blob: 94855a9717de2237b1eeb72449c32ef12d4a5fef [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * gendisk handling
3 */
4
Linus Torvalds1da177e2005-04-16 15:20:36 -07005#include <linux/module.h>
6#include <linux/fs.h>
7#include <linux/genhd.h>
Andrew Mortonb446b602007-02-20 13:57:48 -08008#include <linux/kdev_t.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -07009#include <linux/kernel.h>
10#include <linux/blkdev.h>
11#include <linux/init.h>
12#include <linux/spinlock.h>
Alexey Dobriyanf5009752008-10-04 23:53:21 +040013#include <linux/proc_fs.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070014#include <linux/seq_file.h>
15#include <linux/slab.h>
16#include <linux/kmod.h>
17#include <linux/kobj_map.h>
Christoph Hellwig2ef41632005-05-05 16:15:59 -070018#include <linux/buffer_head.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>
Nao Nishijimaa72c5e52011-08-25 18:04:06 +090022#include <linux/ctype.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 */
30#define MAX_EXT_DEVT (1 << MINORBITS)
31
32/* For extended devt allocation. ext_devt_mutex prevents look up
33 * results from going away underneath its user.
34 */
35static DEFINE_MUTEX(ext_devt_mutex);
36static DEFINE_IDR(ext_devt_idr);
37
Adrian Bunk1826ead2008-03-04 11:23:46 +010038static struct device_type disk_type;
39
Tejun Heo77ea8872010-12-08 20:57:37 +010040static void disk_add_events(struct gendisk *disk);
41static void disk_del_events(struct gendisk *disk);
42static void disk_release_events(struct gendisk *disk);
43
Tejun Heoe71bf0d2008-09-03 09:03:02 +020044/**
45 * disk_get_part - get partition
46 * @disk: disk to look partition from
47 * @partno: partition number
48 *
49 * Look for partition @partno from @disk. If found, increment
50 * reference count and return it.
51 *
52 * CONTEXT:
53 * Don't care.
54 *
55 * RETURNS:
56 * Pointer to the found partition on success, NULL if not found.
57 */
58struct hd_struct *disk_get_part(struct gendisk *disk, int partno)
59{
Tejun Heo540eed52008-08-25 19:56:15 +090060 struct hd_struct *part = NULL;
61 struct disk_part_tbl *ptbl;
Tejun Heoe71bf0d2008-09-03 09:03:02 +020062
Tejun Heo540eed52008-08-25 19:56:15 +090063 if (unlikely(partno < 0))
Tejun Heoe71bf0d2008-09-03 09:03:02 +020064 return NULL;
Tejun Heo540eed52008-08-25 19:56:15 +090065
Tejun Heoe71bf0d2008-09-03 09:03:02 +020066 rcu_read_lock();
Tejun Heo540eed52008-08-25 19:56:15 +090067
68 ptbl = rcu_dereference(disk->part_tbl);
69 if (likely(partno < ptbl->len)) {
70 part = rcu_dereference(ptbl->part[partno]);
71 if (part)
72 get_device(part_to_dev(part));
73 }
74
Tejun Heoe71bf0d2008-09-03 09:03:02 +020075 rcu_read_unlock();
76
77 return part;
78}
79EXPORT_SYMBOL_GPL(disk_get_part);
80
81/**
82 * disk_part_iter_init - initialize partition iterator
83 * @piter: iterator to initialize
84 * @disk: disk to iterate over
85 * @flags: DISK_PITER_* flags
86 *
87 * Initialize @piter so that it iterates over partitions of @disk.
88 *
89 * CONTEXT:
90 * Don't care.
91 */
92void disk_part_iter_init(struct disk_part_iter *piter, struct gendisk *disk,
93 unsigned int flags)
94{
Tejun Heo540eed52008-08-25 19:56:15 +090095 struct disk_part_tbl *ptbl;
96
97 rcu_read_lock();
98 ptbl = rcu_dereference(disk->part_tbl);
99
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200100 piter->disk = disk;
101 piter->part = NULL;
102
103 if (flags & DISK_PITER_REVERSE)
Tejun Heo540eed52008-08-25 19:56:15 +0900104 piter->idx = ptbl->len - 1;
Tejun Heo71982a42009-04-17 08:34:48 +0200105 else if (flags & (DISK_PITER_INCL_PART0 | DISK_PITER_INCL_EMPTY_PART0))
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200106 piter->idx = 0;
Tejun Heob5d0b9d2008-09-03 09:06:42 +0200107 else
108 piter->idx = 1;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200109
110 piter->flags = flags;
Tejun Heo540eed52008-08-25 19:56:15 +0900111
112 rcu_read_unlock();
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200113}
114EXPORT_SYMBOL_GPL(disk_part_iter_init);
115
116/**
117 * disk_part_iter_next - proceed iterator to the next partition and return it
118 * @piter: iterator of interest
119 *
120 * Proceed @piter to the next partition and return it.
121 *
122 * CONTEXT:
123 * Don't care.
124 */
125struct hd_struct *disk_part_iter_next(struct disk_part_iter *piter)
126{
Tejun Heo540eed52008-08-25 19:56:15 +0900127 struct disk_part_tbl *ptbl;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200128 int inc, end;
129
130 /* put the last partition */
131 disk_put_part(piter->part);
132 piter->part = NULL;
133
Tejun Heo540eed52008-08-25 19:56:15 +0900134 /* get part_tbl */
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200135 rcu_read_lock();
Tejun Heo540eed52008-08-25 19:56:15 +0900136 ptbl = rcu_dereference(piter->disk->part_tbl);
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200137
138 /* determine iteration parameters */
139 if (piter->flags & DISK_PITER_REVERSE) {
140 inc = -1;
Tejun Heo71982a42009-04-17 08:34:48 +0200141 if (piter->flags & (DISK_PITER_INCL_PART0 |
142 DISK_PITER_INCL_EMPTY_PART0))
Tejun Heob5d0b9d2008-09-03 09:06:42 +0200143 end = -1;
144 else
145 end = 0;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200146 } else {
147 inc = 1;
Tejun Heo540eed52008-08-25 19:56:15 +0900148 end = ptbl->len;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200149 }
150
151 /* iterate to the next partition */
152 for (; piter->idx != end; piter->idx += inc) {
153 struct hd_struct *part;
154
Tejun Heo540eed52008-08-25 19:56:15 +0900155 part = rcu_dereference(ptbl->part[piter->idx]);
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200156 if (!part)
157 continue;
Tejun Heo71982a42009-04-17 08:34:48 +0200158 if (!part->nr_sects &&
159 !(piter->flags & DISK_PITER_INCL_EMPTY) &&
160 !(piter->flags & DISK_PITER_INCL_EMPTY_PART0 &&
161 piter->idx == 0))
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200162 continue;
163
Tejun Heoed9e1982008-08-25 19:56:05 +0900164 get_device(part_to_dev(part));
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200165 piter->part = part;
166 piter->idx += inc;
167 break;
168 }
169
170 rcu_read_unlock();
171
172 return piter->part;
173}
174EXPORT_SYMBOL_GPL(disk_part_iter_next);
175
176/**
177 * disk_part_iter_exit - finish up partition iteration
178 * @piter: iter of interest
179 *
180 * Called when iteration is over. Cleans up @piter.
181 *
182 * CONTEXT:
183 * Don't care.
184 */
185void disk_part_iter_exit(struct disk_part_iter *piter)
186{
187 disk_put_part(piter->part);
188 piter->part = NULL;
189}
190EXPORT_SYMBOL_GPL(disk_part_iter_exit);
191
Jens Axboea6f23652008-10-24 12:52:42 +0200192static inline int sector_in_part(struct hd_struct *part, sector_t sector)
193{
194 return part->start_sect <= sector &&
195 sector < part->start_sect + part->nr_sects;
196}
197
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200198/**
199 * disk_map_sector_rcu - map sector to partition
200 * @disk: gendisk of interest
201 * @sector: sector to map
202 *
203 * Find out which partition @sector maps to on @disk. This is
204 * primarily used for stats accounting.
205 *
206 * CONTEXT:
207 * RCU read locked. The returned partition pointer is valid only
208 * while preemption is disabled.
209 *
210 * RETURNS:
Tejun Heo074a7ac2008-08-25 19:56:14 +0900211 * Found partition on success, part0 is returned if no partition matches
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200212 */
213struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector)
214{
Tejun Heo540eed52008-08-25 19:56:15 +0900215 struct disk_part_tbl *ptbl;
Jens Axboea6f23652008-10-24 12:52:42 +0200216 struct hd_struct *part;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200217 int i;
218
Tejun Heo540eed52008-08-25 19:56:15 +0900219 ptbl = rcu_dereference(disk->part_tbl);
220
Jens Axboea6f23652008-10-24 12:52:42 +0200221 part = rcu_dereference(ptbl->last_lookup);
222 if (part && sector_in_part(part, sector))
223 return part;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200224
Jens Axboea6f23652008-10-24 12:52:42 +0200225 for (i = 1; i < ptbl->len; i++) {
226 part = rcu_dereference(ptbl->part[i]);
227
228 if (part && sector_in_part(part, sector)) {
229 rcu_assign_pointer(ptbl->last_lookup, part);
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200230 return part;
Jens Axboea6f23652008-10-24 12:52:42 +0200231 }
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200232 }
Tejun Heo074a7ac2008-08-25 19:56:14 +0900233 return &disk->part0;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200234}
235EXPORT_SYMBOL_GPL(disk_map_sector_rcu);
236
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237/*
238 * Can be deleted altogether. Later.
239 *
240 */
241static struct blk_major_name {
242 struct blk_major_name *next;
243 int major;
244 char name[16];
Joe Korty68eef3b2006-03-31 02:30:32 -0800245} *major_names[BLKDEV_MAJOR_HASH_SIZE];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246
247/* index in the above - for now: assume no multimajor ranges */
Yang Zhange61eb2e2010-12-17 09:00:18 +0100248static inline int major_to_index(unsigned major)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249{
Joe Korty68eef3b2006-03-31 02:30:32 -0800250 return major % BLKDEV_MAJOR_HASH_SIZE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251}
252
Joe Korty68eef3b2006-03-31 02:30:32 -0800253#ifdef CONFIG_PROC_FS
Tejun Heocf771cb2008-09-03 09:01:09 +0200254void blkdev_show(struct seq_file *seqf, off_t offset)
Neil Horman7170be52006-01-14 13:20:38 -0800255{
Joe Korty68eef3b2006-03-31 02:30:32 -0800256 struct blk_major_name *dp;
Neil Horman7170be52006-01-14 13:20:38 -0800257
Joe Korty68eef3b2006-03-31 02:30:32 -0800258 if (offset < BLKDEV_MAJOR_HASH_SIZE) {
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200259 mutex_lock(&block_class_lock);
Joe Korty68eef3b2006-03-31 02:30:32 -0800260 for (dp = major_names[offset]; dp; dp = dp->next)
Tejun Heocf771cb2008-09-03 09:01:09 +0200261 seq_printf(seqf, "%3d %s\n", dp->major, dp->name);
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200262 mutex_unlock(&block_class_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700263 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264}
Joe Korty68eef3b2006-03-31 02:30:32 -0800265#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266
Márton Németh9e8c0bc2009-02-20 08:12:51 +0100267/**
268 * register_blkdev - register a new block device
269 *
270 * @major: the requested major device number [1..255]. If @major=0, try to
271 * allocate any unused major number.
272 * @name: the name of the new block device as a zero terminated string
273 *
274 * The @name must be unique within the system.
275 *
276 * The return value depends on the @major input parameter.
277 * - if a major device number was requested in range [1..255] then the
278 * function returns zero on success, or a negative error code
279 * - if any unused major number was requested with @major=0 parameter
280 * then the return value is the allocated major number in range
281 * [1..255] or a negative error code otherwise
282 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700283int register_blkdev(unsigned int major, const char *name)
284{
285 struct blk_major_name **n, *p;
286 int index, ret = 0;
287
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200288 mutex_lock(&block_class_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289
290 /* temporary */
291 if (major == 0) {
292 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
293 if (major_names[index] == NULL)
294 break;
295 }
296
297 if (index == 0) {
298 printk("register_blkdev: failed to get major for %s\n",
299 name);
300 ret = -EBUSY;
301 goto out;
302 }
303 major = index;
304 ret = major;
305 }
306
307 p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
308 if (p == NULL) {
309 ret = -ENOMEM;
310 goto out;
311 }
312
313 p->major = major;
314 strlcpy(p->name, name, sizeof(p->name));
315 p->next = NULL;
316 index = major_to_index(major);
317
318 for (n = &major_names[index]; *n; n = &(*n)->next) {
319 if ((*n)->major == major)
320 break;
321 }
322 if (!*n)
323 *n = p;
324 else
325 ret = -EBUSY;
326
327 if (ret < 0) {
328 printk("register_blkdev: cannot get major %d for %s\n",
329 major, name);
330 kfree(p);
331 }
332out:
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200333 mutex_unlock(&block_class_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700334 return ret;
335}
336
337EXPORT_SYMBOL(register_blkdev);
338
Akinobu Mitaf4480242007-07-17 04:03:47 -0700339void unregister_blkdev(unsigned int major, const char *name)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340{
341 struct blk_major_name **n;
342 struct blk_major_name *p = NULL;
343 int index = major_to_index(major);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200345 mutex_lock(&block_class_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346 for (n = &major_names[index]; *n; n = &(*n)->next)
347 if ((*n)->major == major)
348 break;
Akinobu Mita294462a2007-07-17 04:03:45 -0700349 if (!*n || strcmp((*n)->name, name)) {
350 WARN_ON(1);
Akinobu Mita294462a2007-07-17 04:03:45 -0700351 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352 p = *n;
353 *n = p->next;
354 }
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200355 mutex_unlock(&block_class_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700356 kfree(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357}
358
359EXPORT_SYMBOL(unregister_blkdev);
360
361static struct kobj_map *bdev_map;
362
Tejun Heobcce3de2008-08-25 19:47:22 +0900363/**
Tejun Heo870d6652008-08-25 19:47:25 +0900364 * blk_mangle_minor - scatter minor numbers apart
365 * @minor: minor number to mangle
366 *
367 * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
368 * is enabled. Mangling twice gives the original value.
369 *
370 * RETURNS:
371 * Mangled value.
372 *
373 * CONTEXT:
374 * Don't care.
375 */
376static int blk_mangle_minor(int minor)
377{
378#ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
379 int i;
380
381 for (i = 0; i < MINORBITS / 2; i++) {
382 int low = minor & (1 << i);
383 int high = minor & (1 << (MINORBITS - 1 - i));
384 int distance = MINORBITS - 1 - 2 * i;
385
386 minor ^= low | high; /* clear both bits */
387 low <<= distance; /* swap the positions */
388 high >>= distance;
389 minor |= low | high; /* and set */
390 }
391#endif
392 return minor;
393}
394
395/**
Tejun Heobcce3de2008-08-25 19:47:22 +0900396 * blk_alloc_devt - allocate a dev_t for a partition
397 * @part: partition to allocate dev_t for
Tejun Heobcce3de2008-08-25 19:47:22 +0900398 * @devt: out parameter for resulting dev_t
399 *
400 * Allocate a dev_t for block device.
401 *
402 * RETURNS:
403 * 0 on success, allocated dev_t is returned in *@devt. -errno on
404 * failure.
405 *
406 * CONTEXT:
407 * Might sleep.
408 */
409int blk_alloc_devt(struct hd_struct *part, dev_t *devt)
410{
411 struct gendisk *disk = part_to_disk(part);
412 int idx, rc;
413
414 /* in consecutive minor range? */
415 if (part->partno < disk->minors) {
416 *devt = MKDEV(disk->major, disk->first_minor + part->partno);
417 return 0;
418 }
419
420 /* allocate ext devt */
421 do {
422 if (!idr_pre_get(&ext_devt_idr, GFP_KERNEL))
423 return -ENOMEM;
424 rc = idr_get_new(&ext_devt_idr, part, &idx);
425 } while (rc == -EAGAIN);
426
427 if (rc)
428 return rc;
429
430 if (idx > MAX_EXT_DEVT) {
431 idr_remove(&ext_devt_idr, idx);
432 return -EBUSY;
433 }
434
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{
450 might_sleep();
451
452 if (devt == MKDEV(0, 0))
453 return;
454
455 if (MAJOR(devt) == BLOCK_EXT_MAJOR) {
456 mutex_lock(&ext_devt_mutex);
Tejun Heo870d6652008-08-25 19:47:25 +0900457 idr_remove(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
Tejun Heobcce3de2008-08-25 19:47:22 +0900458 mutex_unlock(&ext_devt_mutex);
459 }
460}
461
Tejun Heo1f014292008-08-25 19:47:23 +0900462static char *bdevt_str(dev_t devt, char *buf)
463{
464 if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
465 char tbuf[BDEVT_SIZE];
466 snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
467 snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
468 } else
469 snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
470
471 return buf;
472}
473
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474/*
475 * Register device numbers dev..(dev+range-1)
476 * range must be nonzero
477 * The hash chain is sorted on range, so that subranges can override.
478 */
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200479void blk_register_region(dev_t devt, unsigned long range, struct module *module,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480 struct kobject *(*probe)(dev_t, int *, void *),
481 int (*lock)(dev_t, void *), void *data)
482{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200483 kobj_map(bdev_map, devt, range, module, probe, lock, data);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484}
485
486EXPORT_SYMBOL(blk_register_region);
487
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200488void blk_unregister_region(dev_t devt, unsigned long range)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200490 kobj_unmap(bdev_map, devt, range);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491}
492
493EXPORT_SYMBOL(blk_unregister_region);
494
Tejun Heocf771cb2008-09-03 09:01:09 +0200495static struct kobject *exact_match(dev_t devt, int *partno, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496{
497 struct gendisk *p = data;
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200498
Tejun Heoed9e1982008-08-25 19:56:05 +0900499 return &disk_to_dev(p)->kobj;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700500}
501
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200502static int exact_lock(dev_t devt, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503{
504 struct gendisk *p = data;
505
506 if (!get_disk(p))
507 return -1;
508 return 0;
509}
510
Tejun Heod2bf1b62010-12-08 20:57:36 +0100511void register_disk(struct gendisk *disk)
512{
513 struct device *ddev = disk_to_dev(disk);
514 struct block_device *bdev;
515 struct disk_part_iter piter;
516 struct hd_struct *part;
517 int err;
518
519 ddev->parent = disk->driverfs_dev;
520
521 dev_set_name(ddev, disk->disk_name);
522
523 /* delay uevents, until we scanned partition table */
524 dev_set_uevent_suppress(ddev, 1);
525
526 if (device_add(ddev))
527 return;
528 if (!sysfs_deprecated) {
529 err = sysfs_create_link(block_depr, &ddev->kobj,
530 kobject_name(&ddev->kobj));
531 if (err) {
532 device_del(ddev);
533 return;
534 }
535 }
536 disk->part0.holder_dir = kobject_create_and_add("holders", &ddev->kobj);
537 disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
538
539 /* No minors to use for partitions */
540 if (!disk_partitionable(disk))
541 goto exit;
542
543 /* No such device (e.g., media were just removed) */
544 if (!get_capacity(disk))
545 goto exit;
546
547 bdev = bdget_disk(disk, 0);
548 if (!bdev)
549 goto exit;
550
551 bdev->bd_invalidated = 1;
552 err = blkdev_get(bdev, FMODE_READ, NULL);
553 if (err < 0)
554 goto exit;
555 blkdev_put(bdev, FMODE_READ);
556
557exit:
558 /* announce disk after possible partitions are created */
559 dev_set_uevent_suppress(ddev, 0);
560 kobject_uevent(&ddev->kobj, KOBJ_ADD);
561
562 /* announce possible partitions */
563 disk_part_iter_init(&piter, disk, 0);
564 while ((part = disk_part_iter_next(&piter)))
565 kobject_uevent(&part_to_dev(part)->kobj, KOBJ_ADD);
566 disk_part_iter_exit(&piter);
567}
568
Linus Torvalds1da177e2005-04-16 15:20:36 -0700569/**
570 * add_disk - add partitioning information to kernel list
571 * @disk: per-device partitioning information
572 *
573 * This function registers the partitioning information in @disk
574 * with the kernel.
Tejun Heo3e1a7ff2008-08-25 19:56:17 +0900575 *
576 * FIXME: error handling
Linus Torvalds1da177e2005-04-16 15:20:36 -0700577 */
578void add_disk(struct gendisk *disk)
579{
Peter Zijlstracf0ca9f2008-04-30 00:54:32 -0700580 struct backing_dev_info *bdi;
Tejun Heo3e1a7ff2008-08-25 19:56:17 +0900581 dev_t devt;
Greg Kroah-Hartman6ffeea72008-05-22 17:21:08 -0400582 int retval;
Peter Zijlstracf0ca9f2008-04-30 00:54:32 -0700583
Tejun Heo3e1a7ff2008-08-25 19:56:17 +0900584 /* minors == 0 indicates to use ext devt from part0 and should
585 * be accompanied with EXT_DEVT flag. Make sure all
586 * parameters make sense.
587 */
588 WARN_ON(disk->minors && !(disk->major || disk->first_minor));
589 WARN_ON(!disk->minors && !(disk->flags & GENHD_FL_EXT_DEVT));
590
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591 disk->flags |= GENHD_FL_UP;
Tejun Heo3e1a7ff2008-08-25 19:56:17 +0900592
593 retval = blk_alloc_devt(&disk->part0, &devt);
594 if (retval) {
595 WARN_ON(1);
596 return;
597 }
598 disk_to_dev(disk)->devt = devt;
599
600 /* ->major and ->first_minor aren't supposed to be
601 * dereferenced from here on, but set them just in case.
602 */
603 disk->major = MAJOR(devt);
604 disk->first_minor = MINOR(devt);
605
Wanlong Gao9f5e4862011-06-13 10:45:43 +0200606 /* Register BDI before referencing it from bdev */
Signed-off-by: Jan Kara01ea5062010-09-16 20:36:36 +0200607 bdi = &disk->queue->backing_dev_info;
608 bdi_register_dev(bdi, disk_devt(disk));
609
Tejun Heof331c022008-09-03 09:01:48 +0200610 blk_register_region(disk_devt(disk), disk->minors, NULL,
611 exact_match, exact_lock, disk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612 register_disk(disk);
613 blk_register_queue(disk);
Peter Zijlstracf0ca9f2008-04-30 00:54:32 -0700614
Tejun Heoed9e1982008-08-25 19:56:05 +0900615 retval = sysfs_create_link(&disk_to_dev(disk)->kobj, &bdi->dev->kobj,
616 "bdi");
Greg Kroah-Hartman6ffeea72008-05-22 17:21:08 -0400617 WARN_ON(retval);
Tejun Heo77ea8872010-12-08 20:57:37 +0100618
619 disk_add_events(disk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700620}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621EXPORT_SYMBOL(add_disk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700622
Tejun Heod2bf1b62010-12-08 20:57:36 +0100623void del_gendisk(struct gendisk *disk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624{
Tejun Heod2bf1b62010-12-08 20:57:36 +0100625 struct disk_part_iter piter;
626 struct hd_struct *part;
627
Tejun Heo77ea8872010-12-08 20:57:37 +0100628 disk_del_events(disk);
629
Tejun Heod2bf1b62010-12-08 20:57:36 +0100630 /* invalidate stuff */
631 disk_part_iter_init(&piter, disk,
632 DISK_PITER_INCL_EMPTY | DISK_PITER_REVERSE);
633 while ((part = disk_part_iter_next(&piter))) {
634 invalidate_partition(disk, part->partno);
635 delete_partition(disk, part->partno);
636 }
637 disk_part_iter_exit(&piter);
638
639 invalidate_partition(disk, 0);
640 blk_free_devt(disk_to_dev(disk)->devt);
641 set_capacity(disk, 0);
642 disk->flags &= ~GENHD_FL_UP;
643
Tejun Heoed9e1982008-08-25 19:56:05 +0900644 sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
Peter Zijlstracf0ca9f2008-04-30 00:54:32 -0700645 bdi_unregister(&disk->queue->backing_dev_info);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700646 blk_unregister_queue(disk);
Tejun Heof331c022008-09-03 09:01:48 +0200647 blk_unregister_region(disk_devt(disk), disk->minors);
Tejun Heod2bf1b62010-12-08 20:57:36 +0100648
649 part_stat_set_all(&disk->part0, 0);
650 disk->part0.stamp = 0;
651
652 kobject_put(disk->part0.holder_dir);
653 kobject_put(disk->slave_dir);
654 disk->driverfs_dev = NULL;
655 if (!sysfs_deprecated)
656 sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
657 device_del(disk_to_dev(disk));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658}
Tejun Heod2bf1b62010-12-08 20:57:36 +0100659EXPORT_SYMBOL(del_gendisk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661/**
662 * get_gendisk - get partitioning information for a given device
Randy Dunlap710027a2008-08-19 20:13:11 +0200663 * @devt: device to get partitioning information for
Randy Dunlap496aa8a2008-10-16 07:46:23 +0200664 * @partno: returned partition index
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665 *
666 * This function gets the structure containing partitioning
Randy Dunlap710027a2008-08-19 20:13:11 +0200667 * information for the given device @devt.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668 */
Tejun Heocf771cb2008-09-03 09:01:09 +0200669struct gendisk *get_gendisk(dev_t devt, int *partno)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670{
Tejun Heobcce3de2008-08-25 19:47:22 +0900671 struct gendisk *disk = NULL;
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200672
Tejun Heobcce3de2008-08-25 19:47:22 +0900673 if (MAJOR(devt) != BLOCK_EXT_MAJOR) {
674 struct kobject *kobj;
675
676 kobj = kobj_lookup(bdev_map, devt, partno);
677 if (kobj)
678 disk = dev_to_disk(kobj_to_dev(kobj));
679 } else {
680 struct hd_struct *part;
681
682 mutex_lock(&ext_devt_mutex);
Tejun Heo870d6652008-08-25 19:47:25 +0900683 part = idr_find(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
Tejun Heobcce3de2008-08-25 19:47:22 +0900684 if (part && get_disk(part_to_disk(part))) {
685 *partno = part->partno;
686 disk = part_to_disk(part);
687 }
688 mutex_unlock(&ext_devt_mutex);
689 }
690
691 return disk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692}
Divyesh Shahb6ac23af2010-04-15 08:54:59 +0200693EXPORT_SYMBOL(get_gendisk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700694
Tejun Heof331c022008-09-03 09:01:48 +0200695/**
696 * bdget_disk - do bdget() by gendisk and partition number
697 * @disk: gendisk of interest
698 * @partno: partition number
699 *
700 * Find partition @partno from @disk, do bdget() on it.
701 *
702 * CONTEXT:
703 * Don't care.
704 *
705 * RETURNS:
706 * Resulting block_device on success, NULL on failure.
707 */
Harvey Harrisonaeb3d3a2008-08-28 09:27:42 +0200708struct block_device *bdget_disk(struct gendisk *disk, int partno)
Tejun Heof331c022008-09-03 09:01:48 +0200709{
Tejun Heo548b10e2008-08-29 09:01:47 +0200710 struct hd_struct *part;
711 struct block_device *bdev = NULL;
Tejun Heof331c022008-09-03 09:01:48 +0200712
Tejun Heo548b10e2008-08-29 09:01:47 +0200713 part = disk_get_part(disk, partno);
Tejun Heo2bbedcb2008-08-29 11:41:51 +0200714 if (part)
Tejun Heo548b10e2008-08-29 09:01:47 +0200715 bdev = bdget(part_devt(part));
716 disk_put_part(part);
Tejun Heof331c022008-09-03 09:01:48 +0200717
Tejun Heo548b10e2008-08-29 09:01:47 +0200718 return bdev;
Tejun Heof331c022008-09-03 09:01:48 +0200719}
720EXPORT_SYMBOL(bdget_disk);
721
Dave Gilbertdd2a3452007-05-09 02:33:24 -0700722/*
723 * print a full list of all partitions - intended for places where the root
724 * filesystem can't be mounted and thus to give the victim some idea of what
725 * went wrong
726 */
727void __init printk_all_partitions(void)
728{
Tejun Heodef4e382008-09-03 08:57:12 +0200729 struct class_dev_iter iter;
730 struct device *dev;
731
732 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
733 while ((dev = class_dev_iter_next(&iter))) {
734 struct gendisk *disk = dev_to_disk(dev);
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200735 struct disk_part_iter piter;
736 struct hd_struct *part;
Tejun Heo1f014292008-08-25 19:47:23 +0900737 char name_buf[BDEVNAME_SIZE];
738 char devt_buf[BDEVT_SIZE];
Will Drewryb5af9212010-08-31 15:47:07 -0500739 u8 uuid[PARTITION_META_INFO_UUIDLTH * 2 + 1];
Tejun Heodef4e382008-09-03 08:57:12 +0200740
741 /*
742 * Don't show empty devices or things that have been
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300743 * suppressed
Tejun Heodef4e382008-09-03 08:57:12 +0200744 */
745 if (get_capacity(disk) == 0 ||
746 (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
747 continue;
748
749 /*
750 * Note, unlike /proc/partitions, I am showing the
751 * numbers in hex - the same format as the root=
752 * option takes.
753 */
Tejun Heo074a7ac2008-08-25 19:56:14 +0900754 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0);
755 while ((part = disk_part_iter_next(&piter))) {
756 bool is_part0 = part == &disk->part0;
Tejun Heodef4e382008-09-03 08:57:12 +0200757
Will Drewryb5af9212010-08-31 15:47:07 -0500758 uuid[0] = 0;
759 if (part->info)
760 part_unpack_uuid(part->info->uuid, uuid);
761
762 printk("%s%s %10llu %s %s", is_part0 ? "" : " ",
Tejun Heo1f014292008-08-25 19:47:23 +0900763 bdevt_str(part_devt(part), devt_buf),
Tejun Heof331c022008-09-03 09:01:48 +0200764 (unsigned long long)part->nr_sects >> 1,
Will Drewryb5af9212010-08-31 15:47:07 -0500765 disk_name(disk, part->partno, name_buf), uuid);
Tejun Heo074a7ac2008-08-25 19:56:14 +0900766 if (is_part0) {
767 if (disk->driverfs_dev != NULL &&
768 disk->driverfs_dev->driver != NULL)
769 printk(" driver: %s\n",
770 disk->driverfs_dev->driver->name);
771 else
772 printk(" (driver?)\n");
773 } else
774 printk("\n");
775 }
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200776 disk_part_iter_exit(&piter);
Tejun Heodef4e382008-09-03 08:57:12 +0200777 }
778 class_dev_iter_exit(&iter);
Dave Gilbertdd2a3452007-05-09 02:33:24 -0700779}
780
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781#ifdef CONFIG_PROC_FS
782/* iterator */
Tejun Heodef4e382008-09-03 08:57:12 +0200783static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
Greg Kroah-Hartman68c4d4a2008-05-22 17:21:08 -0400784{
Tejun Heodef4e382008-09-03 08:57:12 +0200785 loff_t skip = *pos;
786 struct class_dev_iter *iter;
787 struct device *dev;
Greg Kroah-Hartman68c4d4a2008-05-22 17:21:08 -0400788
Harvey Harrisonaeb3d3a2008-08-28 09:27:42 +0200789 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
Tejun Heodef4e382008-09-03 08:57:12 +0200790 if (!iter)
791 return ERR_PTR(-ENOMEM);
792
793 seqf->private = iter;
794 class_dev_iter_init(iter, &block_class, NULL, &disk_type);
795 do {
796 dev = class_dev_iter_next(iter);
797 if (!dev)
798 return NULL;
799 } while (skip--);
800
801 return dev_to_disk(dev);
Greg Kroah-Hartman68c4d4a2008-05-22 17:21:08 -0400802}
803
Tejun Heodef4e382008-09-03 08:57:12 +0200804static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200806 struct device *dev;
Greg Kroah-Hartman66c64af2008-05-22 17:21:08 -0400807
Tejun Heodef4e382008-09-03 08:57:12 +0200808 (*pos)++;
809 dev = class_dev_iter_next(seqf->private);
Tejun Heo2ac3cee2008-09-03 08:53:37 +0200810 if (dev)
Greg Kroah-Hartman68c4d4a2008-05-22 17:21:08 -0400811 return dev_to_disk(dev);
Tejun Heo2ac3cee2008-09-03 08:53:37 +0200812
Linus Torvalds1da177e2005-04-16 15:20:36 -0700813 return NULL;
814}
815
Tejun Heodef4e382008-09-03 08:57:12 +0200816static void disk_seqf_stop(struct seq_file *seqf, void *v)
Greg Kroah-Hartman27f30252008-05-22 17:21:08 -0400817{
Tejun Heodef4e382008-09-03 08:57:12 +0200818 struct class_dev_iter *iter = seqf->private;
Greg Kroah-Hartman27f30252008-05-22 17:21:08 -0400819
Tejun Heodef4e382008-09-03 08:57:12 +0200820 /* stop is called even after start failed :-( */
821 if (iter) {
822 class_dev_iter_exit(iter);
823 kfree(iter);
Kay Sievers5c0ef6d2008-08-16 14:30:30 +0200824 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825}
826
Tejun Heodef4e382008-09-03 08:57:12 +0200827static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828{
Tejun Heodef4e382008-09-03 08:57:12 +0200829 static void *p;
830
831 p = disk_seqf_start(seqf, pos);
Yang Zhangb9f985b2010-12-17 08:58:36 +0100832 if (!IS_ERR_OR_NULL(p) && !*pos)
Tejun Heodef4e382008-09-03 08:57:12 +0200833 seq_puts(seqf, "major minor #blocks name\n\n");
834 return p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835}
836
Tejun Heocf771cb2008-09-03 09:01:09 +0200837static int show_partition(struct seq_file *seqf, void *v)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838{
839 struct gendisk *sgp = v;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200840 struct disk_part_iter piter;
841 struct hd_struct *part;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700842 char buf[BDEVNAME_SIZE];
843
Linus Torvalds1da177e2005-04-16 15:20:36 -0700844 /* Don't show non-partitionable removeable devices or empty devices */
Tejun Heob5d0b9d2008-09-03 09:06:42 +0200845 if (!get_capacity(sgp) || (!disk_partitionable(sgp) &&
Tejun Heof331c022008-09-03 09:01:48 +0200846 (sgp->flags & GENHD_FL_REMOVABLE)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847 return 0;
848 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
849 return 0;
850
851 /* show the full disk and all non-0 size partitions of it */
Tejun Heo074a7ac2008-08-25 19:56:14 +0900852 disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0);
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200853 while ((part = disk_part_iter_next(&piter)))
Tejun Heo1f014292008-08-25 19:47:23 +0900854 seq_printf(seqf, "%4d %7d %10llu %s\n",
Tejun Heof331c022008-09-03 09:01:48 +0200855 MAJOR(part_devt(part)), MINOR(part_devt(part)),
856 (unsigned long long)part->nr_sects >> 1,
857 disk_name(sgp, part->partno, buf));
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200858 disk_part_iter_exit(&piter);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859
860 return 0;
861}
862
Alexey Dobriyanf5009752008-10-04 23:53:21 +0400863static const struct seq_operations partitions_op = {
Tejun Heodef4e382008-09-03 08:57:12 +0200864 .start = show_partition_start,
865 .next = disk_seqf_next,
866 .stop = disk_seqf_stop,
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200867 .show = show_partition
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868};
Alexey Dobriyanf5009752008-10-04 23:53:21 +0400869
870static int partitions_open(struct inode *inode, struct file *file)
871{
872 return seq_open(file, &partitions_op);
873}
874
875static const struct file_operations proc_partitions_operations = {
876 .open = partitions_open,
877 .read = seq_read,
878 .llseek = seq_lseek,
879 .release = seq_release,
880};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881#endif
882
883
Tejun Heocf771cb2008-09-03 09:01:09 +0200884static struct kobject *base_probe(dev_t devt, int *partno, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700885{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200886 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887 /* Make old-style 2.4 aliases work */
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200888 request_module("block-major-%d", MAJOR(devt));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889 return NULL;
890}
891
892static int __init genhd_device_init(void)
893{
Dan Williamse105b8b2008-04-21 10:51:07 -0700894 int error;
895
896 block_class.dev_kobj = sysfs_dev_block_kobj;
897 error = class_register(&block_class);
Roland McGrathee27a552008-03-11 17:13:15 -0700898 if (unlikely(error))
899 return error;
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200900 bdev_map = kobj_map_init(base_probe, &block_class_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700901 blk_dev_init();
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200902
Zhang, Yanmin561ec682008-11-14 08:26:30 +0100903 register_blkdev(BLOCK_EXT_MAJOR, "blkext");
904
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200905 /* create top-level block dir */
Andi Kleene52eec12010-09-08 16:54:17 +0200906 if (!sysfs_deprecated)
907 block_depr = kobject_create_and_add("block", NULL);
Greg Kroah-Hartman830d3cf2007-11-06 10:36:58 -0800908 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700909}
910
911subsys_initcall(genhd_device_init);
912
Nao Nishijimaa72c5e52011-08-25 18:04:06 +0900913static ssize_t alias_show(struct device *dev,
914 struct device_attribute *attr, char *buf)
915{
916 struct gendisk *disk = dev_to_disk(dev);
917 ssize_t ret = 0;
918
919 if (disk->alias)
920 ret = snprintf(buf, ALIAS_LEN, "%s\n", disk->alias);
921 return ret;
922}
923
924static ssize_t alias_store(struct device *dev, struct device_attribute *attr,
925 const char *buf, size_t count)
926{
927 struct gendisk *disk = dev_to_disk(dev);
928 char *alias;
929 char *envp[] = { NULL, NULL };
930 unsigned char c;
931 int i;
932 ssize_t ret = count;
933
934 if (!count)
935 return -EINVAL;
936
937 if (count >= ALIAS_LEN) {
938 printk(KERN_ERR "alias: alias is too long\n");
939 return -EINVAL;
940 }
941
942 /* Validation check */
943 for (i = 0; i < count; i++) {
944 c = buf[i];
945 if (i == count - 1 && c == '\n')
946 break;
947 if (!isalnum(c) && c != '_' && c != '-') {
948 printk(KERN_ERR "alias: invalid alias\n");
949 return -EINVAL;
950 }
951 }
952
953 if (disk->alias) {
954 printk(KERN_INFO "alias: %s is already assigned (%s)\n",
955 disk->disk_name, disk->alias);
956 return -EINVAL;
957 }
958
959 alias = kasprintf(GFP_KERNEL, "%s", buf);
960 if (!alias)
961 return -ENOMEM;
962
963 if (alias[count - 1] == '\n')
964 alias[count - 1] = '\0';
965
966 envp[0] = kasprintf(GFP_KERNEL, "ALIAS=%s", alias);
967 if (!envp[0]) {
968 kfree(alias);
969 return -ENOMEM;
970 }
971
972 disk->alias = alias;
973 printk(KERN_INFO "alias: assigned %s to %s\n", alias, disk->disk_name);
974
975 kobject_uevent_env(&dev->kobj, KOBJ_ADD, envp);
976
977 kfree(envp[0]);
978 return ret;
979}
980
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200981static ssize_t disk_range_show(struct device *dev,
982 struct device_attribute *attr, char *buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700983{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200984 struct gendisk *disk = dev_to_disk(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200986 return sprintf(buf, "%d\n", disk->minors);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700987}
988
Tejun Heo1f014292008-08-25 19:47:23 +0900989static ssize_t disk_ext_range_show(struct device *dev,
990 struct device_attribute *attr, char *buf)
991{
992 struct gendisk *disk = dev_to_disk(dev);
993
Tejun Heob5d0b9d2008-09-03 09:06:42 +0200994 return sprintf(buf, "%d\n", disk_max_parts(disk));
Tejun Heo1f014292008-08-25 19:47:23 +0900995}
996
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200997static ssize_t disk_removable_show(struct device *dev,
998 struct device_attribute *attr, char *buf)
Kay Sieversa7fd6702005-10-01 14:49:43 +0200999{
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001000 struct gendisk *disk = dev_to_disk(dev);
Kay Sieversa7fd6702005-10-01 14:49:43 +02001001
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001002 return sprintf(buf, "%d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001003 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001004}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005
Kay Sievers1c9ce522008-06-13 09:41:00 +02001006static ssize_t disk_ro_show(struct device *dev,
1007 struct device_attribute *attr, char *buf)
1008{
1009 struct gendisk *disk = dev_to_disk(dev);
1010
Tejun Heob7db9952008-08-25 19:56:10 +09001011 return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
Kay Sievers1c9ce522008-06-13 09:41:00 +02001012}
1013
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001014static ssize_t disk_capability_show(struct device *dev,
1015 struct device_attribute *attr, char *buf)
Kristen Carlson Accardi86ce18d2007-05-23 13:57:38 -07001016{
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001017 struct gendisk *disk = dev_to_disk(dev);
1018
1019 return sprintf(buf, "%x\n", disk->flags);
Kristen Carlson Accardi86ce18d2007-05-23 13:57:38 -07001020}
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001021
Martin K. Petersenc72758f2009-05-22 17:17:53 -04001022static ssize_t disk_alignment_offset_show(struct device *dev,
1023 struct device_attribute *attr,
1024 char *buf)
1025{
1026 struct gendisk *disk = dev_to_disk(dev);
1027
1028 return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue));
1029}
1030
Martin K. Petersen86b37282009-11-10 11:50:21 +01001031static ssize_t disk_discard_alignment_show(struct device *dev,
1032 struct device_attribute *attr,
1033 char *buf)
1034{
1035 struct gendisk *disk = dev_to_disk(dev);
1036
Martin K. Petersendd3d1452010-01-11 03:21:48 -05001037 return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue));
Martin K. Petersen86b37282009-11-10 11:50:21 +01001038}
1039
Nao Nishijimaa72c5e52011-08-25 18:04:06 +09001040static DEVICE_ATTR(alias, S_IRUGO|S_IWUSR, alias_show, alias_store);
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001041static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL);
Tejun Heo1f014292008-08-25 19:47:23 +09001042static DEVICE_ATTR(ext_range, S_IRUGO, disk_ext_range_show, NULL);
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001043static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL);
Kay Sievers1c9ce522008-06-13 09:41:00 +02001044static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL);
Tejun Heoe5610522008-08-25 19:56:09 +09001045static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL);
Martin K. Petersenc72758f2009-05-22 17:17:53 -04001046static DEVICE_ATTR(alignment_offset, S_IRUGO, disk_alignment_offset_show, NULL);
Martin K. Petersen86b37282009-11-10 11:50:21 +01001047static DEVICE_ATTR(discard_alignment, S_IRUGO, disk_discard_alignment_show,
1048 NULL);
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001049static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL);
Tejun Heo074a7ac2008-08-25 19:56:14 +09001050static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL);
Nikanth Karthikesan316d3152009-10-06 20:16:55 +02001051static DEVICE_ATTR(inflight, S_IRUGO, part_inflight_show, NULL);
Akinobu Mitac17bb492006-12-08 02:39:46 -08001052#ifdef CONFIG_FAIL_MAKE_REQUEST
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001053static struct device_attribute dev_attr_fail =
Tejun Heoeddb2e22008-08-25 19:56:13 +09001054 __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store);
Akinobu Mitac17bb492006-12-08 02:39:46 -08001055#endif
Jens Axboe581d4e22008-09-14 05:56:33 -07001056#ifdef CONFIG_FAIL_IO_TIMEOUT
1057static struct device_attribute dev_attr_fail_timeout =
1058 __ATTR(io-timeout-fail, S_IRUGO|S_IWUSR, part_timeout_show,
1059 part_timeout_store);
1060#endif
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001061
1062static struct attribute *disk_attrs[] = {
Nao Nishijimaa72c5e52011-08-25 18:04:06 +09001063 &dev_attr_alias.attr,
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001064 &dev_attr_range.attr,
Tejun Heo1f014292008-08-25 19:47:23 +09001065 &dev_attr_ext_range.attr,
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001066 &dev_attr_removable.attr,
Kay Sievers1c9ce522008-06-13 09:41:00 +02001067 &dev_attr_ro.attr,
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001068 &dev_attr_size.attr,
Martin K. Petersenc72758f2009-05-22 17:17:53 -04001069 &dev_attr_alignment_offset.attr,
Martin K. Petersen86b37282009-11-10 11:50:21 +01001070 &dev_attr_discard_alignment.attr,
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001071 &dev_attr_capability.attr,
1072 &dev_attr_stat.attr,
Nikanth Karthikesan316d3152009-10-06 20:16:55 +02001073 &dev_attr_inflight.attr,
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001074#ifdef CONFIG_FAIL_MAKE_REQUEST
1075 &dev_attr_fail.attr,
1076#endif
Jens Axboe581d4e22008-09-14 05:56:33 -07001077#ifdef CONFIG_FAIL_IO_TIMEOUT
1078 &dev_attr_fail_timeout.attr,
1079#endif
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001080 NULL
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081};
1082
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001083static struct attribute_group disk_attr_group = {
1084 .attrs = disk_attrs,
1085};
1086
David Brownella4dbd672009-06-24 10:06:31 -07001087static const struct attribute_group *disk_attr_groups[] = {
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001088 &disk_attr_group,
1089 NULL
1090};
1091
Tejun Heo540eed52008-08-25 19:56:15 +09001092/**
1093 * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way
1094 * @disk: disk to replace part_tbl for
1095 * @new_ptbl: new part_tbl to install
1096 *
1097 * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The
1098 * original ptbl is freed using RCU callback.
1099 *
1100 * LOCKING:
1101 * Matching bd_mutx locked.
1102 */
1103static void disk_replace_part_tbl(struct gendisk *disk,
1104 struct disk_part_tbl *new_ptbl)
1105{
1106 struct disk_part_tbl *old_ptbl = disk->part_tbl;
1107
1108 rcu_assign_pointer(disk->part_tbl, new_ptbl);
Jens Axboea6f23652008-10-24 12:52:42 +02001109
1110 if (old_ptbl) {
1111 rcu_assign_pointer(old_ptbl->last_lookup, NULL);
Lai Jiangshan57bdfbf2011-03-18 11:42:58 +08001112 kfree_rcu(old_ptbl, rcu_head);
Jens Axboea6f23652008-10-24 12:52:42 +02001113 }
Tejun Heo540eed52008-08-25 19:56:15 +09001114}
1115
1116/**
1117 * disk_expand_part_tbl - expand disk->part_tbl
1118 * @disk: disk to expand part_tbl for
1119 * @partno: expand such that this partno can fit in
1120 *
1121 * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl
1122 * uses RCU to allow unlocked dereferencing for stats and other stuff.
1123 *
1124 * LOCKING:
1125 * Matching bd_mutex locked, might sleep.
1126 *
1127 * RETURNS:
1128 * 0 on success, -errno on failure.
1129 */
1130int disk_expand_part_tbl(struct gendisk *disk, int partno)
1131{
1132 struct disk_part_tbl *old_ptbl = disk->part_tbl;
1133 struct disk_part_tbl *new_ptbl;
1134 int len = old_ptbl ? old_ptbl->len : 0;
1135 int target = partno + 1;
1136 size_t size;
1137 int i;
1138
1139 /* disk_max_parts() is zero during initialization, ignore if so */
1140 if (disk_max_parts(disk) && target > disk_max_parts(disk))
1141 return -EINVAL;
1142
1143 if (target <= len)
1144 return 0;
1145
1146 size = sizeof(*new_ptbl) + target * sizeof(new_ptbl->part[0]);
1147 new_ptbl = kzalloc_node(size, GFP_KERNEL, disk->node_id);
1148 if (!new_ptbl)
1149 return -ENOMEM;
1150
Tejun Heo540eed52008-08-25 19:56:15 +09001151 new_ptbl->len = target;
1152
1153 for (i = 0; i < len; i++)
1154 rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]);
1155
1156 disk_replace_part_tbl(disk, new_ptbl);
1157 return 0;
1158}
1159
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001160static void disk_release(struct device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161{
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001162 struct gendisk *disk = dev_to_disk(dev);
1163
Tejun Heo77ea8872010-12-08 20:57:37 +01001164 disk_release_events(disk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001165 kfree(disk->random);
Tejun Heo540eed52008-08-25 19:56:15 +09001166 disk_replace_part_tbl(disk, NULL);
Tejun Heo074a7ac2008-08-25 19:56:14 +09001167 free_part_stats(&disk->part0);
Will Drewry6d1d8052010-08-31 15:47:05 -05001168 free_part_info(&disk->part0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001169 kfree(disk);
1170}
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001171struct class block_class = {
1172 .name = "block",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001173};
1174
Kay Sieverse454cea2009-09-18 23:01:12 +02001175static char *block_devnode(struct device *dev, mode_t *mode)
Kay Sieversb03f38b2009-04-30 15:23:42 +02001176{
1177 struct gendisk *disk = dev_to_disk(dev);
1178
Kay Sieverse454cea2009-09-18 23:01:12 +02001179 if (disk->devnode)
1180 return disk->devnode(disk, mode);
Kay Sieversb03f38b2009-04-30 15:23:42 +02001181 return NULL;
1182}
1183
Adrian Bunk1826ead2008-03-04 11:23:46 +01001184static struct device_type disk_type = {
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001185 .name = "disk",
1186 .groups = disk_attr_groups,
1187 .release = disk_release,
Kay Sieverse454cea2009-09-18 23:01:12 +02001188 .devnode = block_devnode,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001189};
1190
Randy Dunlapa6e2ba82008-05-23 09:44:11 -07001191#ifdef CONFIG_PROC_FS
Tejun Heocf771cb2008-09-03 09:01:09 +02001192/*
1193 * aggregate disk stat collector. Uses the same stats that the sysfs
1194 * entries do, above, but makes them available through one seq_file.
1195 *
1196 * The output looks suspiciously like /proc/partitions with a bunch of
1197 * extra fields.
1198 */
1199static int diskstats_show(struct seq_file *seqf, void *v)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001200{
1201 struct gendisk *gp = v;
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001202 struct disk_part_iter piter;
1203 struct hd_struct *hd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001204 char buf[BDEVNAME_SIZE];
Tejun Heoc9959052008-08-25 19:47:21 +09001205 int cpu;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001206
1207 /*
Tejun Heoed9e1982008-08-25 19:56:05 +09001208 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
Tejun Heocf771cb2008-09-03 09:01:09 +02001209 seq_puts(seqf, "major minor name"
Linus Torvalds1da177e2005-04-16 15:20:36 -07001210 " rio rmerge rsect ruse wio wmerge "
1211 "wsect wuse running use aveq"
1212 "\n\n");
1213 */
Wanlong Gao9f5e4862011-06-13 10:45:43 +02001214
Tejun Heo71982a42009-04-17 08:34:48 +02001215 disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0);
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001216 while ((hd = disk_part_iter_next(&piter))) {
Tejun Heo074a7ac2008-08-25 19:56:14 +09001217 cpu = part_stat_lock();
Tejun Heoc9959052008-08-25 19:47:21 +09001218 part_round_stats(cpu, hd);
Tejun Heo074a7ac2008-08-25 19:56:14 +09001219 part_stat_unlock();
Herbert Poetzlf95fe9c2011-08-02 12:43:50 +02001220 seq_printf(seqf, "%4d %7d %s %lu %lu %lu "
1221 "%u %lu %lu %lu %u %u %u %u\n",
Tejun Heof331c022008-09-03 09:01:48 +02001222 MAJOR(part_devt(hd)), MINOR(part_devt(hd)),
1223 disk_name(gp, hd->partno, buf),
Liu Yuan53f22952011-03-02 11:00:15 -05001224 part_stat_read(hd, ios[READ]),
1225 part_stat_read(hd, merges[READ]),
Herbert Poetzlf95fe9c2011-08-02 12:43:50 +02001226 part_stat_read(hd, sectors[READ]),
Liu Yuan53f22952011-03-02 11:00:15 -05001227 jiffies_to_msecs(part_stat_read(hd, ticks[READ])),
1228 part_stat_read(hd, ios[WRITE]),
1229 part_stat_read(hd, merges[WRITE]),
Herbert Poetzlf95fe9c2011-08-02 12:43:50 +02001230 part_stat_read(hd, sectors[WRITE]),
Liu Yuan53f22952011-03-02 11:00:15 -05001231 jiffies_to_msecs(part_stat_read(hd, ticks[WRITE])),
Nikanth Karthikesan316d3152009-10-06 20:16:55 +02001232 part_in_flight(hd),
Jerome Marchand28f39d52008-02-08 11:04:56 +01001233 jiffies_to_msecs(part_stat_read(hd, io_ticks)),
1234 jiffies_to_msecs(part_stat_read(hd, time_in_queue))
1235 );
Linus Torvalds1da177e2005-04-16 15:20:36 -07001236 }
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001237 disk_part_iter_exit(&piter);
Wanlong Gao9f5e4862011-06-13 10:45:43 +02001238
Linus Torvalds1da177e2005-04-16 15:20:36 -07001239 return 0;
1240}
1241
Alexey Dobriyan31d85ab2008-10-06 12:55:38 +04001242static const struct seq_operations diskstats_op = {
Tejun Heodef4e382008-09-03 08:57:12 +02001243 .start = disk_seqf_start,
1244 .next = disk_seqf_next,
1245 .stop = disk_seqf_stop,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001246 .show = diskstats_show
1247};
Alexey Dobriyanf5009752008-10-04 23:53:21 +04001248
Alexey Dobriyan31d85ab2008-10-06 12:55:38 +04001249static int diskstats_open(struct inode *inode, struct file *file)
1250{
1251 return seq_open(file, &diskstats_op);
1252}
1253
1254static const struct file_operations proc_diskstats_operations = {
1255 .open = diskstats_open,
1256 .read = seq_read,
1257 .llseek = seq_lseek,
1258 .release = seq_release,
1259};
1260
Alexey Dobriyanf5009752008-10-04 23:53:21 +04001261static int __init proc_genhd_init(void)
1262{
Alexey Dobriyan31d85ab2008-10-06 12:55:38 +04001263 proc_create("diskstats", 0, NULL, &proc_diskstats_operations);
Alexey Dobriyanf5009752008-10-04 23:53:21 +04001264 proc_create("partitions", 0, NULL, &proc_partitions_operations);
1265 return 0;
1266}
1267module_init(proc_genhd_init);
Randy Dunlapa6e2ba82008-05-23 09:44:11 -07001268#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001269
Tejun Heocf771cb2008-09-03 09:01:09 +02001270dev_t blk_lookup_devt(const char *name, int partno)
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001271{
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001272 dev_t devt = MKDEV(0, 0);
Tejun Heodef4e382008-09-03 08:57:12 +02001273 struct class_dev_iter iter;
1274 struct device *dev;
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001275
Tejun Heodef4e382008-09-03 08:57:12 +02001276 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1277 while ((dev = class_dev_iter_next(&iter))) {
1278 struct gendisk *disk = dev_to_disk(dev);
Tejun Heo548b10e2008-08-29 09:01:47 +02001279 struct hd_struct *part;
Tejun Heodef4e382008-09-03 08:57:12 +02001280
Kay Sievers3ada8b72009-01-06 10:44:43 -08001281 if (strcmp(dev_name(dev), name))
Tejun Heof331c022008-09-03 09:01:48 +02001282 continue;
Tejun Heof331c022008-09-03 09:01:48 +02001283
Neil Brown41b8c852009-02-18 10:33:59 +01001284 if (partno < disk->minors) {
1285 /* We need to return the right devno, even
1286 * if the partition doesn't exist yet.
1287 */
1288 devt = MKDEV(MAJOR(dev->devt),
1289 MINOR(dev->devt) + partno);
1290 break;
1291 }
Tejun Heo548b10e2008-08-29 09:01:47 +02001292 part = disk_get_part(disk, partno);
Tejun Heo2bbedcb2008-08-29 11:41:51 +02001293 if (part) {
Tejun Heof331c022008-09-03 09:01:48 +02001294 devt = part_devt(part);
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001295 disk_put_part(part);
Tejun Heo548b10e2008-08-29 09:01:47 +02001296 break;
Tejun Heodef4e382008-09-03 08:57:12 +02001297 }
Tejun Heo548b10e2008-08-29 09:01:47 +02001298 disk_put_part(part);
Kay Sievers5c0ef6d2008-08-16 14:30:30 +02001299 }
Tejun Heodef4e382008-09-03 08:57:12 +02001300 class_dev_iter_exit(&iter);
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001301 return devt;
1302}
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001303EXPORT_SYMBOL(blk_lookup_devt);
1304
Linus Torvalds1da177e2005-04-16 15:20:36 -07001305struct gendisk *alloc_disk(int minors)
1306{
Christoph Lameter19460892005-06-23 00:08:19 -07001307 return alloc_disk_node(minors, -1);
1308}
Tejun Heo689d6fa2008-08-25 19:56:16 +09001309EXPORT_SYMBOL(alloc_disk);
Christoph Lameter19460892005-06-23 00:08:19 -07001310
1311struct gendisk *alloc_disk_node(int minors, int node_id)
1312{
1313 struct gendisk *disk;
1314
Christoph Lameter94f60302007-07-17 04:03:29 -07001315 disk = kmalloc_node(sizeof(struct gendisk),
1316 GFP_KERNEL | __GFP_ZERO, node_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001317 if (disk) {
Tejun Heo074a7ac2008-08-25 19:56:14 +09001318 if (!init_part_stats(&disk->part0)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001319 kfree(disk);
1320 return NULL;
1321 }
Cheng Renquanbf91db12008-11-20 08:37:37 +01001322 disk->node_id = node_id;
Tejun Heo540eed52008-08-25 19:56:15 +09001323 if (disk_expand_part_tbl(disk, 0)) {
1324 free_part_stats(&disk->part0);
Tejun Heob5d0b9d2008-09-03 09:06:42 +02001325 kfree(disk);
1326 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001327 }
Tejun Heo540eed52008-08-25 19:56:15 +09001328 disk->part_tbl->part[0] = &disk->part0;
Jens Axboe6c23a962011-01-07 08:43:37 +01001329
1330 hd_ref_init(&disk->part0);
Tejun Heob5d0b9d2008-09-03 09:06:42 +02001331
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332 disk->minors = minors;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001333 rand_initialize_disk(disk);
Tejun Heoed9e1982008-08-25 19:56:05 +09001334 disk_to_dev(disk)->class = &block_class;
1335 disk_to_dev(disk)->type = &disk_type;
1336 device_initialize(disk_to_dev(disk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001337 }
1338 return disk;
1339}
Christoph Lameter19460892005-06-23 00:08:19 -07001340EXPORT_SYMBOL(alloc_disk_node);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001341
1342struct kobject *get_disk(struct gendisk *disk)
1343{
1344 struct module *owner;
1345 struct kobject *kobj;
1346
1347 if (!disk->fops)
1348 return NULL;
1349 owner = disk->fops->owner;
1350 if (owner && !try_module_get(owner))
1351 return NULL;
Tejun Heoed9e1982008-08-25 19:56:05 +09001352 kobj = kobject_get(&disk_to_dev(disk)->kobj);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001353 if (kobj == NULL) {
1354 module_put(owner);
1355 return NULL;
1356 }
1357 return kobj;
1358
1359}
1360
1361EXPORT_SYMBOL(get_disk);
1362
1363void put_disk(struct gendisk *disk)
1364{
1365 if (disk)
Tejun Heoed9e1982008-08-25 19:56:05 +09001366 kobject_put(&disk_to_dev(disk)->kobj);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001367}
1368
1369EXPORT_SYMBOL(put_disk);
1370
Hannes Reineckee3264a42009-07-28 09:13:13 +02001371static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1372{
1373 char event[] = "DISK_RO=1";
1374 char *envp[] = { event, NULL };
1375
1376 if (!ro)
1377 event[8] = '0';
1378 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1379}
1380
Linus Torvalds1da177e2005-04-16 15:20:36 -07001381void set_device_ro(struct block_device *bdev, int flag)
1382{
Tejun Heob7db9952008-08-25 19:56:10 +09001383 bdev->bd_part->policy = flag;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001384}
1385
1386EXPORT_SYMBOL(set_device_ro);
1387
1388void set_disk_ro(struct gendisk *disk, int flag)
1389{
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001390 struct disk_part_iter piter;
1391 struct hd_struct *part;
1392
Hannes Reineckee3264a42009-07-28 09:13:13 +02001393 if (disk->part0.policy != flag) {
1394 set_disk_ro_uevent(disk, flag);
1395 disk->part0.policy = flag;
1396 }
1397
1398 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY);
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001399 while ((part = disk_part_iter_next(&piter)))
1400 part->policy = flag;
1401 disk_part_iter_exit(&piter);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001402}
1403
1404EXPORT_SYMBOL(set_disk_ro);
1405
1406int bdev_read_only(struct block_device *bdev)
1407{
1408 if (!bdev)
1409 return 0;
Tejun Heob7db9952008-08-25 19:56:10 +09001410 return bdev->bd_part->policy;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001411}
1412
1413EXPORT_SYMBOL(bdev_read_only);
1414
Tejun Heocf771cb2008-09-03 09:01:09 +02001415int invalidate_partition(struct gendisk *disk, int partno)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001416{
1417 int res = 0;
Tejun Heocf771cb2008-09-03 09:01:09 +02001418 struct block_device *bdev = bdget_disk(disk, partno);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001419 if (bdev) {
Christoph Hellwig2ef41632005-05-05 16:15:59 -07001420 fsync_bdev(bdev);
NeilBrown93b270f2011-02-24 17:25:47 +11001421 res = __invalidate_device(bdev, true);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001422 bdput(bdev);
1423 }
1424 return res;
1425}
1426
1427EXPORT_SYMBOL(invalidate_partition);
Tejun Heo77ea8872010-12-08 20:57:37 +01001428
1429/*
1430 * Disk events - monitor disk events like media change and eject request.
1431 */
1432struct disk_events {
1433 struct list_head node; /* all disk_event's */
1434 struct gendisk *disk; /* the associated disk */
1435 spinlock_t lock;
1436
Tejun Heofdd514e2011-06-09 20:43:59 +02001437 struct mutex block_mutex; /* protects blocking */
Tejun Heo77ea8872010-12-08 20:57:37 +01001438 int block; /* event blocking depth */
1439 unsigned int pending; /* events already sent out */
1440 unsigned int clearing; /* events being cleared */
1441
1442 long poll_msecs; /* interval, -1 for default */
1443 struct delayed_work dwork;
1444};
1445
1446static const char *disk_events_strs[] = {
1447 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "media_change",
1448 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "eject_request",
1449};
1450
1451static char *disk_uevents[] = {
1452 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "DISK_MEDIA_CHANGE=1",
1453 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "DISK_EJECT_REQUEST=1",
1454};
1455
1456/* list of all disk_events */
1457static DEFINE_MUTEX(disk_events_mutex);
1458static LIST_HEAD(disk_events);
1459
1460/* disable in-kernel polling by default */
1461static unsigned long disk_events_dfl_poll_msecs = 0;
1462
1463static unsigned long disk_events_poll_jiffies(struct gendisk *disk)
1464{
1465 struct disk_events *ev = disk->ev;
1466 long intv_msecs = 0;
1467
1468 /*
1469 * If device-specific poll interval is set, always use it. If
1470 * the default is being used, poll iff there are events which
1471 * can't be monitored asynchronously.
1472 */
1473 if (ev->poll_msecs >= 0)
1474 intv_msecs = ev->poll_msecs;
1475 else if (disk->events & ~disk->async_events)
1476 intv_msecs = disk_events_dfl_poll_msecs;
1477
1478 return msecs_to_jiffies(intv_msecs);
1479}
1480
Tejun Heoc3af54a2011-06-09 20:43:55 +02001481/**
1482 * disk_block_events - block and flush disk event checking
1483 * @disk: disk to block events for
1484 *
1485 * On return from this function, it is guaranteed that event checking
1486 * isn't in progress and won't happen until unblocked by
1487 * disk_unblock_events(). Events blocking is counted and the actual
1488 * unblocking happens after the matching number of unblocks are done.
1489 *
1490 * Note that this intentionally does not block event checking from
1491 * disk_clear_events().
1492 *
1493 * CONTEXT:
1494 * Might sleep.
1495 */
1496void disk_block_events(struct gendisk *disk)
Tejun Heo77ea8872010-12-08 20:57:37 +01001497{
1498 struct disk_events *ev = disk->ev;
1499 unsigned long flags;
1500 bool cancel;
1501
Tejun Heoc3af54a2011-06-09 20:43:55 +02001502 if (!ev)
1503 return;
1504
Tejun Heofdd514e2011-06-09 20:43:59 +02001505 /*
1506 * Outer mutex ensures that the first blocker completes canceling
1507 * the event work before further blockers are allowed to finish.
1508 */
1509 mutex_lock(&ev->block_mutex);
1510
Tejun Heo77ea8872010-12-08 20:57:37 +01001511 spin_lock_irqsave(&ev->lock, flags);
1512 cancel = !ev->block++;
1513 spin_unlock_irqrestore(&ev->lock, flags);
1514
Tejun Heoc3af54a2011-06-09 20:43:55 +02001515 if (cancel)
1516 cancel_delayed_work_sync(&disk->ev->dwork);
Tejun Heofdd514e2011-06-09 20:43:59 +02001517
1518 mutex_unlock(&ev->block_mutex);
Tejun Heo77ea8872010-12-08 20:57:37 +01001519}
1520
1521static void __disk_unblock_events(struct gendisk *disk, bool check_now)
1522{
1523 struct disk_events *ev = disk->ev;
1524 unsigned long intv;
1525 unsigned long flags;
1526
1527 spin_lock_irqsave(&ev->lock, flags);
1528
1529 if (WARN_ON_ONCE(ev->block <= 0))
1530 goto out_unlock;
1531
1532 if (--ev->block)
1533 goto out_unlock;
1534
1535 /*
1536 * Not exactly a latency critical operation, set poll timer
1537 * slack to 25% and kick event check.
1538 */
1539 intv = disk_events_poll_jiffies(disk);
1540 set_timer_slack(&ev->dwork.timer, intv / 4);
1541 if (check_now)
1542 queue_delayed_work(system_nrt_wq, &ev->dwork, 0);
1543 else if (intv)
1544 queue_delayed_work(system_nrt_wq, &ev->dwork, intv);
1545out_unlock:
1546 spin_unlock_irqrestore(&ev->lock, flags);
1547}
1548
1549/**
Tejun Heo77ea8872010-12-08 20:57:37 +01001550 * disk_unblock_events - unblock disk event checking
1551 * @disk: disk to unblock events for
1552 *
1553 * Undo disk_block_events(). When the block count reaches zero, it
1554 * starts events polling if configured.
1555 *
1556 * CONTEXT:
1557 * Don't care. Safe to call from irq context.
1558 */
1559void disk_unblock_events(struct gendisk *disk)
1560{
1561 if (disk->ev)
Tejun Heofacc31d2011-03-09 19:54:27 +01001562 __disk_unblock_events(disk, false);
Tejun Heo77ea8872010-12-08 20:57:37 +01001563}
1564
1565/**
Tejun Heo85ef06d2011-07-01 16:17:47 +02001566 * disk_flush_events - schedule immediate event checking and flushing
1567 * @disk: disk to check and flush events for
1568 * @mask: events to flush
Tejun Heo77ea8872010-12-08 20:57:37 +01001569 *
Tejun Heo85ef06d2011-07-01 16:17:47 +02001570 * Schedule immediate event checking on @disk if not blocked. Events in
1571 * @mask are scheduled to be cleared from the driver. Note that this
1572 * doesn't clear the events from @disk->ev.
Tejun Heo77ea8872010-12-08 20:57:37 +01001573 *
1574 * CONTEXT:
Tejun Heo85ef06d2011-07-01 16:17:47 +02001575 * If @mask is non-zero must be called with bdev->bd_mutex held.
Tejun Heo77ea8872010-12-08 20:57:37 +01001576 */
Tejun Heo85ef06d2011-07-01 16:17:47 +02001577void disk_flush_events(struct gendisk *disk, unsigned int mask)
Tejun Heo77ea8872010-12-08 20:57:37 +01001578{
Tejun Heoa9dce2a2011-06-09 20:43:54 +02001579 struct disk_events *ev = disk->ev;
Tejun Heoa9dce2a2011-06-09 20:43:54 +02001580
1581 if (!ev)
1582 return;
1583
Tejun Heo85ef06d2011-07-01 16:17:47 +02001584 spin_lock_irq(&ev->lock);
1585 ev->clearing |= mask;
Tejun Heoa9dce2a2011-06-09 20:43:54 +02001586 if (!ev->block) {
1587 cancel_delayed_work(&ev->dwork);
1588 queue_delayed_work(system_nrt_wq, &ev->dwork, 0);
Tejun Heo77ea8872010-12-08 20:57:37 +01001589 }
Tejun Heo85ef06d2011-07-01 16:17:47 +02001590 spin_unlock_irq(&ev->lock);
Tejun Heo77ea8872010-12-08 20:57:37 +01001591}
Tejun Heo77ea8872010-12-08 20:57:37 +01001592
1593/**
1594 * disk_clear_events - synchronously check, clear and return pending events
1595 * @disk: disk to fetch and clear events from
1596 * @mask: mask of events to be fetched and clearted
1597 *
1598 * Disk events are synchronously checked and pending events in @mask
1599 * are cleared and returned. This ignores the block count.
1600 *
1601 * CONTEXT:
1602 * Might sleep.
1603 */
1604unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask)
1605{
1606 const struct block_device_operations *bdops = disk->fops;
1607 struct disk_events *ev = disk->ev;
1608 unsigned int pending;
1609
1610 if (!ev) {
1611 /* for drivers still using the old ->media_changed method */
1612 if ((mask & DISK_EVENT_MEDIA_CHANGE) &&
1613 bdops->media_changed && bdops->media_changed(disk))
1614 return DISK_EVENT_MEDIA_CHANGE;
1615 return 0;
1616 }
1617
1618 /* tell the workfn about the events being cleared */
1619 spin_lock_irq(&ev->lock);
1620 ev->clearing |= mask;
1621 spin_unlock_irq(&ev->lock);
1622
1623 /* uncondtionally schedule event check and wait for it to finish */
Tejun Heoc3af54a2011-06-09 20:43:55 +02001624 disk_block_events(disk);
Tejun Heo77ea8872010-12-08 20:57:37 +01001625 queue_delayed_work(system_nrt_wq, &ev->dwork, 0);
1626 flush_delayed_work(&ev->dwork);
1627 __disk_unblock_events(disk, false);
1628
1629 /* then, fetch and clear pending events */
1630 spin_lock_irq(&ev->lock);
1631 WARN_ON_ONCE(ev->clearing & mask); /* cleared by workfn */
1632 pending = ev->pending & mask;
1633 ev->pending &= ~mask;
1634 spin_unlock_irq(&ev->lock);
1635
1636 return pending;
1637}
1638
1639static void disk_events_workfn(struct work_struct *work)
1640{
1641 struct delayed_work *dwork = to_delayed_work(work);
1642 struct disk_events *ev = container_of(dwork, struct disk_events, dwork);
1643 struct gendisk *disk = ev->disk;
1644 char *envp[ARRAY_SIZE(disk_uevents) + 1] = { };
1645 unsigned int clearing = ev->clearing;
1646 unsigned int events;
1647 unsigned long intv;
1648 int nr_events = 0, i;
1649
1650 /* check events */
1651 events = disk->fops->check_events(disk, clearing);
1652
1653 /* accumulate pending events and schedule next poll if necessary */
1654 spin_lock_irq(&ev->lock);
1655
1656 events &= ~ev->pending;
1657 ev->pending |= events;
1658 ev->clearing &= ~clearing;
1659
1660 intv = disk_events_poll_jiffies(disk);
1661 if (!ev->block && intv)
1662 queue_delayed_work(system_nrt_wq, &ev->dwork, intv);
1663
1664 spin_unlock_irq(&ev->lock);
1665
Tejun Heo7c88a162011-04-21 19:43:58 +02001666 /*
1667 * Tell userland about new events. Only the events listed in
1668 * @disk->events are reported. Unlisted events are processed the
1669 * same internally but never get reported to userland.
1670 */
Tejun Heo77ea8872010-12-08 20:57:37 +01001671 for (i = 0; i < ARRAY_SIZE(disk_uevents); i++)
Tejun Heo7c88a162011-04-21 19:43:58 +02001672 if (events & disk->events & (1 << i))
Tejun Heo77ea8872010-12-08 20:57:37 +01001673 envp[nr_events++] = disk_uevents[i];
1674
1675 if (nr_events)
1676 kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp);
1677}
1678
1679/*
1680 * A disk events enabled device has the following sysfs nodes under
1681 * its /sys/block/X/ directory.
1682 *
1683 * events : list of all supported events
1684 * events_async : list of events which can be detected w/o polling
1685 * events_poll_msecs : polling interval, 0: disable, -1: system default
1686 */
1687static ssize_t __disk_events_show(unsigned int events, char *buf)
1688{
1689 const char *delim = "";
1690 ssize_t pos = 0;
1691 int i;
1692
1693 for (i = 0; i < ARRAY_SIZE(disk_events_strs); i++)
1694 if (events & (1 << i)) {
1695 pos += sprintf(buf + pos, "%s%s",
1696 delim, disk_events_strs[i]);
1697 delim = " ";
1698 }
1699 if (pos)
1700 pos += sprintf(buf + pos, "\n");
1701 return pos;
1702}
1703
1704static ssize_t disk_events_show(struct device *dev,
1705 struct device_attribute *attr, char *buf)
1706{
1707 struct gendisk *disk = dev_to_disk(dev);
1708
1709 return __disk_events_show(disk->events, buf);
1710}
1711
1712static ssize_t disk_events_async_show(struct device *dev,
1713 struct device_attribute *attr, char *buf)
1714{
1715 struct gendisk *disk = dev_to_disk(dev);
1716
1717 return __disk_events_show(disk->async_events, buf);
1718}
1719
1720static ssize_t disk_events_poll_msecs_show(struct device *dev,
1721 struct device_attribute *attr,
1722 char *buf)
1723{
1724 struct gendisk *disk = dev_to_disk(dev);
1725
1726 return sprintf(buf, "%ld\n", disk->ev->poll_msecs);
1727}
1728
1729static ssize_t disk_events_poll_msecs_store(struct device *dev,
1730 struct device_attribute *attr,
1731 const char *buf, size_t count)
1732{
1733 struct gendisk *disk = dev_to_disk(dev);
1734 long intv;
1735
1736 if (!count || !sscanf(buf, "%ld", &intv))
1737 return -EINVAL;
1738
1739 if (intv < 0 && intv != -1)
1740 return -EINVAL;
1741
Tejun Heoc3af54a2011-06-09 20:43:55 +02001742 disk_block_events(disk);
Tejun Heo77ea8872010-12-08 20:57:37 +01001743 disk->ev->poll_msecs = intv;
1744 __disk_unblock_events(disk, true);
1745
1746 return count;
1747}
1748
1749static const DEVICE_ATTR(events, S_IRUGO, disk_events_show, NULL);
1750static const DEVICE_ATTR(events_async, S_IRUGO, disk_events_async_show, NULL);
1751static const DEVICE_ATTR(events_poll_msecs, S_IRUGO|S_IWUSR,
1752 disk_events_poll_msecs_show,
1753 disk_events_poll_msecs_store);
1754
1755static const struct attribute *disk_events_attrs[] = {
1756 &dev_attr_events.attr,
1757 &dev_attr_events_async.attr,
1758 &dev_attr_events_poll_msecs.attr,
1759 NULL,
1760};
1761
1762/*
1763 * The default polling interval can be specified by the kernel
1764 * parameter block.events_dfl_poll_msecs which defaults to 0
1765 * (disable). This can also be modified runtime by writing to
1766 * /sys/module/block/events_dfl_poll_msecs.
1767 */
1768static int disk_events_set_dfl_poll_msecs(const char *val,
1769 const struct kernel_param *kp)
1770{
1771 struct disk_events *ev;
1772 int ret;
1773
1774 ret = param_set_ulong(val, kp);
1775 if (ret < 0)
1776 return ret;
1777
1778 mutex_lock(&disk_events_mutex);
1779
1780 list_for_each_entry(ev, &disk_events, node)
Tejun Heo85ef06d2011-07-01 16:17:47 +02001781 disk_flush_events(ev->disk, 0);
Tejun Heo77ea8872010-12-08 20:57:37 +01001782
1783 mutex_unlock(&disk_events_mutex);
1784
1785 return 0;
1786}
1787
1788static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops = {
1789 .set = disk_events_set_dfl_poll_msecs,
1790 .get = param_get_ulong,
1791};
1792
1793#undef MODULE_PARAM_PREFIX
1794#define MODULE_PARAM_PREFIX "block."
1795
1796module_param_cb(events_dfl_poll_msecs, &disk_events_dfl_poll_msecs_param_ops,
1797 &disk_events_dfl_poll_msecs, 0644);
1798
1799/*
1800 * disk_{add|del|release}_events - initialize and destroy disk_events.
1801 */
1802static void disk_add_events(struct gendisk *disk)
1803{
1804 struct disk_events *ev;
1805
Tejun Heo75e3f3e2011-05-26 21:06:50 +02001806 if (!disk->fops->check_events)
Tejun Heo77ea8872010-12-08 20:57:37 +01001807 return;
1808
1809 ev = kzalloc(sizeof(*ev), GFP_KERNEL);
1810 if (!ev) {
1811 pr_warn("%s: failed to initialize events\n", disk->disk_name);
1812 return;
1813 }
1814
1815 if (sysfs_create_files(&disk_to_dev(disk)->kobj,
1816 disk_events_attrs) < 0) {
1817 pr_warn("%s: failed to create sysfs files for events\n",
1818 disk->disk_name);
1819 kfree(ev);
1820 return;
1821 }
1822
1823 disk->ev = ev;
1824
1825 INIT_LIST_HEAD(&ev->node);
1826 ev->disk = disk;
1827 spin_lock_init(&ev->lock);
Tejun Heofdd514e2011-06-09 20:43:59 +02001828 mutex_init(&ev->block_mutex);
Tejun Heo77ea8872010-12-08 20:57:37 +01001829 ev->block = 1;
1830 ev->poll_msecs = -1;
1831 INIT_DELAYED_WORK(&ev->dwork, disk_events_workfn);
1832
1833 mutex_lock(&disk_events_mutex);
1834 list_add_tail(&ev->node, &disk_events);
1835 mutex_unlock(&disk_events_mutex);
1836
1837 /*
1838 * Block count is initialized to 1 and the following initial
1839 * unblock kicks it into action.
1840 */
1841 __disk_unblock_events(disk, true);
1842}
1843
1844static void disk_del_events(struct gendisk *disk)
1845{
1846 if (!disk->ev)
1847 return;
1848
Tejun Heoc3af54a2011-06-09 20:43:55 +02001849 disk_block_events(disk);
Tejun Heo77ea8872010-12-08 20:57:37 +01001850
1851 mutex_lock(&disk_events_mutex);
1852 list_del_init(&disk->ev->node);
1853 mutex_unlock(&disk_events_mutex);
1854
1855 sysfs_remove_files(&disk_to_dev(disk)->kobj, disk_events_attrs);
1856}
1857
1858static void disk_release_events(struct gendisk *disk)
1859{
1860 /* the block count should be 1 from disk_del_events() */
1861 WARN_ON_ONCE(disk->ev && disk->ev->block != 1);
1862 kfree(disk->ev);
1863}