blob: fcd6d4fae657cfdf118274cec31780f9e8b8c1f8 [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>
Vishal Verma99e66082016-01-09 08:36:51 -080023#include <linux/badblocks.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070024
Adrian Bunkff889722008-03-04 11:23:45 +010025#include "blk.h"
26
Kay Sieversedfaa7c2007-05-21 22:08:01 +020027static DEFINE_MUTEX(block_class_lock);
Kay Sieversedfaa7c2007-05-21 22:08:01 +020028struct kobject *block_depr;
Linus Torvalds1da177e2005-04-16 15:20:36 -070029
Tejun Heobcce3de2008-08-25 19:47:22 +090030/* for extended dynamic devt allocation, currently only one major is used */
Tejun Heoce23bba2013-02-27 17:03:56 -080031#define NR_EXT_DEVT (1 << MINORBITS)
Tejun Heobcce3de2008-08-25 19:47:22 +090032
Keith Busch2da78092014-08-26 09:05:36 -060033/* For extended devt allocation. ext_devt_lock prevents look up
Tejun Heobcce3de2008-08-25 19:47:22 +090034 * results from going away underneath its user.
35 */
Keith Busch2da78092014-08-26 09:05:36 -060036static DEFINE_SPINLOCK(ext_devt_lock);
Tejun Heobcce3de2008-08-25 19:47:22 +090037static DEFINE_IDR(ext_devt_idr);
38
Adrian Bunk1826ead2008-03-04 11:23:46 +010039static struct device_type disk_type;
40
Derek Basehore12c2bdb2012-12-18 12:27:20 -080041static void disk_check_events(struct disk_events *ev,
42 unsigned int *clearing_ptr);
Stanislaw Gruszka9f53d2fe2012-03-02 10:43:28 +010043static void disk_alloc_events(struct gendisk *disk);
Tejun Heo77ea8872010-12-08 20:57:37 +010044static void disk_add_events(struct gendisk *disk);
45static void disk_del_events(struct gendisk *disk);
46static void disk_release_events(struct gendisk *disk);
47
Tejun Heoe71bf0d2008-09-03 09:03:02 +020048/**
49 * disk_get_part - get partition
50 * @disk: disk to look partition from
51 * @partno: partition number
52 *
53 * Look for partition @partno from @disk. If found, increment
54 * reference count and return it.
55 *
56 * CONTEXT:
57 * Don't care.
58 *
59 * RETURNS:
60 * Pointer to the found partition on success, NULL if not found.
61 */
62struct hd_struct *disk_get_part(struct gendisk *disk, int partno)
63{
Tejun Heo540eed52008-08-25 19:56:15 +090064 struct hd_struct *part = NULL;
65 struct disk_part_tbl *ptbl;
Tejun Heoe71bf0d2008-09-03 09:03:02 +020066
Tejun Heo540eed52008-08-25 19:56:15 +090067 if (unlikely(partno < 0))
Tejun Heoe71bf0d2008-09-03 09:03:02 +020068 return NULL;
Tejun Heo540eed52008-08-25 19:56:15 +090069
Tejun Heoe71bf0d2008-09-03 09:03:02 +020070 rcu_read_lock();
Tejun Heo540eed52008-08-25 19:56:15 +090071
72 ptbl = rcu_dereference(disk->part_tbl);
73 if (likely(partno < ptbl->len)) {
74 part = rcu_dereference(ptbl->part[partno]);
75 if (part)
76 get_device(part_to_dev(part));
77 }
78
Tejun Heoe71bf0d2008-09-03 09:03:02 +020079 rcu_read_unlock();
80
81 return part;
82}
83EXPORT_SYMBOL_GPL(disk_get_part);
84
85/**
86 * disk_part_iter_init - initialize partition iterator
87 * @piter: iterator to initialize
88 * @disk: disk to iterate over
89 * @flags: DISK_PITER_* flags
90 *
91 * Initialize @piter so that it iterates over partitions of @disk.
92 *
93 * CONTEXT:
94 * Don't care.
95 */
96void disk_part_iter_init(struct disk_part_iter *piter, struct gendisk *disk,
97 unsigned int flags)
98{
Tejun Heo540eed52008-08-25 19:56:15 +090099 struct disk_part_tbl *ptbl;
100
101 rcu_read_lock();
102 ptbl = rcu_dereference(disk->part_tbl);
103
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200104 piter->disk = disk;
105 piter->part = NULL;
106
107 if (flags & DISK_PITER_REVERSE)
Tejun Heo540eed52008-08-25 19:56:15 +0900108 piter->idx = ptbl->len - 1;
Tejun Heo71982a42009-04-17 08:34:48 +0200109 else if (flags & (DISK_PITER_INCL_PART0 | DISK_PITER_INCL_EMPTY_PART0))
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200110 piter->idx = 0;
Tejun Heob5d0b9d2008-09-03 09:06:42 +0200111 else
112 piter->idx = 1;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200113
114 piter->flags = flags;
Tejun Heo540eed52008-08-25 19:56:15 +0900115
116 rcu_read_unlock();
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200117}
118EXPORT_SYMBOL_GPL(disk_part_iter_init);
119
120/**
121 * disk_part_iter_next - proceed iterator to the next partition and return it
122 * @piter: iterator of interest
123 *
124 * Proceed @piter to the next partition and return it.
125 *
126 * CONTEXT:
127 * Don't care.
128 */
129struct hd_struct *disk_part_iter_next(struct disk_part_iter *piter)
130{
Tejun Heo540eed52008-08-25 19:56:15 +0900131 struct disk_part_tbl *ptbl;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200132 int inc, end;
133
134 /* put the last partition */
135 disk_put_part(piter->part);
136 piter->part = NULL;
137
Tejun Heo540eed52008-08-25 19:56:15 +0900138 /* get part_tbl */
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200139 rcu_read_lock();
Tejun Heo540eed52008-08-25 19:56:15 +0900140 ptbl = rcu_dereference(piter->disk->part_tbl);
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200141
142 /* determine iteration parameters */
143 if (piter->flags & DISK_PITER_REVERSE) {
144 inc = -1;
Tejun Heo71982a42009-04-17 08:34:48 +0200145 if (piter->flags & (DISK_PITER_INCL_PART0 |
146 DISK_PITER_INCL_EMPTY_PART0))
Tejun Heob5d0b9d2008-09-03 09:06:42 +0200147 end = -1;
148 else
149 end = 0;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200150 } else {
151 inc = 1;
Tejun Heo540eed52008-08-25 19:56:15 +0900152 end = ptbl->len;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200153 }
154
155 /* iterate to the next partition */
156 for (; piter->idx != end; piter->idx += inc) {
157 struct hd_struct *part;
158
Tejun Heo540eed52008-08-25 19:56:15 +0900159 part = rcu_dereference(ptbl->part[piter->idx]);
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200160 if (!part)
161 continue;
Vivek Goyalc83f6bf2012-08-01 12:24:18 +0200162 if (!part_nr_sects_read(part) &&
Tejun Heo71982a42009-04-17 08:34:48 +0200163 !(piter->flags & DISK_PITER_INCL_EMPTY) &&
164 !(piter->flags & DISK_PITER_INCL_EMPTY_PART0 &&
165 piter->idx == 0))
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200166 continue;
167
Tejun Heoed9e1982008-08-25 19:56:05 +0900168 get_device(part_to_dev(part));
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200169 piter->part = part;
170 piter->idx += inc;
171 break;
172 }
173
174 rcu_read_unlock();
175
176 return piter->part;
177}
178EXPORT_SYMBOL_GPL(disk_part_iter_next);
179
180/**
181 * disk_part_iter_exit - finish up partition iteration
182 * @piter: iter of interest
183 *
184 * Called when iteration is over. Cleans up @piter.
185 *
186 * CONTEXT:
187 * Don't care.
188 */
189void disk_part_iter_exit(struct disk_part_iter *piter)
190{
191 disk_put_part(piter->part);
192 piter->part = NULL;
193}
194EXPORT_SYMBOL_GPL(disk_part_iter_exit);
195
Jens Axboea6f23652008-10-24 12:52:42 +0200196static inline int sector_in_part(struct hd_struct *part, sector_t sector)
197{
198 return part->start_sect <= sector &&
Vivek Goyalc83f6bf2012-08-01 12:24:18 +0200199 sector < part->start_sect + part_nr_sects_read(part);
Jens Axboea6f23652008-10-24 12:52:42 +0200200}
201
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200202/**
203 * disk_map_sector_rcu - map sector to partition
204 * @disk: gendisk of interest
205 * @sector: sector to map
206 *
207 * Find out which partition @sector maps to on @disk. This is
208 * primarily used for stats accounting.
209 *
210 * CONTEXT:
211 * RCU read locked. The returned partition pointer is valid only
212 * while preemption is disabled.
213 *
214 * RETURNS:
Tejun Heo074a7ac2008-08-25 19:56:14 +0900215 * Found partition on success, part0 is returned if no partition matches
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200216 */
217struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector)
218{
Tejun Heo540eed52008-08-25 19:56:15 +0900219 struct disk_part_tbl *ptbl;
Jens Axboea6f23652008-10-24 12:52:42 +0200220 struct hd_struct *part;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200221 int i;
222
Tejun Heo540eed52008-08-25 19:56:15 +0900223 ptbl = rcu_dereference(disk->part_tbl);
224
Jens Axboea6f23652008-10-24 12:52:42 +0200225 part = rcu_dereference(ptbl->last_lookup);
226 if (part && sector_in_part(part, sector))
227 return part;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200228
Jens Axboea6f23652008-10-24 12:52:42 +0200229 for (i = 1; i < ptbl->len; i++) {
230 part = rcu_dereference(ptbl->part[i]);
231
232 if (part && sector_in_part(part, sector)) {
233 rcu_assign_pointer(ptbl->last_lookup, part);
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200234 return part;
Jens Axboea6f23652008-10-24 12:52:42 +0200235 }
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200236 }
Tejun Heo074a7ac2008-08-25 19:56:14 +0900237 return &disk->part0;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200238}
239EXPORT_SYMBOL_GPL(disk_map_sector_rcu);
240
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241/*
242 * Can be deleted altogether. Later.
243 *
244 */
245static struct blk_major_name {
246 struct blk_major_name *next;
247 int major;
248 char name[16];
Joe Korty68eef3b2006-03-31 02:30:32 -0800249} *major_names[BLKDEV_MAJOR_HASH_SIZE];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250
251/* index in the above - for now: assume no multimajor ranges */
Yang Zhange61eb2e2010-12-17 09:00:18 +0100252static inline int major_to_index(unsigned major)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253{
Joe Korty68eef3b2006-03-31 02:30:32 -0800254 return major % BLKDEV_MAJOR_HASH_SIZE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255}
256
Joe Korty68eef3b2006-03-31 02:30:32 -0800257#ifdef CONFIG_PROC_FS
Tejun Heocf771cb2008-09-03 09:01:09 +0200258void blkdev_show(struct seq_file *seqf, off_t offset)
Neil Horman7170be52006-01-14 13:20:38 -0800259{
Joe Korty68eef3b2006-03-31 02:30:32 -0800260 struct blk_major_name *dp;
Neil Horman7170be52006-01-14 13:20:38 -0800261
Joe Korty68eef3b2006-03-31 02:30:32 -0800262 if (offset < BLKDEV_MAJOR_HASH_SIZE) {
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200263 mutex_lock(&block_class_lock);
Joe Korty68eef3b2006-03-31 02:30:32 -0800264 for (dp = major_names[offset]; dp; dp = dp->next)
Tejun Heocf771cb2008-09-03 09:01:09 +0200265 seq_printf(seqf, "%3d %s\n", dp->major, dp->name);
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200266 mutex_unlock(&block_class_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700267 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700268}
Joe Korty68eef3b2006-03-31 02:30:32 -0800269#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270
Márton Németh9e8c0bc2009-02-20 08:12:51 +0100271/**
272 * register_blkdev - register a new block device
273 *
274 * @major: the requested major device number [1..255]. If @major=0, try to
275 * allocate any unused major number.
276 * @name: the name of the new block device as a zero terminated string
277 *
278 * The @name must be unique within the system.
279 *
280 * The return value depends on the @major input parameter.
281 * - if a major device number was requested in range [1..255] then the
282 * function returns zero on success, or a negative error code
283 * - if any unused major number was requested with @major=0 parameter
284 * then the return value is the allocated major number in range
285 * [1..255] or a negative error code otherwise
286 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287int register_blkdev(unsigned int major, const char *name)
288{
289 struct blk_major_name **n, *p;
290 int index, ret = 0;
291
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200292 mutex_lock(&block_class_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293
294 /* temporary */
295 if (major == 0) {
296 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
297 if (major_names[index] == NULL)
298 break;
299 }
300
301 if (index == 0) {
302 printk("register_blkdev: failed to get major for %s\n",
303 name);
304 ret = -EBUSY;
305 goto out;
306 }
307 major = index;
308 ret = major;
309 }
310
311 p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
312 if (p == NULL) {
313 ret = -ENOMEM;
314 goto out;
315 }
316
317 p->major = major;
318 strlcpy(p->name, name, sizeof(p->name));
319 p->next = NULL;
320 index = major_to_index(major);
321
322 for (n = &major_names[index]; *n; n = &(*n)->next) {
323 if ((*n)->major == major)
324 break;
325 }
326 if (!*n)
327 *n = p;
328 else
329 ret = -EBUSY;
330
331 if (ret < 0) {
332 printk("register_blkdev: cannot get major %d for %s\n",
333 major, name);
334 kfree(p);
335 }
336out:
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200337 mutex_unlock(&block_class_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338 return ret;
339}
340
341EXPORT_SYMBOL(register_blkdev);
342
Akinobu Mitaf4480242007-07-17 04:03:47 -0700343void unregister_blkdev(unsigned int major, const char *name)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344{
345 struct blk_major_name **n;
346 struct blk_major_name *p = NULL;
347 int index = major_to_index(major);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200349 mutex_lock(&block_class_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350 for (n = &major_names[index]; *n; n = &(*n)->next)
351 if ((*n)->major == major)
352 break;
Akinobu Mita294462a2007-07-17 04:03:45 -0700353 if (!*n || strcmp((*n)->name, name)) {
354 WARN_ON(1);
Akinobu Mita294462a2007-07-17 04:03:45 -0700355 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700356 p = *n;
357 *n = p->next;
358 }
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200359 mutex_unlock(&block_class_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360 kfree(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700361}
362
363EXPORT_SYMBOL(unregister_blkdev);
364
365static struct kobj_map *bdev_map;
366
Tejun Heobcce3de2008-08-25 19:47:22 +0900367/**
Tejun Heo870d6652008-08-25 19:47:25 +0900368 * blk_mangle_minor - scatter minor numbers apart
369 * @minor: minor number to mangle
370 *
371 * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
372 * is enabled. Mangling twice gives the original value.
373 *
374 * RETURNS:
375 * Mangled value.
376 *
377 * CONTEXT:
378 * Don't care.
379 */
380static int blk_mangle_minor(int minor)
381{
382#ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
383 int i;
384
385 for (i = 0; i < MINORBITS / 2; i++) {
386 int low = minor & (1 << i);
387 int high = minor & (1 << (MINORBITS - 1 - i));
388 int distance = MINORBITS - 1 - 2 * i;
389
390 minor ^= low | high; /* clear both bits */
391 low <<= distance; /* swap the positions */
392 high >>= distance;
393 minor |= low | high; /* and set */
394 }
395#endif
396 return minor;
397}
398
399/**
Tejun Heobcce3de2008-08-25 19:47:22 +0900400 * blk_alloc_devt - allocate a dev_t for a partition
401 * @part: partition to allocate dev_t for
Tejun Heobcce3de2008-08-25 19:47:22 +0900402 * @devt: out parameter for resulting dev_t
403 *
404 * Allocate a dev_t for block device.
405 *
406 * RETURNS:
407 * 0 on success, allocated dev_t is returned in *@devt. -errno on
408 * failure.
409 *
410 * CONTEXT:
411 * Might sleep.
412 */
413int blk_alloc_devt(struct hd_struct *part, dev_t *devt)
414{
415 struct gendisk *disk = part_to_disk(part);
Tejun Heobab998d2013-02-27 17:03:57 -0800416 int idx;
Tejun Heobcce3de2008-08-25 19:47:22 +0900417
418 /* in consecutive minor range? */
419 if (part->partno < disk->minors) {
420 *devt = MKDEV(disk->major, disk->first_minor + part->partno);
421 return 0;
422 }
423
424 /* allocate ext devt */
Keith Busch2da78092014-08-26 09:05:36 -0600425 idr_preload(GFP_KERNEL);
426
Dan Williams4d66e5e2015-06-10 23:47:14 -0400427 spin_lock_bh(&ext_devt_lock);
Keith Busch2da78092014-08-26 09:05:36 -0600428 idx = idr_alloc(&ext_devt_idr, part, 0, NR_EXT_DEVT, GFP_NOWAIT);
Dan Williams4d66e5e2015-06-10 23:47:14 -0400429 spin_unlock_bh(&ext_devt_lock);
Keith Busch2da78092014-08-26 09:05:36 -0600430
431 idr_preload_end();
Tejun Heobab998d2013-02-27 17:03:57 -0800432 if (idx < 0)
433 return idx == -ENOSPC ? -EBUSY : idx;
Tejun Heobcce3de2008-08-25 19:47:22 +0900434
Tejun Heo870d6652008-08-25 19:47:25 +0900435 *devt = MKDEV(BLOCK_EXT_MAJOR, blk_mangle_minor(idx));
Tejun Heobcce3de2008-08-25 19:47:22 +0900436 return 0;
437}
438
439/**
440 * blk_free_devt - free a dev_t
441 * @devt: dev_t to free
442 *
443 * Free @devt which was allocated using blk_alloc_devt().
444 *
445 * CONTEXT:
446 * Might sleep.
447 */
448void blk_free_devt(dev_t devt)
449{
Tejun Heobcce3de2008-08-25 19:47:22 +0900450 if (devt == MKDEV(0, 0))
451 return;
452
453 if (MAJOR(devt) == BLOCK_EXT_MAJOR) {
Dan Williams4d66e5e2015-06-10 23:47:14 -0400454 spin_lock_bh(&ext_devt_lock);
Tejun Heo870d6652008-08-25 19:47:25 +0900455 idr_remove(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
Dan Williams4d66e5e2015-06-10 23:47:14 -0400456 spin_unlock_bh(&ext_devt_lock);
Tejun Heobcce3de2008-08-25 19:47:22 +0900457 }
458}
459
Tejun Heo1f014292008-08-25 19:47:23 +0900460static char *bdevt_str(dev_t devt, char *buf)
461{
462 if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
463 char tbuf[BDEVT_SIZE];
464 snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
465 snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
466 } else
467 snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
468
469 return buf;
470}
471
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472/*
473 * Register device numbers dev..(dev+range-1)
474 * range must be nonzero
475 * The hash chain is sorted on range, so that subranges can override.
476 */
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200477void blk_register_region(dev_t devt, unsigned long range, struct module *module,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700478 struct kobject *(*probe)(dev_t, int *, void *),
479 int (*lock)(dev_t, void *), void *data)
480{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200481 kobj_map(bdev_map, devt, range, module, probe, lock, data);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482}
483
484EXPORT_SYMBOL(blk_register_region);
485
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200486void blk_unregister_region(dev_t devt, unsigned long range)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200488 kobj_unmap(bdev_map, devt, range);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489}
490
491EXPORT_SYMBOL(blk_unregister_region);
492
Tejun Heocf771cb2008-09-03 09:01:09 +0200493static struct kobject *exact_match(dev_t devt, int *partno, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494{
495 struct gendisk *p = data;
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200496
Tejun Heoed9e1982008-08-25 19:56:05 +0900497 return &disk_to_dev(p)->kobj;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498}
499
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200500static int exact_lock(dev_t devt, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501{
502 struct gendisk *p = data;
503
504 if (!get_disk(p))
505 return -1;
506 return 0;
507}
508
Dan Williamse63a46b2016-06-15 18:17:27 -0700509static void register_disk(struct device *parent, struct gendisk *disk)
Tejun Heod2bf1b62010-12-08 20:57:36 +0100510{
511 struct device *ddev = disk_to_dev(disk);
512 struct block_device *bdev;
513 struct disk_part_iter piter;
514 struct hd_struct *part;
515 int err;
516
Dan Williamse63a46b2016-06-15 18:17:27 -0700517 ddev->parent = parent;
Tejun Heod2bf1b62010-12-08 20:57:36 +0100518
Kees Cookffc8b302013-07-03 15:01:14 -0700519 dev_set_name(ddev, "%s", disk->disk_name);
Tejun Heod2bf1b62010-12-08 20:57:36 +0100520
521 /* delay uevents, until we scanned partition table */
522 dev_set_uevent_suppress(ddev, 1);
523
524 if (device_add(ddev))
525 return;
526 if (!sysfs_deprecated) {
527 err = sysfs_create_link(block_depr, &ddev->kobj,
528 kobject_name(&ddev->kobj));
529 if (err) {
530 device_del(ddev);
531 return;
532 }
533 }
Ming Lei25e823c2013-02-22 16:34:13 -0800534
535 /*
536 * avoid probable deadlock caused by allocating memory with
537 * GFP_KERNEL in runtime_resume callback of its all ancestor
538 * devices
539 */
540 pm_runtime_set_memalloc_noio(ddev, true);
541
Tejun Heod2bf1b62010-12-08 20:57:36 +0100542 disk->part0.holder_dir = kobject_create_and_add("holders", &ddev->kobj);
543 disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
544
545 /* No minors to use for partitions */
Tejun Heod27769e2011-08-23 20:01:04 +0200546 if (!disk_part_scan_enabled(disk))
Tejun Heod2bf1b62010-12-08 20:57:36 +0100547 goto exit;
548
549 /* No such device (e.g., media were just removed) */
550 if (!get_capacity(disk))
551 goto exit;
552
553 bdev = bdget_disk(disk, 0);
554 if (!bdev)
555 goto exit;
556
557 bdev->bd_invalidated = 1;
558 err = blkdev_get(bdev, FMODE_READ, NULL);
559 if (err < 0)
560 goto exit;
561 blkdev_put(bdev, FMODE_READ);
562
563exit:
564 /* announce disk after possible partitions are created */
565 dev_set_uevent_suppress(ddev, 0);
566 kobject_uevent(&ddev->kobj, KOBJ_ADD);
567
568 /* announce possible partitions */
569 disk_part_iter_init(&piter, disk, 0);
570 while ((part = disk_part_iter_next(&piter)))
571 kobject_uevent(&part_to_dev(part)->kobj, KOBJ_ADD);
572 disk_part_iter_exit(&piter);
573}
574
Linus Torvalds1da177e2005-04-16 15:20:36 -0700575/**
Dan Williamse63a46b2016-06-15 18:17:27 -0700576 * device_add_disk - add partitioning information to kernel list
577 * @parent: parent device for the disk
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578 * @disk: per-device partitioning information
579 *
580 * This function registers the partitioning information in @disk
581 * with the kernel.
Tejun Heo3e1a7ff2008-08-25 19:56:17 +0900582 *
583 * FIXME: error handling
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584 */
Dan Williamse63a46b2016-06-15 18:17:27 -0700585void device_add_disk(struct device *parent, struct gendisk *disk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700586{
Peter Zijlstracf0ca9f2008-04-30 00:54:32 -0700587 struct backing_dev_info *bdi;
Tejun Heo3e1a7ff2008-08-25 19:56:17 +0900588 dev_t devt;
Greg Kroah-Hartman6ffeea72008-05-22 17:21:08 -0400589 int retval;
Peter Zijlstracf0ca9f2008-04-30 00:54:32 -0700590
Tejun Heo3e1a7ff2008-08-25 19:56:17 +0900591 /* minors == 0 indicates to use ext devt from part0 and should
592 * be accompanied with EXT_DEVT flag. Make sure all
593 * parameters make sense.
594 */
595 WARN_ON(disk->minors && !(disk->major || disk->first_minor));
596 WARN_ON(!disk->minors && !(disk->flags & GENHD_FL_EXT_DEVT));
597
Linus Torvalds1da177e2005-04-16 15:20:36 -0700598 disk->flags |= GENHD_FL_UP;
Tejun Heo3e1a7ff2008-08-25 19:56:17 +0900599
600 retval = blk_alloc_devt(&disk->part0, &devt);
601 if (retval) {
602 WARN_ON(1);
603 return;
604 }
605 disk_to_dev(disk)->devt = devt;
606
607 /* ->major and ->first_minor aren't supposed to be
608 * dereferenced from here on, but set them just in case.
609 */
610 disk->major = MAJOR(devt);
611 disk->first_minor = MINOR(devt);
612
Stanislaw Gruszka9f53d2fe2012-03-02 10:43:28 +0100613 disk_alloc_events(disk);
614
Wanlong Gao9f5e4862011-06-13 10:45:43 +0200615 /* Register BDI before referencing it from bdev */
Signed-off-by: Jan Kara01ea5062010-09-16 20:36:36 +0200616 bdi = &disk->queue->backing_dev_info;
Dan Williamsdf08c322016-07-31 11:15:13 -0700617 bdi_register_owner(bdi, disk_to_dev(disk));
Signed-off-by: Jan Kara01ea5062010-09-16 20:36:36 +0200618
Tejun Heof331c022008-09-03 09:01:48 +0200619 blk_register_region(disk_devt(disk), disk->minors, NULL,
620 exact_match, exact_lock, disk);
Dan Williamse63a46b2016-06-15 18:17:27 -0700621 register_disk(parent, disk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700622 blk_register_queue(disk);
Peter Zijlstracf0ca9f2008-04-30 00:54:32 -0700623
Tejun Heo523e1d32011-10-19 14:31:07 +0200624 /*
625 * Take an extra ref on queue which will be put on disk_release()
626 * so that it sticks around as long as @disk is there.
627 */
Tejun Heo09ac46c2011-12-14 00:33:38 +0100628 WARN_ON_ONCE(!blk_get_queue(disk->queue));
Tejun Heo523e1d32011-10-19 14:31:07 +0200629
Tejun Heoed9e1982008-08-25 19:56:05 +0900630 retval = sysfs_create_link(&disk_to_dev(disk)->kobj, &bdi->dev->kobj,
631 "bdi");
Greg Kroah-Hartman6ffeea72008-05-22 17:21:08 -0400632 WARN_ON(retval);
Tejun Heo77ea8872010-12-08 20:57:37 +0100633
634 disk_add_events(disk);
Martin K. Petersen25520d52015-10-21 13:19:49 -0400635 blk_integrity_add(disk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636}
Dan Williamse63a46b2016-06-15 18:17:27 -0700637EXPORT_SYMBOL(device_add_disk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638
Tejun Heod2bf1b62010-12-08 20:57:36 +0100639void del_gendisk(struct gendisk *disk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700640{
Tejun Heod2bf1b62010-12-08 20:57:36 +0100641 struct disk_part_iter piter;
642 struct hd_struct *part;
643
Martin K. Petersen25520d52015-10-21 13:19:49 -0400644 blk_integrity_del(disk);
Tejun Heo77ea8872010-12-08 20:57:37 +0100645 disk_del_events(disk);
646
Tejun Heod2bf1b62010-12-08 20:57:36 +0100647 /* invalidate stuff */
648 disk_part_iter_init(&piter, disk,
649 DISK_PITER_INCL_EMPTY | DISK_PITER_REVERSE);
650 while ((part = disk_part_iter_next(&piter))) {
651 invalidate_partition(disk, part->partno);
652 delete_partition(disk, part->partno);
653 }
654 disk_part_iter_exit(&piter);
655
656 invalidate_partition(disk, 0);
Tejun Heod2bf1b62010-12-08 20:57:36 +0100657 set_capacity(disk, 0);
658 disk->flags &= ~GENHD_FL_UP;
659
Tejun Heoed9e1982008-08-25 19:56:05 +0900660 sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661 blk_unregister_queue(disk);
Tejun Heof331c022008-09-03 09:01:48 +0200662 blk_unregister_region(disk_devt(disk), disk->minors);
Tejun Heod2bf1b62010-12-08 20:57:36 +0100663
664 part_stat_set_all(&disk->part0, 0);
665 disk->part0.stamp = 0;
666
667 kobject_put(disk->part0.holder_dir);
668 kobject_put(disk->slave_dir);
Tejun Heod2bf1b62010-12-08 20:57:36 +0100669 if (!sysfs_deprecated)
670 sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
Ming Lei25e823c2013-02-22 16:34:13 -0800671 pm_runtime_set_memalloc_noio(disk_to_dev(disk), false);
Tejun Heod2bf1b62010-12-08 20:57:36 +0100672 device_del(disk_to_dev(disk));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673}
Tejun Heod2bf1b62010-12-08 20:57:36 +0100674EXPORT_SYMBOL(del_gendisk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675
Vishal Verma99e66082016-01-09 08:36:51 -0800676/* sysfs access to bad-blocks list. */
677static ssize_t disk_badblocks_show(struct device *dev,
678 struct device_attribute *attr,
679 char *page)
680{
681 struct gendisk *disk = dev_to_disk(dev);
682
683 if (!disk->bb)
684 return sprintf(page, "\n");
685
686 return badblocks_show(disk->bb, page, 0);
687}
688
689static ssize_t disk_badblocks_store(struct device *dev,
690 struct device_attribute *attr,
691 const char *page, size_t len)
692{
693 struct gendisk *disk = dev_to_disk(dev);
694
695 if (!disk->bb)
696 return -ENXIO;
697
698 return badblocks_store(disk->bb, page, len, 0);
699}
700
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701/**
702 * get_gendisk - get partitioning information for a given device
Randy Dunlap710027a2008-08-19 20:13:11 +0200703 * @devt: device to get partitioning information for
Randy Dunlap496aa8a2008-10-16 07:46:23 +0200704 * @partno: returned partition index
Linus Torvalds1da177e2005-04-16 15:20:36 -0700705 *
706 * This function gets the structure containing partitioning
Randy Dunlap710027a2008-08-19 20:13:11 +0200707 * information for the given device @devt.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708 */
Tejun Heocf771cb2008-09-03 09:01:09 +0200709struct gendisk *get_gendisk(dev_t devt, int *partno)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700710{
Tejun Heobcce3de2008-08-25 19:47:22 +0900711 struct gendisk *disk = NULL;
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200712
Tejun Heobcce3de2008-08-25 19:47:22 +0900713 if (MAJOR(devt) != BLOCK_EXT_MAJOR) {
714 struct kobject *kobj;
715
716 kobj = kobj_lookup(bdev_map, devt, partno);
717 if (kobj)
718 disk = dev_to_disk(kobj_to_dev(kobj));
719 } else {
720 struct hd_struct *part;
721
Dan Williams4d66e5e2015-06-10 23:47:14 -0400722 spin_lock_bh(&ext_devt_lock);
Tejun Heo870d6652008-08-25 19:47:25 +0900723 part = idr_find(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
Tejun Heobcce3de2008-08-25 19:47:22 +0900724 if (part && get_disk(part_to_disk(part))) {
725 *partno = part->partno;
726 disk = part_to_disk(part);
727 }
Dan Williams4d66e5e2015-06-10 23:47:14 -0400728 spin_unlock_bh(&ext_devt_lock);
Tejun Heobcce3de2008-08-25 19:47:22 +0900729 }
730
731 return disk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700732}
Divyesh Shahb6ac23af2010-04-15 08:54:59 +0200733EXPORT_SYMBOL(get_gendisk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734
Tejun Heof331c022008-09-03 09:01:48 +0200735/**
736 * bdget_disk - do bdget() by gendisk and partition number
737 * @disk: gendisk of interest
738 * @partno: partition number
739 *
740 * Find partition @partno from @disk, do bdget() on it.
741 *
742 * CONTEXT:
743 * Don't care.
744 *
745 * RETURNS:
746 * Resulting block_device on success, NULL on failure.
747 */
Harvey Harrisonaeb3d3a2008-08-28 09:27:42 +0200748struct block_device *bdget_disk(struct gendisk *disk, int partno)
Tejun Heof331c022008-09-03 09:01:48 +0200749{
Tejun Heo548b10e2008-08-29 09:01:47 +0200750 struct hd_struct *part;
751 struct block_device *bdev = NULL;
Tejun Heof331c022008-09-03 09:01:48 +0200752
Tejun Heo548b10e2008-08-29 09:01:47 +0200753 part = disk_get_part(disk, partno);
Tejun Heo2bbedcb2008-08-29 11:41:51 +0200754 if (part)
Tejun Heo548b10e2008-08-29 09:01:47 +0200755 bdev = bdget(part_devt(part));
756 disk_put_part(part);
Tejun Heof331c022008-09-03 09:01:48 +0200757
Tejun Heo548b10e2008-08-29 09:01:47 +0200758 return bdev;
Tejun Heof331c022008-09-03 09:01:48 +0200759}
760EXPORT_SYMBOL(bdget_disk);
761
Dave Gilbertdd2a3452007-05-09 02:33:24 -0700762/*
763 * print a full list of all partitions - intended for places where the root
764 * filesystem can't be mounted and thus to give the victim some idea of what
765 * went wrong
766 */
767void __init printk_all_partitions(void)
768{
Tejun Heodef4e382008-09-03 08:57:12 +0200769 struct class_dev_iter iter;
770 struct device *dev;
771
772 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
773 while ((dev = class_dev_iter_next(&iter))) {
774 struct gendisk *disk = dev_to_disk(dev);
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200775 struct disk_part_iter piter;
776 struct hd_struct *part;
Tejun Heo1f014292008-08-25 19:47:23 +0900777 char name_buf[BDEVNAME_SIZE];
778 char devt_buf[BDEVT_SIZE];
Tejun Heodef4e382008-09-03 08:57:12 +0200779
780 /*
781 * Don't show empty devices or things that have been
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300782 * suppressed
Tejun Heodef4e382008-09-03 08:57:12 +0200783 */
784 if (get_capacity(disk) == 0 ||
785 (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
786 continue;
787
788 /*
789 * Note, unlike /proc/partitions, I am showing the
790 * numbers in hex - the same format as the root=
791 * option takes.
792 */
Tejun Heo074a7ac2008-08-25 19:56:14 +0900793 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0);
794 while ((part = disk_part_iter_next(&piter))) {
795 bool is_part0 = part == &disk->part0;
Tejun Heodef4e382008-09-03 08:57:12 +0200796
Will Drewryb5af9212010-08-31 15:47:07 -0500797 printk("%s%s %10llu %s %s", is_part0 ? "" : " ",
Tejun Heo1f014292008-08-25 19:47:23 +0900798 bdevt_str(part_devt(part), devt_buf),
Vivek Goyalc83f6bf2012-08-01 12:24:18 +0200799 (unsigned long long)part_nr_sects_read(part) >> 1
800 , disk_name(disk, part->partno, name_buf),
Stephen Warren1ad7e892012-11-08 16:12:25 -0800801 part->info ? part->info->uuid : "");
Tejun Heo074a7ac2008-08-25 19:56:14 +0900802 if (is_part0) {
Dan Williams52c44d92016-06-15 19:43:07 -0700803 if (dev->parent && dev->parent->driver)
Tejun Heo074a7ac2008-08-25 19:56:14 +0900804 printk(" driver: %s\n",
Dan Williams52c44d92016-06-15 19:43:07 -0700805 dev->parent->driver->name);
Tejun Heo074a7ac2008-08-25 19:56:14 +0900806 else
807 printk(" (driver?)\n");
808 } else
809 printk("\n");
810 }
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200811 disk_part_iter_exit(&piter);
Tejun Heodef4e382008-09-03 08:57:12 +0200812 }
813 class_dev_iter_exit(&iter);
Dave Gilbertdd2a3452007-05-09 02:33:24 -0700814}
815
Linus Torvalds1da177e2005-04-16 15:20:36 -0700816#ifdef CONFIG_PROC_FS
817/* iterator */
Tejun Heodef4e382008-09-03 08:57:12 +0200818static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
Greg Kroah-Hartman68c4d4a2008-05-22 17:21:08 -0400819{
Tejun Heodef4e382008-09-03 08:57:12 +0200820 loff_t skip = *pos;
821 struct class_dev_iter *iter;
822 struct device *dev;
Greg Kroah-Hartman68c4d4a2008-05-22 17:21:08 -0400823
Harvey Harrisonaeb3d3a2008-08-28 09:27:42 +0200824 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
Tejun Heodef4e382008-09-03 08:57:12 +0200825 if (!iter)
826 return ERR_PTR(-ENOMEM);
827
828 seqf->private = iter;
829 class_dev_iter_init(iter, &block_class, NULL, &disk_type);
830 do {
831 dev = class_dev_iter_next(iter);
832 if (!dev)
833 return NULL;
834 } while (skip--);
835
836 return dev_to_disk(dev);
Greg Kroah-Hartman68c4d4a2008-05-22 17:21:08 -0400837}
838
Tejun Heodef4e382008-09-03 08:57:12 +0200839static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200841 struct device *dev;
Greg Kroah-Hartman66c64af2008-05-22 17:21:08 -0400842
Tejun Heodef4e382008-09-03 08:57:12 +0200843 (*pos)++;
844 dev = class_dev_iter_next(seqf->private);
Tejun Heo2ac3cee2008-09-03 08:53:37 +0200845 if (dev)
Greg Kroah-Hartman68c4d4a2008-05-22 17:21:08 -0400846 return dev_to_disk(dev);
Tejun Heo2ac3cee2008-09-03 08:53:37 +0200847
Linus Torvalds1da177e2005-04-16 15:20:36 -0700848 return NULL;
849}
850
Tejun Heodef4e382008-09-03 08:57:12 +0200851static void disk_seqf_stop(struct seq_file *seqf, void *v)
Greg Kroah-Hartman27f30252008-05-22 17:21:08 -0400852{
Tejun Heodef4e382008-09-03 08:57:12 +0200853 struct class_dev_iter *iter = seqf->private;
Greg Kroah-Hartman27f30252008-05-22 17:21:08 -0400854
Tejun Heodef4e382008-09-03 08:57:12 +0200855 /* stop is called even after start failed :-( */
856 if (iter) {
857 class_dev_iter_exit(iter);
858 kfree(iter);
Vegard Nossum77da1602016-07-29 10:40:31 +0200859 seqf->private = NULL;
Kay Sievers5c0ef6d2008-08-16 14:30:30 +0200860 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861}
862
Tejun Heodef4e382008-09-03 08:57:12 +0200863static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864{
Jianpeng Ma06768062012-08-03 10:42:00 +0200865 void *p;
Tejun Heodef4e382008-09-03 08:57:12 +0200866
867 p = disk_seqf_start(seqf, pos);
Yang Zhangb9f985b2010-12-17 08:58:36 +0100868 if (!IS_ERR_OR_NULL(p) && !*pos)
Tejun Heodef4e382008-09-03 08:57:12 +0200869 seq_puts(seqf, "major minor #blocks name\n\n");
870 return p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871}
872
Tejun Heocf771cb2008-09-03 09:01:09 +0200873static int show_partition(struct seq_file *seqf, void *v)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700874{
875 struct gendisk *sgp = v;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200876 struct disk_part_iter piter;
877 struct hd_struct *part;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878 char buf[BDEVNAME_SIZE];
879
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880 /* Don't show non-partitionable removeable devices or empty devices */
Tejun Heod27769e2011-08-23 20:01:04 +0200881 if (!get_capacity(sgp) || (!disk_max_parts(sgp) &&
Tejun Heof331c022008-09-03 09:01:48 +0200882 (sgp->flags & GENHD_FL_REMOVABLE)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883 return 0;
884 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
885 return 0;
886
887 /* show the full disk and all non-0 size partitions of it */
Tejun Heo074a7ac2008-08-25 19:56:14 +0900888 disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0);
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200889 while ((part = disk_part_iter_next(&piter)))
Tejun Heo1f014292008-08-25 19:47:23 +0900890 seq_printf(seqf, "%4d %7d %10llu %s\n",
Tejun Heof331c022008-09-03 09:01:48 +0200891 MAJOR(part_devt(part)), MINOR(part_devt(part)),
Vivek Goyalc83f6bf2012-08-01 12:24:18 +0200892 (unsigned long long)part_nr_sects_read(part) >> 1,
Tejun Heof331c022008-09-03 09:01:48 +0200893 disk_name(sgp, part->partno, buf));
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200894 disk_part_iter_exit(&piter);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895
896 return 0;
897}
898
Alexey Dobriyanf5009752008-10-04 23:53:21 +0400899static const struct seq_operations partitions_op = {
Tejun Heodef4e382008-09-03 08:57:12 +0200900 .start = show_partition_start,
901 .next = disk_seqf_next,
902 .stop = disk_seqf_stop,
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200903 .show = show_partition
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904};
Alexey Dobriyanf5009752008-10-04 23:53:21 +0400905
906static int partitions_open(struct inode *inode, struct file *file)
907{
908 return seq_open(file, &partitions_op);
909}
910
911static const struct file_operations proc_partitions_operations = {
912 .open = partitions_open,
913 .read = seq_read,
914 .llseek = seq_lseek,
915 .release = seq_release,
916};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700917#endif
918
919
Tejun Heocf771cb2008-09-03 09:01:09 +0200920static struct kobject *base_probe(dev_t devt, int *partno, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700921{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200922 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700923 /* Make old-style 2.4 aliases work */
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200924 request_module("block-major-%d", MAJOR(devt));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700925 return NULL;
926}
927
928static int __init genhd_device_init(void)
929{
Dan Williamse105b8b2008-04-21 10:51:07 -0700930 int error;
931
932 block_class.dev_kobj = sysfs_dev_block_kobj;
933 error = class_register(&block_class);
Roland McGrathee27a552008-03-11 17:13:15 -0700934 if (unlikely(error))
935 return error;
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200936 bdev_map = kobj_map_init(base_probe, &block_class_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700937 blk_dev_init();
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200938
Zhang, Yanmin561ec682008-11-14 08:26:30 +0100939 register_blkdev(BLOCK_EXT_MAJOR, "blkext");
940
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200941 /* create top-level block dir */
Andi Kleene52eec12010-09-08 16:54:17 +0200942 if (!sysfs_deprecated)
943 block_depr = kobject_create_and_add("block", NULL);
Greg Kroah-Hartman830d3cf2007-11-06 10:36:58 -0800944 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700945}
946
947subsys_initcall(genhd_device_init);
948
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200949static ssize_t disk_range_show(struct device *dev,
950 struct device_attribute *attr, char *buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700951{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200952 struct gendisk *disk = dev_to_disk(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700953
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200954 return sprintf(buf, "%d\n", disk->minors);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955}
956
Tejun Heo1f014292008-08-25 19:47:23 +0900957static ssize_t disk_ext_range_show(struct device *dev,
958 struct device_attribute *attr, char *buf)
959{
960 struct gendisk *disk = dev_to_disk(dev);
961
Tejun Heob5d0b9d2008-09-03 09:06:42 +0200962 return sprintf(buf, "%d\n", disk_max_parts(disk));
Tejun Heo1f014292008-08-25 19:47:23 +0900963}
964
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200965static ssize_t disk_removable_show(struct device *dev,
966 struct device_attribute *attr, char *buf)
Kay Sieversa7fd6702005-10-01 14:49:43 +0200967{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200968 struct gendisk *disk = dev_to_disk(dev);
Kay Sieversa7fd6702005-10-01 14:49:43 +0200969
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200970 return sprintf(buf, "%d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700971 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200972}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973
Kay Sievers1c9ce522008-06-13 09:41:00 +0200974static ssize_t disk_ro_show(struct device *dev,
975 struct device_attribute *attr, char *buf)
976{
977 struct gendisk *disk = dev_to_disk(dev);
978
Tejun Heob7db9952008-08-25 19:56:10 +0900979 return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
Kay Sievers1c9ce522008-06-13 09:41:00 +0200980}
981
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200982static ssize_t disk_capability_show(struct device *dev,
983 struct device_attribute *attr, char *buf)
Kristen Carlson Accardi86ce18d2007-05-23 13:57:38 -0700984{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200985 struct gendisk *disk = dev_to_disk(dev);
986
987 return sprintf(buf, "%x\n", disk->flags);
Kristen Carlson Accardi86ce18d2007-05-23 13:57:38 -0700988}
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200989
Martin K. Petersenc72758f2009-05-22 17:17:53 -0400990static ssize_t disk_alignment_offset_show(struct device *dev,
991 struct device_attribute *attr,
992 char *buf)
993{
994 struct gendisk *disk = dev_to_disk(dev);
995
996 return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue));
997}
998
Martin K. Petersen86b37282009-11-10 11:50:21 +0100999static ssize_t disk_discard_alignment_show(struct device *dev,
1000 struct device_attribute *attr,
1001 char *buf)
1002{
1003 struct gendisk *disk = dev_to_disk(dev);
1004
Martin K. Petersendd3d1452010-01-11 03:21:48 -05001005 return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue));
Martin K. Petersen86b37282009-11-10 11:50:21 +01001006}
1007
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001008static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL);
Tejun Heo1f014292008-08-25 19:47:23 +09001009static DEVICE_ATTR(ext_range, S_IRUGO, disk_ext_range_show, NULL);
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001010static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL);
Kay Sievers1c9ce522008-06-13 09:41:00 +02001011static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL);
Tejun Heoe5610522008-08-25 19:56:09 +09001012static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL);
Martin K. Petersenc72758f2009-05-22 17:17:53 -04001013static DEVICE_ATTR(alignment_offset, S_IRUGO, disk_alignment_offset_show, NULL);
Martin K. Petersen86b37282009-11-10 11:50:21 +01001014static DEVICE_ATTR(discard_alignment, S_IRUGO, disk_discard_alignment_show,
1015 NULL);
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001016static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL);
Tejun Heo074a7ac2008-08-25 19:56:14 +09001017static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL);
Nikanth Karthikesan316d3152009-10-06 20:16:55 +02001018static DEVICE_ATTR(inflight, S_IRUGO, part_inflight_show, NULL);
Vishal Verma99e66082016-01-09 08:36:51 -08001019static DEVICE_ATTR(badblocks, S_IRUGO | S_IWUSR, disk_badblocks_show,
1020 disk_badblocks_store);
Akinobu Mitac17bb492006-12-08 02:39:46 -08001021#ifdef CONFIG_FAIL_MAKE_REQUEST
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001022static struct device_attribute dev_attr_fail =
Tejun Heoeddb2e22008-08-25 19:56:13 +09001023 __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store);
Akinobu Mitac17bb492006-12-08 02:39:46 -08001024#endif
Jens Axboe581d4e22008-09-14 05:56:33 -07001025#ifdef CONFIG_FAIL_IO_TIMEOUT
1026static struct device_attribute dev_attr_fail_timeout =
1027 __ATTR(io-timeout-fail, S_IRUGO|S_IWUSR, part_timeout_show,
1028 part_timeout_store);
1029#endif
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001030
1031static struct attribute *disk_attrs[] = {
1032 &dev_attr_range.attr,
Tejun Heo1f014292008-08-25 19:47:23 +09001033 &dev_attr_ext_range.attr,
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001034 &dev_attr_removable.attr,
Kay Sievers1c9ce522008-06-13 09:41:00 +02001035 &dev_attr_ro.attr,
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001036 &dev_attr_size.attr,
Martin K. Petersenc72758f2009-05-22 17:17:53 -04001037 &dev_attr_alignment_offset.attr,
Martin K. Petersen86b37282009-11-10 11:50:21 +01001038 &dev_attr_discard_alignment.attr,
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001039 &dev_attr_capability.attr,
1040 &dev_attr_stat.attr,
Nikanth Karthikesan316d3152009-10-06 20:16:55 +02001041 &dev_attr_inflight.attr,
Vishal Verma99e66082016-01-09 08:36:51 -08001042 &dev_attr_badblocks.attr,
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001043#ifdef CONFIG_FAIL_MAKE_REQUEST
1044 &dev_attr_fail.attr,
1045#endif
Jens Axboe581d4e22008-09-14 05:56:33 -07001046#ifdef CONFIG_FAIL_IO_TIMEOUT
1047 &dev_attr_fail_timeout.attr,
1048#endif
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001049 NULL
Linus Torvalds1da177e2005-04-16 15:20:36 -07001050};
1051
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001052static struct attribute_group disk_attr_group = {
1053 .attrs = disk_attrs,
1054};
1055
David Brownella4dbd672009-06-24 10:06:31 -07001056static const struct attribute_group *disk_attr_groups[] = {
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001057 &disk_attr_group,
1058 NULL
1059};
1060
Tejun Heo540eed52008-08-25 19:56:15 +09001061/**
1062 * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way
1063 * @disk: disk to replace part_tbl for
1064 * @new_ptbl: new part_tbl to install
1065 *
1066 * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The
1067 * original ptbl is freed using RCU callback.
1068 *
1069 * LOCKING:
1070 * Matching bd_mutx locked.
1071 */
1072static void disk_replace_part_tbl(struct gendisk *disk,
1073 struct disk_part_tbl *new_ptbl)
1074{
1075 struct disk_part_tbl *old_ptbl = disk->part_tbl;
1076
1077 rcu_assign_pointer(disk->part_tbl, new_ptbl);
Jens Axboea6f23652008-10-24 12:52:42 +02001078
1079 if (old_ptbl) {
1080 rcu_assign_pointer(old_ptbl->last_lookup, NULL);
Lai Jiangshan57bdfbf2011-03-18 11:42:58 +08001081 kfree_rcu(old_ptbl, rcu_head);
Jens Axboea6f23652008-10-24 12:52:42 +02001082 }
Tejun Heo540eed52008-08-25 19:56:15 +09001083}
1084
1085/**
1086 * disk_expand_part_tbl - expand disk->part_tbl
1087 * @disk: disk to expand part_tbl for
1088 * @partno: expand such that this partno can fit in
1089 *
1090 * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl
1091 * uses RCU to allow unlocked dereferencing for stats and other stuff.
1092 *
1093 * LOCKING:
1094 * Matching bd_mutex locked, might sleep.
1095 *
1096 * RETURNS:
1097 * 0 on success, -errno on failure.
1098 */
1099int disk_expand_part_tbl(struct gendisk *disk, int partno)
1100{
1101 struct disk_part_tbl *old_ptbl = disk->part_tbl;
1102 struct disk_part_tbl *new_ptbl;
1103 int len = old_ptbl ? old_ptbl->len : 0;
Jens Axboe5fabcb42014-11-19 13:06:22 -07001104 int i, target;
Tejun Heo540eed52008-08-25 19:56:15 +09001105 size_t size;
Jens Axboe5fabcb42014-11-19 13:06:22 -07001106
1107 /*
1108 * check for int overflow, since we can get here from blkpg_ioctl()
1109 * with a user passed 'partno'.
1110 */
1111 target = partno + 1;
1112 if (target < 0)
1113 return -EINVAL;
Tejun Heo540eed52008-08-25 19:56:15 +09001114
1115 /* disk_max_parts() is zero during initialization, ignore if so */
1116 if (disk_max_parts(disk) && target > disk_max_parts(disk))
1117 return -EINVAL;
1118
1119 if (target <= len)
1120 return 0;
1121
1122 size = sizeof(*new_ptbl) + target * sizeof(new_ptbl->part[0]);
1123 new_ptbl = kzalloc_node(size, GFP_KERNEL, disk->node_id);
1124 if (!new_ptbl)
1125 return -ENOMEM;
1126
Tejun Heo540eed52008-08-25 19:56:15 +09001127 new_ptbl->len = target;
1128
1129 for (i = 0; i < len; i++)
1130 rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]);
1131
1132 disk_replace_part_tbl(disk, new_ptbl);
1133 return 0;
1134}
1135
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001136static void disk_release(struct device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001137{
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001138 struct gendisk *disk = dev_to_disk(dev);
1139
Keith Busch2da78092014-08-26 09:05:36 -06001140 blk_free_devt(dev->devt);
Tejun Heo77ea8872010-12-08 20:57:37 +01001141 disk_release_events(disk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001142 kfree(disk->random);
Tejun Heo540eed52008-08-25 19:56:15 +09001143 disk_replace_part_tbl(disk, NULL);
Ming Leib54e5ed2015-07-16 11:16:44 +08001144 hd_free_part(&disk->part0);
Tejun Heo523e1d32011-10-19 14:31:07 +02001145 if (disk->queue)
1146 blk_put_queue(disk->queue);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001147 kfree(disk);
1148}
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001149struct class block_class = {
1150 .name = "block",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001151};
1152
Kay Sievers3c2670e2013-04-06 09:56:00 -07001153static char *block_devnode(struct device *dev, umode_t *mode,
Greg Kroah-Hartman4e4098a2013-04-11 11:43:29 -07001154 kuid_t *uid, kgid_t *gid)
Kay Sieversb03f38b2009-04-30 15:23:42 +02001155{
1156 struct gendisk *disk = dev_to_disk(dev);
1157
Kay Sieverse454cea2009-09-18 23:01:12 +02001158 if (disk->devnode)
1159 return disk->devnode(disk, mode);
Kay Sieversb03f38b2009-04-30 15:23:42 +02001160 return NULL;
1161}
1162
Adrian Bunk1826ead2008-03-04 11:23:46 +01001163static struct device_type disk_type = {
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001164 .name = "disk",
1165 .groups = disk_attr_groups,
1166 .release = disk_release,
Kay Sieverse454cea2009-09-18 23:01:12 +02001167 .devnode = block_devnode,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168};
1169
Randy Dunlapa6e2ba82008-05-23 09:44:11 -07001170#ifdef CONFIG_PROC_FS
Tejun Heocf771cb2008-09-03 09:01:09 +02001171/*
1172 * aggregate disk stat collector. Uses the same stats that the sysfs
1173 * entries do, above, but makes them available through one seq_file.
1174 *
1175 * The output looks suspiciously like /proc/partitions with a bunch of
1176 * extra fields.
1177 */
1178static int diskstats_show(struct seq_file *seqf, void *v)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001179{
1180 struct gendisk *gp = v;
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001181 struct disk_part_iter piter;
1182 struct hd_struct *hd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183 char buf[BDEVNAME_SIZE];
Tejun Heoc9959052008-08-25 19:47:21 +09001184 int cpu;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001185
1186 /*
Tejun Heoed9e1982008-08-25 19:56:05 +09001187 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
Tejun Heocf771cb2008-09-03 09:01:09 +02001188 seq_puts(seqf, "major minor name"
Linus Torvalds1da177e2005-04-16 15:20:36 -07001189 " rio rmerge rsect ruse wio wmerge "
1190 "wsect wuse running use aveq"
1191 "\n\n");
1192 */
Wanlong Gao9f5e4862011-06-13 10:45:43 +02001193
Tejun Heo71982a42009-04-17 08:34:48 +02001194 disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0);
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001195 while ((hd = disk_part_iter_next(&piter))) {
Tejun Heo074a7ac2008-08-25 19:56:14 +09001196 cpu = part_stat_lock();
Tejun Heoc9959052008-08-25 19:47:21 +09001197 part_round_stats(cpu, hd);
Tejun Heo074a7ac2008-08-25 19:56:14 +09001198 part_stat_unlock();
Herbert Poetzlf95fe9c2011-08-02 12:43:50 +02001199 seq_printf(seqf, "%4d %7d %s %lu %lu %lu "
1200 "%u %lu %lu %lu %u %u %u %u\n",
Tejun Heof331c022008-09-03 09:01:48 +02001201 MAJOR(part_devt(hd)), MINOR(part_devt(hd)),
1202 disk_name(gp, hd->partno, buf),
Liu Yuan53f22952011-03-02 11:00:15 -05001203 part_stat_read(hd, ios[READ]),
1204 part_stat_read(hd, merges[READ]),
Herbert Poetzlf95fe9c2011-08-02 12:43:50 +02001205 part_stat_read(hd, sectors[READ]),
Liu Yuan53f22952011-03-02 11:00:15 -05001206 jiffies_to_msecs(part_stat_read(hd, ticks[READ])),
1207 part_stat_read(hd, ios[WRITE]),
1208 part_stat_read(hd, merges[WRITE]),
Herbert Poetzlf95fe9c2011-08-02 12:43:50 +02001209 part_stat_read(hd, sectors[WRITE]),
Liu Yuan53f22952011-03-02 11:00:15 -05001210 jiffies_to_msecs(part_stat_read(hd, ticks[WRITE])),
Nikanth Karthikesan316d3152009-10-06 20:16:55 +02001211 part_in_flight(hd),
Jerome Marchand28f39d52008-02-08 11:04:56 +01001212 jiffies_to_msecs(part_stat_read(hd, io_ticks)),
1213 jiffies_to_msecs(part_stat_read(hd, time_in_queue))
1214 );
Linus Torvalds1da177e2005-04-16 15:20:36 -07001215 }
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001216 disk_part_iter_exit(&piter);
Wanlong Gao9f5e4862011-06-13 10:45:43 +02001217
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218 return 0;
1219}
1220
Alexey Dobriyan31d85ab2008-10-06 12:55:38 +04001221static const struct seq_operations diskstats_op = {
Tejun Heodef4e382008-09-03 08:57:12 +02001222 .start = disk_seqf_start,
1223 .next = disk_seqf_next,
1224 .stop = disk_seqf_stop,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001225 .show = diskstats_show
1226};
Alexey Dobriyanf5009752008-10-04 23:53:21 +04001227
Alexey Dobriyan31d85ab2008-10-06 12:55:38 +04001228static int diskstats_open(struct inode *inode, struct file *file)
1229{
1230 return seq_open(file, &diskstats_op);
1231}
1232
1233static const struct file_operations proc_diskstats_operations = {
1234 .open = diskstats_open,
1235 .read = seq_read,
1236 .llseek = seq_lseek,
1237 .release = seq_release,
1238};
1239
Alexey Dobriyanf5009752008-10-04 23:53:21 +04001240static int __init proc_genhd_init(void)
1241{
Alexey Dobriyan31d85ab2008-10-06 12:55:38 +04001242 proc_create("diskstats", 0, NULL, &proc_diskstats_operations);
Alexey Dobriyanf5009752008-10-04 23:53:21 +04001243 proc_create("partitions", 0, NULL, &proc_partitions_operations);
1244 return 0;
1245}
1246module_init(proc_genhd_init);
Randy Dunlapa6e2ba82008-05-23 09:44:11 -07001247#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001248
Tejun Heocf771cb2008-09-03 09:01:09 +02001249dev_t blk_lookup_devt(const char *name, int partno)
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001250{
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001251 dev_t devt = MKDEV(0, 0);
Tejun Heodef4e382008-09-03 08:57:12 +02001252 struct class_dev_iter iter;
1253 struct device *dev;
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001254
Tejun Heodef4e382008-09-03 08:57:12 +02001255 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1256 while ((dev = class_dev_iter_next(&iter))) {
1257 struct gendisk *disk = dev_to_disk(dev);
Tejun Heo548b10e2008-08-29 09:01:47 +02001258 struct hd_struct *part;
Tejun Heodef4e382008-09-03 08:57:12 +02001259
Kay Sievers3ada8b72009-01-06 10:44:43 -08001260 if (strcmp(dev_name(dev), name))
Tejun Heof331c022008-09-03 09:01:48 +02001261 continue;
Tejun Heof331c022008-09-03 09:01:48 +02001262
Neil Brown41b8c852009-02-18 10:33:59 +01001263 if (partno < disk->minors) {
1264 /* We need to return the right devno, even
1265 * if the partition doesn't exist yet.
1266 */
1267 devt = MKDEV(MAJOR(dev->devt),
1268 MINOR(dev->devt) + partno);
1269 break;
1270 }
Tejun Heo548b10e2008-08-29 09:01:47 +02001271 part = disk_get_part(disk, partno);
Tejun Heo2bbedcb2008-08-29 11:41:51 +02001272 if (part) {
Tejun Heof331c022008-09-03 09:01:48 +02001273 devt = part_devt(part);
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001274 disk_put_part(part);
Tejun Heo548b10e2008-08-29 09:01:47 +02001275 break;
Tejun Heodef4e382008-09-03 08:57:12 +02001276 }
Tejun Heo548b10e2008-08-29 09:01:47 +02001277 disk_put_part(part);
Kay Sievers5c0ef6d2008-08-16 14:30:30 +02001278 }
Tejun Heodef4e382008-09-03 08:57:12 +02001279 class_dev_iter_exit(&iter);
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001280 return devt;
1281}
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001282EXPORT_SYMBOL(blk_lookup_devt);
1283
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284struct gendisk *alloc_disk(int minors)
1285{
Ezequiel Garciac304a512012-11-10 10:39:44 +01001286 return alloc_disk_node(minors, NUMA_NO_NODE);
Christoph Lameter19460892005-06-23 00:08:19 -07001287}
Tejun Heo689d6fa2008-08-25 19:56:16 +09001288EXPORT_SYMBOL(alloc_disk);
Christoph Lameter19460892005-06-23 00:08:19 -07001289
1290struct gendisk *alloc_disk_node(int minors, int node_id)
1291{
1292 struct gendisk *disk;
1293
Joe Perchesc1b511e2013-08-29 15:21:42 -07001294 disk = kzalloc_node(sizeof(struct gendisk), GFP_KERNEL, node_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001295 if (disk) {
Tejun Heo074a7ac2008-08-25 19:56:14 +09001296 if (!init_part_stats(&disk->part0)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001297 kfree(disk);
1298 return NULL;
1299 }
Cheng Renquanbf91db12008-11-20 08:37:37 +01001300 disk->node_id = node_id;
Tejun Heo540eed52008-08-25 19:56:15 +09001301 if (disk_expand_part_tbl(disk, 0)) {
1302 free_part_stats(&disk->part0);
Tejun Heob5d0b9d2008-09-03 09:06:42 +02001303 kfree(disk);
1304 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001305 }
Tejun Heo540eed52008-08-25 19:56:15 +09001306 disk->part_tbl->part[0] = &disk->part0;
Jens Axboe6c23a962011-01-07 08:43:37 +01001307
Vivek Goyalc83f6bf2012-08-01 12:24:18 +02001308 /*
1309 * set_capacity() and get_capacity() currently don't use
1310 * seqcounter to read/update the part0->nr_sects. Still init
1311 * the counter as we can read the sectors in IO submission
1312 * patch using seqence counters.
1313 *
1314 * TODO: Ideally set_capacity() and get_capacity() should be
1315 * converted to make use of bd_mutex and sequence counters.
1316 */
1317 seqcount_init(&disk->part0.nr_sects_seq);
Ming Lei6c710132015-07-16 11:16:45 +08001318 if (hd_ref_init(&disk->part0)) {
1319 hd_free_part(&disk->part0);
1320 kfree(disk);
1321 return NULL;
1322 }
Tejun Heob5d0b9d2008-09-03 09:06:42 +02001323
Linus Torvalds1da177e2005-04-16 15:20:36 -07001324 disk->minors = minors;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325 rand_initialize_disk(disk);
Tejun Heoed9e1982008-08-25 19:56:05 +09001326 disk_to_dev(disk)->class = &block_class;
1327 disk_to_dev(disk)->type = &disk_type;
1328 device_initialize(disk_to_dev(disk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001329 }
1330 return disk;
1331}
Christoph Lameter19460892005-06-23 00:08:19 -07001332EXPORT_SYMBOL(alloc_disk_node);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001333
1334struct kobject *get_disk(struct gendisk *disk)
1335{
1336 struct module *owner;
1337 struct kobject *kobj;
1338
1339 if (!disk->fops)
1340 return NULL;
1341 owner = disk->fops->owner;
1342 if (owner && !try_module_get(owner))
1343 return NULL;
Tejun Heoed9e1982008-08-25 19:56:05 +09001344 kobj = kobject_get(&disk_to_dev(disk)->kobj);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001345 if (kobj == NULL) {
1346 module_put(owner);
1347 return NULL;
1348 }
1349 return kobj;
1350
1351}
1352
1353EXPORT_SYMBOL(get_disk);
1354
1355void put_disk(struct gendisk *disk)
1356{
1357 if (disk)
Tejun Heoed9e1982008-08-25 19:56:05 +09001358 kobject_put(&disk_to_dev(disk)->kobj);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001359}
1360
1361EXPORT_SYMBOL(put_disk);
1362
Hannes Reineckee3264a42009-07-28 09:13:13 +02001363static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1364{
1365 char event[] = "DISK_RO=1";
1366 char *envp[] = { event, NULL };
1367
1368 if (!ro)
1369 event[8] = '0';
1370 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1371}
1372
Linus Torvalds1da177e2005-04-16 15:20:36 -07001373void set_device_ro(struct block_device *bdev, int flag)
1374{
Tejun Heob7db9952008-08-25 19:56:10 +09001375 bdev->bd_part->policy = flag;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001376}
1377
1378EXPORT_SYMBOL(set_device_ro);
1379
1380void set_disk_ro(struct gendisk *disk, int flag)
1381{
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001382 struct disk_part_iter piter;
1383 struct hd_struct *part;
1384
Hannes Reineckee3264a42009-07-28 09:13:13 +02001385 if (disk->part0.policy != flag) {
1386 set_disk_ro_uevent(disk, flag);
1387 disk->part0.policy = flag;
1388 }
1389
1390 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY);
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001391 while ((part = disk_part_iter_next(&piter)))
1392 part->policy = flag;
1393 disk_part_iter_exit(&piter);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394}
1395
1396EXPORT_SYMBOL(set_disk_ro);
1397
1398int bdev_read_only(struct block_device *bdev)
1399{
1400 if (!bdev)
1401 return 0;
Tejun Heob7db9952008-08-25 19:56:10 +09001402 return bdev->bd_part->policy;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001403}
1404
1405EXPORT_SYMBOL(bdev_read_only);
1406
Tejun Heocf771cb2008-09-03 09:01:09 +02001407int invalidate_partition(struct gendisk *disk, int partno)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001408{
1409 int res = 0;
Tejun Heocf771cb2008-09-03 09:01:09 +02001410 struct block_device *bdev = bdget_disk(disk, partno);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001411 if (bdev) {
Christoph Hellwig2ef41632005-05-05 16:15:59 -07001412 fsync_bdev(bdev);
NeilBrown93b270f2011-02-24 17:25:47 +11001413 res = __invalidate_device(bdev, true);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001414 bdput(bdev);
1415 }
1416 return res;
1417}
1418
1419EXPORT_SYMBOL(invalidate_partition);
Tejun Heo77ea8872010-12-08 20:57:37 +01001420
1421/*
1422 * Disk events - monitor disk events like media change and eject request.
1423 */
1424struct disk_events {
1425 struct list_head node; /* all disk_event's */
1426 struct gendisk *disk; /* the associated disk */
1427 spinlock_t lock;
1428
Tejun Heofdd514e2011-06-09 20:43:59 +02001429 struct mutex block_mutex; /* protects blocking */
Tejun Heo77ea8872010-12-08 20:57:37 +01001430 int block; /* event blocking depth */
1431 unsigned int pending; /* events already sent out */
1432 unsigned int clearing; /* events being cleared */
1433
1434 long poll_msecs; /* interval, -1 for default */
1435 struct delayed_work dwork;
1436};
1437
1438static const char *disk_events_strs[] = {
1439 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "media_change",
1440 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "eject_request",
1441};
1442
1443static char *disk_uevents[] = {
1444 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "DISK_MEDIA_CHANGE=1",
1445 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "DISK_EJECT_REQUEST=1",
1446};
1447
1448/* list of all disk_events */
1449static DEFINE_MUTEX(disk_events_mutex);
1450static LIST_HEAD(disk_events);
1451
1452/* disable in-kernel polling by default */
Wei Tang1fe8f342015-11-24 09:58:46 +08001453static unsigned long disk_events_dfl_poll_msecs;
Tejun Heo77ea8872010-12-08 20:57:37 +01001454
1455static unsigned long disk_events_poll_jiffies(struct gendisk *disk)
1456{
1457 struct disk_events *ev = disk->ev;
1458 long intv_msecs = 0;
1459
1460 /*
1461 * If device-specific poll interval is set, always use it. If
1462 * the default is being used, poll iff there are events which
1463 * can't be monitored asynchronously.
1464 */
1465 if (ev->poll_msecs >= 0)
1466 intv_msecs = ev->poll_msecs;
1467 else if (disk->events & ~disk->async_events)
1468 intv_msecs = disk_events_dfl_poll_msecs;
1469
1470 return msecs_to_jiffies(intv_msecs);
1471}
1472
Tejun Heoc3af54a2011-06-09 20:43:55 +02001473/**
1474 * disk_block_events - block and flush disk event checking
1475 * @disk: disk to block events for
1476 *
1477 * On return from this function, it is guaranteed that event checking
1478 * isn't in progress and won't happen until unblocked by
1479 * disk_unblock_events(). Events blocking is counted and the actual
1480 * unblocking happens after the matching number of unblocks are done.
1481 *
1482 * Note that this intentionally does not block event checking from
1483 * disk_clear_events().
1484 *
1485 * CONTEXT:
1486 * Might sleep.
1487 */
1488void disk_block_events(struct gendisk *disk)
Tejun Heo77ea8872010-12-08 20:57:37 +01001489{
1490 struct disk_events *ev = disk->ev;
1491 unsigned long flags;
1492 bool cancel;
1493
Tejun Heoc3af54a2011-06-09 20:43:55 +02001494 if (!ev)
1495 return;
1496
Tejun Heofdd514e2011-06-09 20:43:59 +02001497 /*
1498 * Outer mutex ensures that the first blocker completes canceling
1499 * the event work before further blockers are allowed to finish.
1500 */
1501 mutex_lock(&ev->block_mutex);
1502
Tejun Heo77ea8872010-12-08 20:57:37 +01001503 spin_lock_irqsave(&ev->lock, flags);
1504 cancel = !ev->block++;
1505 spin_unlock_irqrestore(&ev->lock, flags);
1506
Tejun Heoc3af54a2011-06-09 20:43:55 +02001507 if (cancel)
1508 cancel_delayed_work_sync(&disk->ev->dwork);
Tejun Heofdd514e2011-06-09 20:43:59 +02001509
1510 mutex_unlock(&ev->block_mutex);
Tejun Heo77ea8872010-12-08 20:57:37 +01001511}
1512
1513static void __disk_unblock_events(struct gendisk *disk, bool check_now)
1514{
1515 struct disk_events *ev = disk->ev;
1516 unsigned long intv;
1517 unsigned long flags;
1518
1519 spin_lock_irqsave(&ev->lock, flags);
1520
1521 if (WARN_ON_ONCE(ev->block <= 0))
1522 goto out_unlock;
1523
1524 if (--ev->block)
1525 goto out_unlock;
1526
Tejun Heo77ea8872010-12-08 20:57:37 +01001527 intv = disk_events_poll_jiffies(disk);
Tejun Heo77ea8872010-12-08 20:57:37 +01001528 if (check_now)
Viresh Kumar695588f2013-04-24 17:12:56 +05301529 queue_delayed_work(system_freezable_power_efficient_wq,
1530 &ev->dwork, 0);
Tejun Heo77ea8872010-12-08 20:57:37 +01001531 else if (intv)
Viresh Kumar695588f2013-04-24 17:12:56 +05301532 queue_delayed_work(system_freezable_power_efficient_wq,
1533 &ev->dwork, intv);
Tejun Heo77ea8872010-12-08 20:57:37 +01001534out_unlock:
1535 spin_unlock_irqrestore(&ev->lock, flags);
1536}
1537
1538/**
Tejun Heo77ea8872010-12-08 20:57:37 +01001539 * disk_unblock_events - unblock disk event checking
1540 * @disk: disk to unblock events for
1541 *
1542 * Undo disk_block_events(). When the block count reaches zero, it
1543 * starts events polling if configured.
1544 *
1545 * CONTEXT:
1546 * Don't care. Safe to call from irq context.
1547 */
1548void disk_unblock_events(struct gendisk *disk)
1549{
1550 if (disk->ev)
Tejun Heofacc31d2011-03-09 19:54:27 +01001551 __disk_unblock_events(disk, false);
Tejun Heo77ea8872010-12-08 20:57:37 +01001552}
1553
1554/**
Tejun Heo85ef06d2011-07-01 16:17:47 +02001555 * disk_flush_events - schedule immediate event checking and flushing
1556 * @disk: disk to check and flush events for
1557 * @mask: events to flush
Tejun Heo77ea8872010-12-08 20:57:37 +01001558 *
Tejun Heo85ef06d2011-07-01 16:17:47 +02001559 * Schedule immediate event checking on @disk if not blocked. Events in
1560 * @mask are scheduled to be cleared from the driver. Note that this
1561 * doesn't clear the events from @disk->ev.
Tejun Heo77ea8872010-12-08 20:57:37 +01001562 *
1563 * CONTEXT:
Tejun Heo85ef06d2011-07-01 16:17:47 +02001564 * If @mask is non-zero must be called with bdev->bd_mutex held.
Tejun Heo77ea8872010-12-08 20:57:37 +01001565 */
Tejun Heo85ef06d2011-07-01 16:17:47 +02001566void disk_flush_events(struct gendisk *disk, unsigned int mask)
Tejun Heo77ea8872010-12-08 20:57:37 +01001567{
Tejun Heoa9dce2a2011-06-09 20:43:54 +02001568 struct disk_events *ev = disk->ev;
Tejun Heoa9dce2a2011-06-09 20:43:54 +02001569
1570 if (!ev)
1571 return;
1572
Tejun Heo85ef06d2011-07-01 16:17:47 +02001573 spin_lock_irq(&ev->lock);
1574 ev->clearing |= mask;
Tejun Heo41f63c52012-08-03 10:30:47 -07001575 if (!ev->block)
Viresh Kumar695588f2013-04-24 17:12:56 +05301576 mod_delayed_work(system_freezable_power_efficient_wq,
1577 &ev->dwork, 0);
Tejun Heo85ef06d2011-07-01 16:17:47 +02001578 spin_unlock_irq(&ev->lock);
Tejun Heo77ea8872010-12-08 20:57:37 +01001579}
Tejun Heo77ea8872010-12-08 20:57:37 +01001580
1581/**
1582 * disk_clear_events - synchronously check, clear and return pending events
1583 * @disk: disk to fetch and clear events from
Masanari Iidada3dae52014-09-09 01:27:23 +09001584 * @mask: mask of events to be fetched and cleared
Tejun Heo77ea8872010-12-08 20:57:37 +01001585 *
1586 * Disk events are synchronously checked and pending events in @mask
1587 * are cleared and returned. This ignores the block count.
1588 *
1589 * CONTEXT:
1590 * Might sleep.
1591 */
1592unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask)
1593{
1594 const struct block_device_operations *bdops = disk->fops;
1595 struct disk_events *ev = disk->ev;
1596 unsigned int pending;
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001597 unsigned int clearing = mask;
Tejun Heo77ea8872010-12-08 20:57:37 +01001598
1599 if (!ev) {
1600 /* for drivers still using the old ->media_changed method */
1601 if ((mask & DISK_EVENT_MEDIA_CHANGE) &&
1602 bdops->media_changed && bdops->media_changed(disk))
1603 return DISK_EVENT_MEDIA_CHANGE;
1604 return 0;
1605 }
1606
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001607 disk_block_events(disk);
1608
1609 /*
1610 * store the union of mask and ev->clearing on the stack so that the
1611 * race with disk_flush_events does not cause ambiguity (ev->clearing
1612 * can still be modified even if events are blocked).
1613 */
Tejun Heo77ea8872010-12-08 20:57:37 +01001614 spin_lock_irq(&ev->lock);
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001615 clearing |= ev->clearing;
1616 ev->clearing = 0;
Tejun Heo77ea8872010-12-08 20:57:37 +01001617 spin_unlock_irq(&ev->lock);
1618
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001619 disk_check_events(ev, &clearing);
Derek Basehoreaea24a82012-12-18 12:27:18 -08001620 /*
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001621 * if ev->clearing is not 0, the disk_flush_events got called in the
1622 * middle of this function, so we want to run the workfn without delay.
Derek Basehoreaea24a82012-12-18 12:27:18 -08001623 */
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001624 __disk_unblock_events(disk, ev->clearing ? true : false);
Tejun Heo77ea8872010-12-08 20:57:37 +01001625
1626 /* then, fetch and clear pending events */
1627 spin_lock_irq(&ev->lock);
Tejun Heo77ea8872010-12-08 20:57:37 +01001628 pending = ev->pending & mask;
1629 ev->pending &= ~mask;
1630 spin_unlock_irq(&ev->lock);
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001631 WARN_ON_ONCE(clearing & mask);
Tejun Heo77ea8872010-12-08 20:57:37 +01001632
1633 return pending;
1634}
1635
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001636/*
1637 * Separate this part out so that a different pointer for clearing_ptr can be
1638 * passed in for disk_clear_events.
1639 */
Tejun Heo77ea8872010-12-08 20:57:37 +01001640static void disk_events_workfn(struct work_struct *work)
1641{
1642 struct delayed_work *dwork = to_delayed_work(work);
1643 struct disk_events *ev = container_of(dwork, struct disk_events, dwork);
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001644
1645 disk_check_events(ev, &ev->clearing);
1646}
1647
1648static void disk_check_events(struct disk_events *ev,
1649 unsigned int *clearing_ptr)
1650{
Tejun Heo77ea8872010-12-08 20:57:37 +01001651 struct gendisk *disk = ev->disk;
1652 char *envp[ARRAY_SIZE(disk_uevents) + 1] = { };
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001653 unsigned int clearing = *clearing_ptr;
Tejun Heo77ea8872010-12-08 20:57:37 +01001654 unsigned int events;
1655 unsigned long intv;
1656 int nr_events = 0, i;
1657
1658 /* check events */
1659 events = disk->fops->check_events(disk, clearing);
1660
1661 /* accumulate pending events and schedule next poll if necessary */
1662 spin_lock_irq(&ev->lock);
1663
1664 events &= ~ev->pending;
1665 ev->pending |= events;
Derek Basehore12c2bdb2012-12-18 12:27:20 -08001666 *clearing_ptr &= ~clearing;
Tejun Heo77ea8872010-12-08 20:57:37 +01001667
1668 intv = disk_events_poll_jiffies(disk);
1669 if (!ev->block && intv)
Viresh Kumar695588f2013-04-24 17:12:56 +05301670 queue_delayed_work(system_freezable_power_efficient_wq,
1671 &ev->dwork, intv);
Tejun Heo77ea8872010-12-08 20:57:37 +01001672
1673 spin_unlock_irq(&ev->lock);
1674
Tejun Heo7c88a162011-04-21 19:43:58 +02001675 /*
1676 * Tell userland about new events. Only the events listed in
1677 * @disk->events are reported. Unlisted events are processed the
1678 * same internally but never get reported to userland.
1679 */
Tejun Heo77ea8872010-12-08 20:57:37 +01001680 for (i = 0; i < ARRAY_SIZE(disk_uevents); i++)
Tejun Heo7c88a162011-04-21 19:43:58 +02001681 if (events & disk->events & (1 << i))
Tejun Heo77ea8872010-12-08 20:57:37 +01001682 envp[nr_events++] = disk_uevents[i];
1683
1684 if (nr_events)
1685 kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp);
1686}
1687
1688/*
1689 * A disk events enabled device has the following sysfs nodes under
1690 * its /sys/block/X/ directory.
1691 *
1692 * events : list of all supported events
1693 * events_async : list of events which can be detected w/o polling
1694 * events_poll_msecs : polling interval, 0: disable, -1: system default
1695 */
1696static ssize_t __disk_events_show(unsigned int events, char *buf)
1697{
1698 const char *delim = "";
1699 ssize_t pos = 0;
1700 int i;
1701
1702 for (i = 0; i < ARRAY_SIZE(disk_events_strs); i++)
1703 if (events & (1 << i)) {
1704 pos += sprintf(buf + pos, "%s%s",
1705 delim, disk_events_strs[i]);
1706 delim = " ";
1707 }
1708 if (pos)
1709 pos += sprintf(buf + pos, "\n");
1710 return pos;
1711}
1712
1713static ssize_t disk_events_show(struct device *dev,
1714 struct device_attribute *attr, char *buf)
1715{
1716 struct gendisk *disk = dev_to_disk(dev);
1717
1718 return __disk_events_show(disk->events, buf);
1719}
1720
1721static ssize_t disk_events_async_show(struct device *dev,
1722 struct device_attribute *attr, char *buf)
1723{
1724 struct gendisk *disk = dev_to_disk(dev);
1725
1726 return __disk_events_show(disk->async_events, buf);
1727}
1728
1729static ssize_t disk_events_poll_msecs_show(struct device *dev,
1730 struct device_attribute *attr,
1731 char *buf)
1732{
1733 struct gendisk *disk = dev_to_disk(dev);
1734
1735 return sprintf(buf, "%ld\n", disk->ev->poll_msecs);
1736}
1737
1738static ssize_t disk_events_poll_msecs_store(struct device *dev,
1739 struct device_attribute *attr,
1740 const char *buf, size_t count)
1741{
1742 struct gendisk *disk = dev_to_disk(dev);
1743 long intv;
1744
1745 if (!count || !sscanf(buf, "%ld", &intv))
1746 return -EINVAL;
1747
1748 if (intv < 0 && intv != -1)
1749 return -EINVAL;
1750
Tejun Heoc3af54a2011-06-09 20:43:55 +02001751 disk_block_events(disk);
Tejun Heo77ea8872010-12-08 20:57:37 +01001752 disk->ev->poll_msecs = intv;
1753 __disk_unblock_events(disk, true);
1754
1755 return count;
1756}
1757
1758static const DEVICE_ATTR(events, S_IRUGO, disk_events_show, NULL);
1759static const DEVICE_ATTR(events_async, S_IRUGO, disk_events_async_show, NULL);
1760static const DEVICE_ATTR(events_poll_msecs, S_IRUGO|S_IWUSR,
1761 disk_events_poll_msecs_show,
1762 disk_events_poll_msecs_store);
1763
1764static const struct attribute *disk_events_attrs[] = {
1765 &dev_attr_events.attr,
1766 &dev_attr_events_async.attr,
1767 &dev_attr_events_poll_msecs.attr,
1768 NULL,
1769};
1770
1771/*
1772 * The default polling interval can be specified by the kernel
1773 * parameter block.events_dfl_poll_msecs which defaults to 0
1774 * (disable). This can also be modified runtime by writing to
1775 * /sys/module/block/events_dfl_poll_msecs.
1776 */
1777static int disk_events_set_dfl_poll_msecs(const char *val,
1778 const struct kernel_param *kp)
1779{
1780 struct disk_events *ev;
1781 int ret;
1782
1783 ret = param_set_ulong(val, kp);
1784 if (ret < 0)
1785 return ret;
1786
1787 mutex_lock(&disk_events_mutex);
1788
1789 list_for_each_entry(ev, &disk_events, node)
Tejun Heo85ef06d2011-07-01 16:17:47 +02001790 disk_flush_events(ev->disk, 0);
Tejun Heo77ea8872010-12-08 20:57:37 +01001791
1792 mutex_unlock(&disk_events_mutex);
1793
1794 return 0;
1795}
1796
1797static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops = {
1798 .set = disk_events_set_dfl_poll_msecs,
1799 .get = param_get_ulong,
1800};
1801
1802#undef MODULE_PARAM_PREFIX
1803#define MODULE_PARAM_PREFIX "block."
1804
1805module_param_cb(events_dfl_poll_msecs, &disk_events_dfl_poll_msecs_param_ops,
1806 &disk_events_dfl_poll_msecs, 0644);
1807
1808/*
Stanislaw Gruszka9f53d2fe2012-03-02 10:43:28 +01001809 * disk_{alloc|add|del|release}_events - initialize and destroy disk_events.
Tejun Heo77ea8872010-12-08 20:57:37 +01001810 */
Stanislaw Gruszka9f53d2fe2012-03-02 10:43:28 +01001811static void disk_alloc_events(struct gendisk *disk)
Tejun Heo77ea8872010-12-08 20:57:37 +01001812{
1813 struct disk_events *ev;
1814
Tejun Heo75e3f3e2011-05-26 21:06:50 +02001815 if (!disk->fops->check_events)
Tejun Heo77ea8872010-12-08 20:57:37 +01001816 return;
1817
1818 ev = kzalloc(sizeof(*ev), GFP_KERNEL);
1819 if (!ev) {
1820 pr_warn("%s: failed to initialize events\n", disk->disk_name);
1821 return;
1822 }
1823
Tejun Heo77ea8872010-12-08 20:57:37 +01001824 INIT_LIST_HEAD(&ev->node);
1825 ev->disk = disk;
1826 spin_lock_init(&ev->lock);
Tejun Heofdd514e2011-06-09 20:43:59 +02001827 mutex_init(&ev->block_mutex);
Tejun Heo77ea8872010-12-08 20:57:37 +01001828 ev->block = 1;
1829 ev->poll_msecs = -1;
1830 INIT_DELAYED_WORK(&ev->dwork, disk_events_workfn);
1831
Stanislaw Gruszka9f53d2fe2012-03-02 10:43:28 +01001832 disk->ev = ev;
1833}
1834
1835static void disk_add_events(struct gendisk *disk)
1836{
1837 if (!disk->ev)
1838 return;
1839
1840 /* FIXME: error handling */
1841 if (sysfs_create_files(&disk_to_dev(disk)->kobj, disk_events_attrs) < 0)
1842 pr_warn("%s: failed to create sysfs files for events\n",
1843 disk->disk_name);
1844
Tejun Heo77ea8872010-12-08 20:57:37 +01001845 mutex_lock(&disk_events_mutex);
Stanislaw Gruszka9f53d2fe2012-03-02 10:43:28 +01001846 list_add_tail(&disk->ev->node, &disk_events);
Tejun Heo77ea8872010-12-08 20:57:37 +01001847 mutex_unlock(&disk_events_mutex);
1848
1849 /*
1850 * Block count is initialized to 1 and the following initial
1851 * unblock kicks it into action.
1852 */
1853 __disk_unblock_events(disk, true);
1854}
1855
1856static void disk_del_events(struct gendisk *disk)
1857{
1858 if (!disk->ev)
1859 return;
1860
Tejun Heoc3af54a2011-06-09 20:43:55 +02001861 disk_block_events(disk);
Tejun Heo77ea8872010-12-08 20:57:37 +01001862
1863 mutex_lock(&disk_events_mutex);
1864 list_del_init(&disk->ev->node);
1865 mutex_unlock(&disk_events_mutex);
1866
1867 sysfs_remove_files(&disk_to_dev(disk)->kobj, disk_events_attrs);
1868}
1869
1870static void disk_release_events(struct gendisk *disk)
1871{
1872 /* the block count should be 1 from disk_del_events() */
1873 WARN_ON_ONCE(disk->ev && disk->ev->block != 1);
1874 kfree(disk->ev);
1875}