwork, intv);
Tejun Heo77ea8872010-12-08 20:57:37 +01001510out_unlock:
1511 spin_unlock_irqrestore(&ev->lock, flags);
1512}
1513
1514/**
Tejun Heo77ea8872010-12-08 20:57:37 +01001515 * disk_unblock_events - unblock disk event checking
1516 * @disk: disk to unblock events for
1517 *
1518 * Undo disk_block_events(). When the block count reaches zero, it
1519 * starts events polling if configured.
1520 *
1521 * CONTEXT:
1522 * Don't care. Safe to call from irq context.
1523 */
1524void disk_unblock_events(struct gendisk *disk)
1525{
1526 if (disk->ev)
Tejun Heofacc31d2011-03-09 19:54:27 +01001527 __disk_unblock_events(disk, false);
Tejun Heo77ea8872010-12-08 20:57:37 +01001528}
1529
1530/**
Tejun Heo85ef06d2011-07-01 16:17:47 +02001531 * disk_flush_events - schedule immediate event checking and flushing
1532 * @disk: disk to check and flush events for
1533 * @mask: events to flush
Tejun Heo77ea8872010-12-08 20:57:37 +01001534 *
Tejun Heo85ef06d2011-07-01 16:17:47 +02001535 * Schedule immediate event checking on @disk if not blocked. Events in
1536 * @mask are scheduled to be cleared from the driver. Note that this
1537 * doesn't clear the events from @disk->ev.
Tejun Heo77ea8872010-12-08 20:57:37 +01001538 *
1539 * CONTEXT:
Tejun Heo85ef06d2011-07-01 16:17:47 +02001540 * If @mask is non-zero must be called with bdev->bd_mutex held.
Tejun Heo77ea8872010-12-08 20:57:37 +01001541 */
Tejun Heo85ef06d2011-07-01 16:17:47 +02001542void disk_flush_events(struct gendisk *disk, unsigned int mask)
Tejun Heo77ea8872010-12-08 20:57:37 +01001543{
Tejun Heoa9dce2a2011-06-09 20:43:54 +02001544 struct disk_events *ev = disk->ev;
Tejun Heoa9dce2a2011-06-09 20:43:54 +02001545
1546 if (!ev)
1547 return;
1548
Tejun Heo85ef06d2011-07-01 16:17:47 +02001549 spin_lock_irq(&ev->lock);
1550 ev->clearing |= mask;
Tejun Heo41f63c52012-08-03 10:30:47 -07001551 if (!ev->block)
Viresh Kumar695588f2013-04-24 17:12:56 +05301552 mod_delayed_work(system_freezable_power_efficient_wq,
1553 &ev->dwork, 0);
Tejun Heo85ef06d2011-07-01 16:17:47 +02001554 spin_unlock_irq(&ev->lock);
Tejun Heo77ea8872010-12-08 20:57:37 +01001555}
Tejun Heo77ea8872010-12-08 20:57:37 +01001556
1557/**
1558 * disk_clear_events - synchronously check, clear and return pending events
1559 * @disk: disk to fetch and clear events from
Masanari Iidada3dae52014-09-09 01:27:23 +09001560 * @mask: mask of events to be fetched and cleared
Tejun Heo77ea8872010-12-08 20:57:37 +01001561 *
1562 * Disk events are synchronously checked and pending events in @mask
1563 * are cleared and returned. This ignores the block count.
1564 *
1565 * CONTEXT:
1566 * Might sleep.
1567 */
1568unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask)
1569{
1570 const struct block_device_operations *bdops = disk->fops;
1571 struct disk_events *ev = disk->ev;
1572 unsigned int pending;
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001573 unsigned int clearing = mask;
Tejun Heo77ea8872010-12-08 20:57:37 +01001574
1575 if (!ev) {
1576 /* for drivers still using the old ->media_changed method */
1577 if ((mask & DISK_EVENT_MEDIA_CHANGE) &&
1578 bdops->media_changed && bdops->media_changed(disk))
1579 return DISK_EVENT_MEDIA_CHANGE;
1580 return 0;
1581 }
1582
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001583 disk_block_events(disk);
1584
1585 /*
1586 * store the union of mask and ev->clearing on the stack so that the
1587 * race with disk_flush_events does not cause ambiguity (ev->clearing
1588 * can still be modified even if events are blocked).
1589 */
Tejun Heo77ea8872010-12-08 20:57:37 +01001590 spin_lock_irq(&ev->lock);
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001591 clearing |= ev->clearing;
1592 ev->clearing = 0;
Tejun Heo77ea8872010-12-08 20:57:37 +01001593 spin_unlock_irq(&ev->lock);
1594
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001595 disk_check_events(ev, &clearing);
Derek Basehoreaea24a82012-12-18 12:27:18 -08001596 /*
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001597 * if ev->clearing is not 0, the disk_flush_events got called in the
1598 * middle of this function, so we want to run the workfn without delay.
Derek Basehoreaea24a82012-12-18 12:27:18 -08001599 */
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001600 __disk_unblock_events(disk, ev->clearing ? true : false);
Tejun Heo77ea8872010-12-08 20:57:37 +01001601
1602 /* then, fetch and clear pending events */
1603 spin_lock_irq(&ev->lock);
Tejun Heo77ea8872010-12-08 20:57:37 +01001604 pending = ev->pending & mask;
1605 ev->pending &= ~mask;
1606 spin_unlock_irq(&ev->lock);
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001607 WARN_ON_ONCE(clearing & mask);
Tejun Heo77ea8872010-12-08 20:57:37 +01001608
1609 return pending;
1610}
1611
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001612/*
1613 * Separate this part out so that a different pointer for clearing_ptr can be
1614 * passed in for disk_clear_events.
1615 */
Tejun Heo77ea8872010-12-08 20:57:37 +01001616static void disk_events_workfn(struct work_struct *work)
1617{
1618 struct delayed_work *dwork = to_delayed_work(work);
1619 struct disk_events *ev = container_of(dwork, struct disk_events, dwork);
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001620
1621 disk_check_events(ev, &ev->clearing);
1622}
1623
1624static void disk_check_events(struct disk_events *ev,
1625 unsigned int *clearing_ptr)
1626{
Tejun Heo77ea8872010-12-08 20:57:37 +01001627 struct gendisk *disk = ev->disk;
1628 char *envp[ARRAY_SIZE(disk_uevents) + 1] = { };
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001629 unsigned int clearing = *clearing_ptr;
Tejun Heo77ea8872010-12-08 20:57:37 +01001630 unsigned int events;
1631 unsigned long intv;
1632 int nr_events = 0, i;
1633
1634 /* check events */
1635 events = disk->fops->check_events(disk, clearing);
1636
1637 /* accumulate pending events and schedule next poll if necessary */
1638 spin_lock_irq(&ev->lock);
1639
1640 events &= ~ev->pending;
1641 ev->pending |= events;
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001642 *clearing_ptr &= ~clearing;
Tejun Heo77ea8872010-12-08 20:57:37 +01001643
1644 intv = disk_events_poll_jiffies(disk);
1645 if (!ev->block && intv)
Viresh Kumar695588f2013-04-24 17:12:56 +05301646 queue_delayed_work(system_freezable_power_efficient_wq,
1647 &ev->dwork, intv);
Tejun Heo77ea8872010-12-08 20:57:37 +01001648
1649 spin_unlock_irq(&ev->lock);
1650
Tejun Heo7c88a162011-04-21 19:43:58 +02001651 /*
1652 * Tell userland about new events. Only the events listed in
1653 * @disk->events are reported. Unlisted events are processed the
1654 * same internally but never get reported to userland.
1655 */
Tejun Heo77ea8872010-12-08 20:57:37 +01001656 for (i = 0; i < ARRAY_SIZE(disk_uevents); i++)
Tejun Heo7c88a162011-04-21 19:43:58 +02001657 if (events & disk->events & (1 << i))
Tejun Heo77ea8872010-12-08 20:57:37 +01001658 envp[nr_events++] = disk_uevents[i];
1659
1660 if (nr_events)
1661 kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp);
1662}
1663
1664/*
1665 * A disk events enabled device has the following sysfs nodes under
1666 * its /sys/block/X/ directory.
1667 *
1668 * events : list of all supported events
1669 * events_async : list of events which can be detected w/o polling
1670 * events_poll_msecs : polling interval, 0: disable, -1: system default
1671 */
1672static ssize_t __disk_events_show(unsigned int events, char *buf)
1673{
1674 const char *delim = "";
1675 ssize_t pos = 0;
1676 int i;
1677
1678 for (i = 0; i < ARRAY_SIZE(disk_events_strs); i++)
1679 if (events & (1 << i)) {
1680 pos += sprintf(buf + pos, "%s%s",
1681 delim, disk_events_strs[i]);
1682 delim = " ";
1683 }
1684 if (pos)
1685 pos += sprintf(buf + pos, "\n");
1686 return pos;
1687}
1688
1689static ssize_t disk_events_show(struct device *dev,
1690 struct device_attribute *attr, char *buf)
1691{
1692 struct gendisk *disk = dev_to_disk(dev);
1693
1694 return __disk_events_show(disk->events, buf);
1695}
1696
1697static ssize_t disk_events_async_show(struct device *dev,
1698 struct device_attribute *attr, char *buf)
1699{
1700 struct gendisk *disk = dev_to_disk(dev);
1701
1702 return __disk_events_show(disk->async_events, buf);
1703}
1704
1705static ssize_t disk_events_poll_msecs_show(struct device *dev,
1706 struct device_attribute *attr,
1707 char *buf)
1708{
1709 struct gendisk *disk = dev_to_disk(dev);
1710
1711 return sprintf(buf, "%ld\n", disk->ev->poll_msecs);
1712}
1713
1714static ssize_t disk_events_poll_msecs_store(struct device *dev,
1715 struct device_attribute *attr,
1716 const char *buf, size_t count)
1717{
1718 struct gendisk *disk = dev_to_disk(dev);
1719 long intv;
1720
1721 if (!count || !sscanf(buf, "%ld", &intv))
1722 return -EINVAL;
1723
1724 if (intv < 0 && intv != -1)
1725 return -EINVAL;
1726
Tejun Heoc3af54a2011-06-09 20:43:55 +02001727 disk_block_events(disk);
Tejun Heo77ea8872010-12-08 20:57:37 +01001728 disk->ev->poll_msecs = intv;
1729 __disk_unblock_events(disk, true);
1730
1731 return count;
1732}
1733
1734static const DEVICE_ATTR(events, S_IRUGO, disk_events_show, NULL);
1735static const DEVICE_ATTR(events_async, S_IRUGO, disk_events_async_show, NULL);
1736static const DEVICE_ATTR(events_poll_msecs, S_IRUGO|S_IWUSR,
1737 disk_events_poll_msecs_show,
1738 disk_events_poll_msecs_store);
1739
1740static const struct attribute *disk_events_attrs[] = {
1741 &dev_attr_events.attr,
1742 &dev_attr_events_async.attr,
1743 &dev_attr_events_poll_msecs.attr,
1744 NULL,
1745};
1746
1747/*
1748 * The default polling interval can be specified by the kernel
1749 * parameter block.events_dfl_poll_msecs which defaults to 0
1750 * (disable). This can also be modified runtime by writing to
1751 * /sys/module/block/events_dfl_poll_msecs.
1752 */
1753static int disk_events_set_dfl_poll_msecs(const char *val,
1754 const struct kernel_param *kp)
1755{
1756 struct disk_events *ev;
1757 int ret;
1758
1759 ret = param_set_ulong(val, kp);
1760 if (ret < 0)
1761 return ret;
1762
1763 mutex_lock(&disk_events_mutex);
1764
1765 list_for_each_entry(ev, &disk_events, node)
Tejun Heo85ef06d2011-07-01 16:17:47 +02001766 disk_flush_events(ev->disk, 0);
Tejun Heo77ea8872010-12-08 20:57:37 +01001767
1768 mutex_unlock(&disk_events_mutex);
1769
1770 return 0;
1771}
1772
1773static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops = {
1774 .set = disk_events_set_dfl_poll_msecs,
1775 .get = param_get_ulong,
1776};
1777
1778#undef MODULE_PARAM_PREFIX
1779#define MODULE_PARAM_PREFIX "block."
1780
1781module_param_cb(events_dfl_poll_msecs, &disk_events_dfl_poll_msecs_param_ops,
1782 &disk_events_dfl_poll_msecs, 0644);
1783
1784/*
Stanislaw Gruszka9f53d2fe2012-03-02 10:43:28 +01001785 * disk_{alloc|add|del|release}_events - initialize and destroy disk_events.
Tejun Heo77ea8872010-12-08 20:57:37 +01001786 */
Stanislaw Gruszka9f53d2fe2012-03-02 10:43:28 +01001787static void disk_alloc_events(struct gendisk *disk)
Tejun Heo77ea8872010-12-08 20:57:37 +01001788{
1789 struct disk_events *ev;
1790
Tejun Heo75e3f3e2011-05-26 21:06:50 +02001791 if (!disk->fops->check_events)
Tejun Heo77ea8872010-12-08 20:57:37 +01001792 return;
1793
1794 ev = kzalloc(sizeof(*ev), GFP_KERNEL);
1795 if (!ev) {
1796 pr_warn("%s: failed to initialize events\n", disk->disk_name);
1797 return;
1798 }
1799
Tejun Heo77ea8872010-12-08 20:57:37 +01001800 INIT_LIST_HEAD(&ev->node);
1801 ev->disk = disk;
1802 spin_lock_init(&ev->lock);
Tejun Heofdd514e2011-06-09 20:43:59 +02001803 mutex_init(&ev->block_mutex);
Tejun Heo77ea8872010-12-08 20:57:37 +01001804 ev->block = 1;
1805 ev->poll_msecs = -1;
1806 INIT_DELAYED_WORK(&ev->dwork, disk_events_workfn);
1807
Stanislaw Gruszka9f53d2fe2012-03-02 10:43:28 +01001808 disk->ev = ev;
1809}
1810
1811static void disk_add_events(struct gendisk *disk)
1812{
1813 if (!disk->ev)
1814 return;
1815
1816 /* FIXME: error handling */
1817 if (sysfs_create_files(&disk_to_dev(disk)->kobj, disk_events_attrs) < 0)
1818 pr_warn("%s: failed to create sysfs files for events\n",
1819 disk->disk_name);
1820
Tejun Heo77ea8872010-12-08 20:57:37 +01001821 mutex_lock(&disk_events_mutex);
Stanislaw Gruszka9f53d2fe2012-03-02 10:43:28 +01001822 list_add_tail(&disk->ev->node, &disk_events);
Tejun Heo77ea8872010-12-08 20:57:37 +01001823 mutex_unlock(&disk_events_mutex);
1824
1825 /*
1826 * Block count is initialized to 1 and the following initial
1827 * unblock kicks it into action.
1828 */
1829 __disk_unblock_events(disk, true);
1830}
1831
1832static void disk_del_events(struct gendisk *disk)
1833{
1834 if (!disk->ev)
1835 return;
1836
Tejun Heoc3af54a2011-06-09 20:43:55 +02001837 disk_block_events(disk);
Tejun Heo77ea8872010-12-08 20:57:37 +01001838
1839 mutex_lock(&disk_events_mutex);
1840 list_del_init(&disk->ev->node);
1841 mutex_unlock(&disk_events_mutex);
1842
1843 sysfs_remove_files(&disk_to_dev(disk)->kobj, disk_events_attrs);
1844}
1845
1846static void disk_release_events(struct gendisk *disk)
1847{
1848 /* the block count should be 1 from disk_del_events() */
1849 WARN_ON_ONCE(disk->ev && disk->ev->block != 1);
1850 kfree(disk->ev);
1851}