blob: b26c4085590da7917a666e068ac469cfbb8fe69d [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * gendisk handling
3 */
4
Linus Torvalds1da177e2005-04-16 15:20:36 -07005#include <linux/module.h>
6#include <linux/fs.h>
7#include <linux/genhd.h>
Andrew Mortonb446b602007-02-20 13:57:48 -08008#include <linux/kdev_t.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -07009#include <linux/kernel.h>
10#include <linux/blkdev.h>
11#include <linux/init.h>
12#include <linux/spinlock.h>
Alexey Dobriyanf5009752008-10-04 23:53:21 +040013#include <linux/proc_fs.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070014#include <linux/seq_file.h>
15#include <linux/slab.h>
16#include <linux/kmod.h>
17#include <linux/kobj_map.h>
Jes Sorensen58383af2006-02-06 14:12:43 -080018#include <linux/mutex.h>
Tejun Heobcce3de2008-08-25 19:47:22 +090019#include <linux/idr.h>
Tejun Heo77ea8872010-12-08 20:57:37 +010020#include <linux/log2.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070021
Adrian Bunkff889722008-03-04 11:23:45 +010022#include "blk.h"
23
Kay Sieversedfaa7c2007-05-21 22:08:01 +020024static DEFINE_MUTEX(block_class_lock);
Kay Sieversedfaa7c2007-05-21 22:08:01 +020025struct kobject *block_depr;
Linus Torvalds1da177e2005-04-16 15:20:36 -070026
Tejun Heobcce3de2008-08-25 19:47:22 +090027/* for extended dynamic devt allocation, currently only one major is used */
28#define MAX_EXT_DEVT (1 << MINORBITS)
29
30/* For extended devt allocation. ext_devt_mutex prevents look up
31 * results from going away underneath its user.
32 */
33static DEFINE_MUTEX(ext_devt_mutex);
34static DEFINE_IDR(ext_devt_idr);
35
Adrian Bunk1826ead2008-03-04 11:23:46 +010036static struct device_type disk_type;
37
Stanislaw Gruszka9f53d2fe2012-03-02 10:43:28 +010038static void disk_alloc_events(struct gendisk *disk);
Tejun Heo77ea8872010-12-08 20:57:37 +010039static void disk_add_events(struct gendisk *disk);
40static void disk_del_events(struct gendisk *disk);
41static void disk_release_events(struct gendisk *disk);
42
Tejun Heoe71bf0d2008-09-03 09:03:02 +020043/**
44 * disk_get_part - get partition
45 * @disk: disk to look partition from
46 * @partno: partition number
47 *
48 * Look for partition @partno from @disk. If found, increment
49 * reference count and return it.
50 *
51 * CONTEXT:
52 * Don't care.
53 *
54 * RETURNS:
55 * Pointer to the found partition on success, NULL if not found.
56 */
57struct hd_struct *disk_get_part(struct gendisk *disk, int partno)
58{
Tejun Heo540eed52008-08-25 19:56:15 +090059 struct hd_struct *part = NULL;
60 struct disk_part_tbl *ptbl;
Tejun Heoe71bf0d2008-09-03 09:03:02 +020061
Tejun Heo540eed52008-08-25 19:56:15 +090062 if (unlikely(partno < 0))
Tejun Heoe71bf0d2008-09-03 09:03:02 +020063 return NULL;
Tejun Heo540eed52008-08-25 19:56:15 +090064
Tejun Heoe71bf0d2008-09-03 09:03:02 +020065 rcu_read_lock();
Tejun Heo540eed52008-08-25 19:56:15 +090066
67 ptbl = rcu_dereference(disk->part_tbl);
68 if (likely(partno < ptbl->len)) {
69 part = rcu_dereference(ptbl->part[partno]);
70 if (part)
71 get_device(part_to_dev(part));
72 }
73
Tejun Heoe71bf0d2008-09-03 09:03:02 +020074 rcu_read_unlock();
75
76 return part;
77}
78EXPORT_SYMBOL_GPL(disk_get_part);
79
80/**
81 * disk_part_iter_init - initialize partition iterator
82 * @piter: iterator to initialize
83 * @disk: disk to iterate over
84 * @flags: DISK_PITER_* flags
85 *
86 * Initialize @piter so that it iterates over partitions of @disk.
87 *
88 * CONTEXT:
89 * Don't care.
90 */
91void disk_part_iter_init(struct disk_part_iter *piter, struct gendisk *disk,
92 unsigned int flags)
93{
Tejun Heo540eed52008-08-25 19:56:15 +090094 struct disk_part_tbl *ptbl;
95
96 rcu_read_lock();
97 ptbl = rcu_dereference(disk->part_tbl);
98
Tejun Heoe71bf0d2008-09-03 09:03:02 +020099 piter->disk = disk;
100 piter->part = NULL;
101
102 if (flags & DISK_PITER_REVERSE)
Tejun Heo540eed52008-08-25 19:56:15 +0900103 piter->idx = ptbl->len - 1;
Tejun Heo71982a42009-04-17 08:34:48 +0200104 else if (flags & (DISK_PITER_INCL_PART0 | DISK_PITER_INCL_EMPTY_PART0))
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200105 piter->idx = 0;
Tejun Heob5d0b9d2008-09-03 09:06:42 +0200106 else
107 piter->idx = 1;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200108
109 piter->flags = flags;
Tejun Heo540eed52008-08-25 19:56:15 +0900110
111 rcu_read_unlock();
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200112}
113EXPORT_SYMBOL_GPL(disk_part_iter_init);
114
115/**
116 * disk_part_iter_next - proceed iterator to the next partition and return it
117 * @piter: iterator of interest
118 *
119 * Proceed @piter to the next partition and return it.
120 *
121 * CONTEXT:
122 * Don't care.
123 */
124struct hd_struct *disk_part_iter_next(struct disk_part_iter *piter)
125{
Tejun Heo540eed52008-08-25 19:56:15 +0900126 struct disk_part_tbl *ptbl;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200127 int inc, end;
128
129 /* put the last partition */
130 disk_put_part(piter->part);
131 piter->part = NULL;
132
Tejun Heo540eed52008-08-25 19:56:15 +0900133 /* get part_tbl */
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200134 rcu_read_lock();
Tejun Heo540eed52008-08-25 19:56:15 +0900135 ptbl = rcu_dereference(piter->disk->part_tbl);
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200136
137 /* determine iteration parameters */
138 if (piter->flags & DISK_PITER_REVERSE) {
139 inc = -1;
Tejun Heo71982a42009-04-17 08:34:48 +0200140 if (piter->flags & (DISK_PITER_INCL_PART0 |
141 DISK_PITER_INCL_EMPTY_PART0))
Tejun Heob5d0b9d2008-09-03 09:06:42 +0200142 end = -1;
143 else
144 end = 0;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200145 } else {
146 inc = 1;
Tejun Heo540eed52008-08-25 19:56:15 +0900147 end = ptbl->len;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200148 }
149
150 /* iterate to the next partition */
151 for (; piter->idx != end; piter->idx += inc) {
152 struct hd_struct *part;
153
Tejun Heo540eed52008-08-25 19:56:15 +0900154 part = rcu_dereference(ptbl->part[piter->idx]);
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200155 if (!part)
156 continue;
Tejun Heo71982a42009-04-17 08:34:48 +0200157 if (!part->nr_sects &&
158 !(piter->flags & DISK_PITER_INCL_EMPTY) &&
159 !(piter->flags & DISK_PITER_INCL_EMPTY_PART0 &&
160 piter->idx == 0))
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200161 continue;
162
Tejun Heoed9e1982008-08-25 19:56:05 +0900163 get_device(part_to_dev(part));
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200164 piter->part = part;
165 piter->idx += inc;
166 break;
167 }
168
169 rcu_read_unlock();
170
171 return piter->part;
172}
173EXPORT_SYMBOL_GPL(disk_part_iter_next);
174
175/**
176 * disk_part_iter_exit - finish up partition iteration
177 * @piter: iter of interest
178 *
179 * Called when iteration is over. Cleans up @piter.
180 *
181 * CONTEXT:
182 * Don't care.
183 */
184void disk_part_iter_exit(struct disk_part_iter *piter)
185{
186 disk_put_part(piter->part);
187 piter->part = NULL;
188}
189EXPORT_SYMBOL_GPL(disk_part_iter_exit);
190
Jens Axboea6f23652008-10-24 12:52:42 +0200191static inline int sector_in_part(struct hd_struct *part, sector_t sector)
192{
193 return part->start_sect <= sector &&
194 sector < part->start_sect + part->nr_sects;
195}
196
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200197/**
198 * disk_map_sector_rcu - map sector to partition
199 * @disk: gendisk of interest
200 * @sector: sector to map
201 *
202 * Find out which partition @sector maps to on @disk. This is
203 * primarily used for stats accounting.
204 *
205 * CONTEXT:
206 * RCU read locked. The returned partition pointer is valid only
207 * while preemption is disabled.
208 *
209 * RETURNS:
Tejun Heo074a7ac2008-08-25 19:56:14 +0900210 * Found partition on success, part0 is returned if no partition matches
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200211 */
212struct hd_struct *disk_map_sector_rcu(struct gendisk *disk, sector_t sector)
213{
Tejun Heo540eed52008-08-25 19:56:15 +0900214 struct disk_part_tbl *ptbl;
Jens Axboea6f23652008-10-24 12:52:42 +0200215 struct hd_struct *part;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200216 int i;
217
Tejun Heo540eed52008-08-25 19:56:15 +0900218 ptbl = rcu_dereference(disk->part_tbl);
219
Jens Axboea6f23652008-10-24 12:52:42 +0200220 part = rcu_dereference(ptbl->last_lookup);
221 if (part && sector_in_part(part, sector))
222 return part;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200223
Jens Axboea6f23652008-10-24 12:52:42 +0200224 for (i = 1; i < ptbl->len; i++) {
225 part = rcu_dereference(ptbl->part[i]);
226
227 if (part && sector_in_part(part, sector)) {
228 rcu_assign_pointer(ptbl->last_lookup, part);
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200229 return part;
Jens Axboea6f23652008-10-24 12:52:42 +0200230 }
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200231 }
Tejun Heo074a7ac2008-08-25 19:56:14 +0900232 return &disk->part0;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200233}
234EXPORT_SYMBOL_GPL(disk_map_sector_rcu);
235
Linus Torvalds1da177e2005-04-16 15:20:36 -0700236/*
237 * Can be deleted altogether. Later.
238 *
239 */
240static struct blk_major_name {
241 struct blk_major_name *next;
242 int major;
243 char name[16];
Joe Korty68eef3b2006-03-31 02:30:32 -0800244} *major_names[BLKDEV_MAJOR_HASH_SIZE];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245
246/* index in the above - for now: assume no multimajor ranges */
Yang Zhange61eb2e2010-12-17 09:00:18 +0100247static inline int major_to_index(unsigned major)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248{
Joe Korty68eef3b2006-03-31 02:30:32 -0800249 return major % BLKDEV_MAJOR_HASH_SIZE;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250}
251
Joe Korty68eef3b2006-03-31 02:30:32 -0800252#ifdef CONFIG_PROC_FS
Tejun Heocf771cb2008-09-03 09:01:09 +0200253void blkdev_show(struct seq_file *seqf, off_t offset)
Neil Horman7170be52006-01-14 13:20:38 -0800254{
Joe Korty68eef3b2006-03-31 02:30:32 -0800255 struct blk_major_name *dp;
Neil Horman7170be52006-01-14 13:20:38 -0800256
Joe Korty68eef3b2006-03-31 02:30:32 -0800257 if (offset < BLKDEV_MAJOR_HASH_SIZE) {
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200258 mutex_lock(&block_class_lock);
Joe Korty68eef3b2006-03-31 02:30:32 -0800259 for (dp = major_names[offset]; dp; dp = dp->next)
Tejun Heocf771cb2008-09-03 09:01:09 +0200260 seq_printf(seqf, "%3d %s\n", dp->major, dp->name);
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200261 mutex_unlock(&block_class_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700263}
Joe Korty68eef3b2006-03-31 02:30:32 -0800264#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700265
Márton Németh9e8c0bc2009-02-20 08:12:51 +0100266/**
267 * register_blkdev - register a new block device
268 *
269 * @major: the requested major device number [1..255]. If @major=0, try to
270 * allocate any unused major number.
271 * @name: the name of the new block device as a zero terminated string
272 *
273 * The @name must be unique within the system.
274 *
275 * The return value depends on the @major input parameter.
276 * - if a major device number was requested in range [1..255] then the
277 * function returns zero on success, or a negative error code
278 * - if any unused major number was requested with @major=0 parameter
279 * then the return value is the allocated major number in range
280 * [1..255] or a negative error code otherwise
281 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282int register_blkdev(unsigned int major, const char *name)
283{
284 struct blk_major_name **n, *p;
285 int index, ret = 0;
286
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200287 mutex_lock(&block_class_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288
289 /* temporary */
290 if (major == 0) {
291 for (index = ARRAY_SIZE(major_names)-1; index > 0; index--) {
292 if (major_names[index] == NULL)
293 break;
294 }
295
296 if (index == 0) {
297 printk("register_blkdev: failed to get major for %s\n",
298 name);
299 ret = -EBUSY;
300 goto out;
301 }
302 major = index;
303 ret = major;
304 }
305
306 p = kmalloc(sizeof(struct blk_major_name), GFP_KERNEL);
307 if (p == NULL) {
308 ret = -ENOMEM;
309 goto out;
310 }
311
312 p->major = major;
313 strlcpy(p->name, name, sizeof(p->name));
314 p->next = NULL;
315 index = major_to_index(major);
316
317 for (n = &major_names[index]; *n; n = &(*n)->next) {
318 if ((*n)->major == major)
319 break;
320 }
321 if (!*n)
322 *n = p;
323 else
324 ret = -EBUSY;
325
326 if (ret < 0) {
327 printk("register_blkdev: cannot get major %d for %s\n",
328 major, name);
329 kfree(p);
330 }
331out:
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200332 mutex_unlock(&block_class_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333 return ret;
334}
335
336EXPORT_SYMBOL(register_blkdev);
337
Akinobu Mitaf4480242007-07-17 04:03:47 -0700338void unregister_blkdev(unsigned int major, const char *name)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700339{
340 struct blk_major_name **n;
341 struct blk_major_name *p = NULL;
342 int index = major_to_index(major);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700343
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200344 mutex_lock(&block_class_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345 for (n = &major_names[index]; *n; n = &(*n)->next)
346 if ((*n)->major == major)
347 break;
Akinobu Mita294462a2007-07-17 04:03:45 -0700348 if (!*n || strcmp((*n)->name, name)) {
349 WARN_ON(1);
Akinobu Mita294462a2007-07-17 04:03:45 -0700350 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351 p = *n;
352 *n = p->next;
353 }
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200354 mutex_unlock(&block_class_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355 kfree(p);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700356}
357
358EXPORT_SYMBOL(unregister_blkdev);
359
360static struct kobj_map *bdev_map;
361
Tejun Heobcce3de2008-08-25 19:47:22 +0900362/**
Tejun Heo870d6652008-08-25 19:47:25 +0900363 * blk_mangle_minor - scatter minor numbers apart
364 * @minor: minor number to mangle
365 *
366 * Scatter consecutively allocated @minor number apart if MANGLE_DEVT
367 * is enabled. Mangling twice gives the original value.
368 *
369 * RETURNS:
370 * Mangled value.
371 *
372 * CONTEXT:
373 * Don't care.
374 */
375static int blk_mangle_minor(int minor)
376{
377#ifdef CONFIG_DEBUG_BLOCK_EXT_DEVT
378 int i;
379
380 for (i = 0; i < MINORBITS / 2; i++) {
381 int low = minor & (1 << i);
382 int high = minor & (1 << (MINORBITS - 1 - i));
383 int distance = MINORBITS - 1 - 2 * i;
384
385 minor ^= low | high; /* clear both bits */
386 low <<= distance; /* swap the positions */
387 high >>= distance;
388 minor |= low | high; /* and set */
389 }
390#endif
391 return minor;
392}
393
394/**
Tejun Heobcce3de2008-08-25 19:47:22 +0900395 * blk_alloc_devt - allocate a dev_t for a partition
396 * @part: partition to allocate dev_t for
Tejun Heobcce3de2008-08-25 19:47:22 +0900397 * @devt: out parameter for resulting dev_t
398 *
399 * Allocate a dev_t for block device.
400 *
401 * RETURNS:
402 * 0 on success, allocated dev_t is returned in *@devt. -errno on
403 * failure.
404 *
405 * CONTEXT:
406 * Might sleep.
407 */
408int blk_alloc_devt(struct hd_struct *part, dev_t *devt)
409{
410 struct gendisk *disk = part_to_disk(part);
411 int idx, rc;
412
413 /* in consecutive minor range? */
414 if (part->partno < disk->minors) {
415 *devt = MKDEV(disk->major, disk->first_minor + part->partno);
416 return 0;
417 }
418
419 /* allocate ext devt */
420 do {
421 if (!idr_pre_get(&ext_devt_idr, GFP_KERNEL))
422 return -ENOMEM;
423 rc = idr_get_new(&ext_devt_idr, part, &idx);
424 } while (rc == -EAGAIN);
425
426 if (rc)
427 return rc;
428
429 if (idx > MAX_EXT_DEVT) {
430 idr_remove(&ext_devt_idr, idx);
431 return -EBUSY;
432 }
433
Tejun Heo870d6652008-08-25 19:47:25 +0900434 *devt = MKDEV(BLOCK_EXT_MAJOR, blk_mangle_minor(idx));
Tejun Heobcce3de2008-08-25 19:47:22 +0900435 return 0;
436}
437
438/**
439 * blk_free_devt - free a dev_t
440 * @devt: dev_t to free
441 *
442 * Free @devt which was allocated using blk_alloc_devt().
443 *
444 * CONTEXT:
445 * Might sleep.
446 */
447void blk_free_devt(dev_t devt)
448{
449 might_sleep();
450
451 if (devt == MKDEV(0, 0))
452 return;
453
454 if (MAJOR(devt) == BLOCK_EXT_MAJOR) {
455 mutex_lock(&ext_devt_mutex);
Tejun Heo870d6652008-08-25 19:47:25 +0900456 idr_remove(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
Tejun Heobcce3de2008-08-25 19:47:22 +0900457 mutex_unlock(&ext_devt_mutex);
458 }
459}
460
Tejun Heo1f014292008-08-25 19:47:23 +0900461static char *bdevt_str(dev_t devt, char *buf)
462{
463 if (MAJOR(devt) <= 0xff && MINOR(devt) <= 0xff) {
464 char tbuf[BDEVT_SIZE];
465 snprintf(tbuf, BDEVT_SIZE, "%02x%02x", MAJOR(devt), MINOR(devt));
466 snprintf(buf, BDEVT_SIZE, "%-9s", tbuf);
467 } else
468 snprintf(buf, BDEVT_SIZE, "%03x:%05x", MAJOR(devt), MINOR(devt));
469
470 return buf;
471}
472
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473/*
474 * Register device numbers dev..(dev+range-1)
475 * range must be nonzero
476 * The hash chain is sorted on range, so that subranges can override.
477 */
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200478void blk_register_region(dev_t devt, unsigned long range, struct module *module,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479 struct kobject *(*probe)(dev_t, int *, void *),
480 int (*lock)(dev_t, void *), void *data)
481{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200482 kobj_map(bdev_map, devt, range, module, probe, lock, data);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483}
484
485EXPORT_SYMBOL(blk_register_region);
486
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200487void blk_unregister_region(dev_t devt, unsigned long range)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200489 kobj_unmap(bdev_map, devt, range);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490}
491
492EXPORT_SYMBOL(blk_unregister_region);
493
Tejun Heocf771cb2008-09-03 09:01:09 +0200494static struct kobject *exact_match(dev_t devt, int *partno, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495{
496 struct gendisk *p = data;
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200497
Tejun Heoed9e1982008-08-25 19:56:05 +0900498 return &disk_to_dev(p)->kobj;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499}
500
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200501static int exact_lock(dev_t devt, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502{
503 struct gendisk *p = data;
504
505 if (!get_disk(p))
506 return -1;
507 return 0;
508}
509
Al Viro4752bc32011-09-16 00:21:54 -0400510static void register_disk(struct gendisk *disk)
Tejun Heod2bf1b62010-12-08 20:57:36 +0100511{
512 struct device *ddev = disk_to_dev(disk);
513 struct block_device *bdev;
514 struct disk_part_iter piter;
515 struct hd_struct *part;
516 int err;
517
518 ddev->parent = disk->driverfs_dev;
519
520 dev_set_name(ddev, disk->disk_name);
521
522 /* delay uevents, until we scanned partition table */
523 dev_set_uevent_suppress(ddev, 1);
524
525 if (device_add(ddev))
526 return;
527 if (!sysfs_deprecated) {
528 err = sysfs_create_link(block_depr, &ddev->kobj,
529 kobject_name(&ddev->kobj));
530 if (err) {
531 device_del(ddev);
532 return;
533 }
534 }
535 disk->part0.holder_dir = kobject_create_and_add("holders", &ddev->kobj);
536 disk->slave_dir = kobject_create_and_add("slaves", &ddev->kobj);
537
538 /* No minors to use for partitions */
Tejun Heod27769e2011-08-23 20:01:04 +0200539 if (!disk_part_scan_enabled(disk))
Tejun Heod2bf1b62010-12-08 20:57:36 +0100540 goto exit;
541
542 /* No such device (e.g., media were just removed) */
543 if (!get_capacity(disk))
544 goto exit;
545
546 bdev = bdget_disk(disk, 0);
547 if (!bdev)
548 goto exit;
549
550 bdev->bd_invalidated = 1;
551 err = blkdev_get(bdev, FMODE_READ, NULL);
552 if (err < 0)
553 goto exit;
554 blkdev_put(bdev, FMODE_READ);
555
556exit:
557 /* announce disk after possible partitions are created */
558 dev_set_uevent_suppress(ddev, 0);
559 kobject_uevent(&ddev->kobj, KOBJ_ADD);
560
561 /* announce possible partitions */
562 disk_part_iter_init(&piter, disk, 0);
563 while ((part = disk_part_iter_next(&piter)))
564 kobject_uevent(&part_to_dev(part)->kobj, KOBJ_ADD);
565 disk_part_iter_exit(&piter);
566}
567
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568/**
569 * add_disk - add partitioning information to kernel list
570 * @disk: per-device partitioning information
571 *
572 * This function registers the partitioning information in @disk
573 * with the kernel.
Tejun Heo3e1a7ff2008-08-25 19:56:17 +0900574 *
575 * FIXME: error handling
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576 */
577void add_disk(struct gendisk *disk)
578{
Peter Zijlstracf0ca9f2008-04-30 00:54:32 -0700579 struct backing_dev_info *bdi;
Tejun Heo3e1a7ff2008-08-25 19:56:17 +0900580 dev_t devt;
Greg Kroah-Hartman6ffeea72008-05-22 17:21:08 -0400581 int retval;
Peter Zijlstracf0ca9f2008-04-30 00:54:32 -0700582
Tejun Heo3e1a7ff2008-08-25 19:56:17 +0900583 /* minors == 0 indicates to use ext devt from part0 and should
584 * be accompanied with EXT_DEVT flag. Make sure all
585 * parameters make sense.
586 */
587 WARN_ON(disk->minors && !(disk->major || disk->first_minor));
588 WARN_ON(!disk->minors && !(disk->flags & GENHD_FL_EXT_DEVT));
589
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590 disk->flags |= GENHD_FL_UP;
Tejun Heo3e1a7ff2008-08-25 19:56:17 +0900591
592 retval = blk_alloc_devt(&disk->part0, &devt);
593 if (retval) {
594 WARN_ON(1);
595 return;
596 }
597 disk_to_dev(disk)->devt = devt;
598
599 /* ->major and ->first_minor aren't supposed to be
600 * dereferenced from here on, but set them just in case.
601 */
602 disk->major = MAJOR(devt);
603 disk->first_minor = MINOR(devt);
604
Stanislaw Gruszka9f53d2fe2012-03-02 10:43:28 +0100605 disk_alloc_events(disk);
606
Wanlong Gao9f5e4862011-06-13 10:45:43 +0200607 /* Register BDI before referencing it from bdev */
Signed-off-by: Jan Kara01ea5062010-09-16 20:36:36 +0200608 bdi = &disk->queue->backing_dev_info;
609 bdi_register_dev(bdi, disk_devt(disk));
610
Tejun Heof331c022008-09-03 09:01:48 +0200611 blk_register_region(disk_devt(disk), disk->minors, NULL,
612 exact_match, exact_lock, disk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700613 register_disk(disk);
614 blk_register_queue(disk);
Peter Zijlstracf0ca9f2008-04-30 00:54:32 -0700615
Tejun Heo523e1d32011-10-19 14:31:07 +0200616 /*
617 * Take an extra ref on queue which will be put on disk_release()
618 * so that it sticks around as long as @disk is there.
619 */
Tejun Heo09ac46c2011-12-14 00:33:38 +0100620 WARN_ON_ONCE(!blk_get_queue(disk->queue));
Tejun Heo523e1d32011-10-19 14:31:07 +0200621
Tejun Heoed9e1982008-08-25 19:56:05 +0900622 retval = sysfs_create_link(&disk_to_dev(disk)->kobj, &bdi->dev->kobj,
623 "bdi");
Greg Kroah-Hartman6ffeea72008-05-22 17:21:08 -0400624 WARN_ON(retval);
Tejun Heo77ea8872010-12-08 20:57:37 +0100625
626 disk_add_events(disk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700628EXPORT_SYMBOL(add_disk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629
Tejun Heod2bf1b62010-12-08 20:57:36 +0100630void del_gendisk(struct gendisk *disk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631{
Tejun Heod2bf1b62010-12-08 20:57:36 +0100632 struct disk_part_iter piter;
633 struct hd_struct *part;
634
Tejun Heo77ea8872010-12-08 20:57:37 +0100635 disk_del_events(disk);
636
Tejun Heod2bf1b62010-12-08 20:57:36 +0100637 /* invalidate stuff */
638 disk_part_iter_init(&piter, disk,
639 DISK_PITER_INCL_EMPTY | DISK_PITER_REVERSE);
640 while ((part = disk_part_iter_next(&piter))) {
641 invalidate_partition(disk, part->partno);
642 delete_partition(disk, part->partno);
643 }
644 disk_part_iter_exit(&piter);
645
646 invalidate_partition(disk, 0);
647 blk_free_devt(disk_to_dev(disk)->devt);
648 set_capacity(disk, 0);
649 disk->flags &= ~GENHD_FL_UP;
650
Tejun Heoed9e1982008-08-25 19:56:05 +0900651 sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
Peter Zijlstracf0ca9f2008-04-30 00:54:32 -0700652 bdi_unregister(&disk->queue->backing_dev_info);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653 blk_unregister_queue(disk);
Tejun Heof331c022008-09-03 09:01:48 +0200654 blk_unregister_region(disk_devt(disk), disk->minors);
Tejun Heod2bf1b62010-12-08 20:57:36 +0100655
656 part_stat_set_all(&disk->part0, 0);
657 disk->part0.stamp = 0;
658
659 kobject_put(disk->part0.holder_dir);
660 kobject_put(disk->slave_dir);
661 disk->driverfs_dev = NULL;
662 if (!sysfs_deprecated)
663 sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
664 device_del(disk_to_dev(disk));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665}
Tejun Heod2bf1b62010-12-08 20:57:36 +0100666EXPORT_SYMBOL(del_gendisk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700667
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668/**
669 * get_gendisk - get partitioning information for a given device
Randy Dunlap710027a2008-08-19 20:13:11 +0200670 * @devt: device to get partitioning information for
Randy Dunlap496aa8a2008-10-16 07:46:23 +0200671 * @partno: returned partition index
Linus Torvalds1da177e2005-04-16 15:20:36 -0700672 *
673 * This function gets the structure containing partitioning
Randy Dunlap710027a2008-08-19 20:13:11 +0200674 * information for the given device @devt.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675 */
Tejun Heocf771cb2008-09-03 09:01:09 +0200676struct gendisk *get_gendisk(dev_t devt, int *partno)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677{
Tejun Heobcce3de2008-08-25 19:47:22 +0900678 struct gendisk *disk = NULL;
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200679
Tejun Heobcce3de2008-08-25 19:47:22 +0900680 if (MAJOR(devt) != BLOCK_EXT_MAJOR) {
681 struct kobject *kobj;
682
683 kobj = kobj_lookup(bdev_map, devt, partno);
684 if (kobj)
685 disk = dev_to_disk(kobj_to_dev(kobj));
686 } else {
687 struct hd_struct *part;
688
689 mutex_lock(&ext_devt_mutex);
Tejun Heo870d6652008-08-25 19:47:25 +0900690 part = idr_find(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
Tejun Heobcce3de2008-08-25 19:47:22 +0900691 if (part && get_disk(part_to_disk(part))) {
692 *partno = part->partno;
693 disk = part_to_disk(part);
694 }
695 mutex_unlock(&ext_devt_mutex);
696 }
697
698 return disk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700699}
Divyesh Shahb6ac23af2010-04-15 08:54:59 +0200700EXPORT_SYMBOL(get_gendisk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701
Tejun Heof331c022008-09-03 09:01:48 +0200702/**
703 * bdget_disk - do bdget() by gendisk and partition number
704 * @disk: gendisk of interest
705 * @partno: partition number
706 *
707 * Find partition @partno from @disk, do bdget() on it.
708 *
709 * CONTEXT:
710 * Don't care.
711 *
712 * RETURNS:
713 * Resulting block_device on success, NULL on failure.
714 */
Harvey Harrisonaeb3d3a2008-08-28 09:27:42 +0200715struct block_device *bdget_disk(struct gendisk *disk, int partno)
Tejun Heof331c022008-09-03 09:01:48 +0200716{
Tejun Heo548b10e2008-08-29 09:01:47 +0200717 struct hd_struct *part;
718 struct block_device *bdev = NULL;
Tejun Heof331c022008-09-03 09:01:48 +0200719
Tejun Heo548b10e2008-08-29 09:01:47 +0200720 part = disk_get_part(disk, partno);
Tejun Heo2bbedcb2008-08-29 11:41:51 +0200721 if (part)
Tejun Heo548b10e2008-08-29 09:01:47 +0200722 bdev = bdget(part_devt(part));
723 disk_put_part(part);
Tejun Heof331c022008-09-03 09:01:48 +0200724
Tejun Heo548b10e2008-08-29 09:01:47 +0200725 return bdev;
Tejun Heof331c022008-09-03 09:01:48 +0200726}
727EXPORT_SYMBOL(bdget_disk);
728
Dave Gilbertdd2a3452007-05-09 02:33:24 -0700729/*
730 * print a full list of all partitions - intended for places where the root
731 * filesystem can't be mounted and thus to give the victim some idea of what
732 * went wrong
733 */
734void __init printk_all_partitions(void)
735{
Tejun Heodef4e382008-09-03 08:57:12 +0200736 struct class_dev_iter iter;
737 struct device *dev;
738
739 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
740 while ((dev = class_dev_iter_next(&iter))) {
741 struct gendisk *disk = dev_to_disk(dev);
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200742 struct disk_part_iter piter;
743 struct hd_struct *part;
Tejun Heo1f014292008-08-25 19:47:23 +0900744 char name_buf[BDEVNAME_SIZE];
745 char devt_buf[BDEVT_SIZE];
Will Drewryb5af9212010-08-31 15:47:07 -0500746 u8 uuid[PARTITION_META_INFO_UUIDLTH * 2 + 1];
Tejun Heodef4e382008-09-03 08:57:12 +0200747
748 /*
749 * Don't show empty devices or things that have been
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300750 * suppressed
Tejun Heodef4e382008-09-03 08:57:12 +0200751 */
752 if (get_capacity(disk) == 0 ||
753 (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
754 continue;
755
756 /*
757 * Note, unlike /proc/partitions, I am showing the
758 * numbers in hex - the same format as the root=
759 * option takes.
760 */
Tejun Heo074a7ac2008-08-25 19:56:14 +0900761 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0);
762 while ((part = disk_part_iter_next(&piter))) {
763 bool is_part0 = part == &disk->part0;
Tejun Heodef4e382008-09-03 08:57:12 +0200764
Will Drewryb5af9212010-08-31 15:47:07 -0500765 uuid[0] = 0;
766 if (part->info)
767 part_unpack_uuid(part->info->uuid, uuid);
768
769 printk("%s%s %10llu %s %s", is_part0 ? "" : " ",
Tejun Heo1f014292008-08-25 19:47:23 +0900770 bdevt_str(part_devt(part), devt_buf),
Tejun Heof331c022008-09-03 09:01:48 +0200771 (unsigned long long)part->nr_sects >> 1,
Will Drewryb5af9212010-08-31 15:47:07 -0500772 disk_name(disk, part->partno, name_buf), uuid);
Tejun Heo074a7ac2008-08-25 19:56:14 +0900773 if (is_part0) {
774 if (disk->driverfs_dev != NULL &&
775 disk->driverfs_dev->driver != NULL)
776 printk(" driver: %s\n",
777 disk->driverfs_dev->driver->name);
778 else
779 printk(" (driver?)\n");
780 } else
781 printk("\n");
782 }
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200783 disk_part_iter_exit(&piter);
Tejun Heodef4e382008-09-03 08:57:12 +0200784 }
785 class_dev_iter_exit(&iter);
Dave Gilbertdd2a3452007-05-09 02:33:24 -0700786}
787
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788#ifdef CONFIG_PROC_FS
789/* iterator */
Tejun Heodef4e382008-09-03 08:57:12 +0200790static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
Greg Kroah-Hartman68c4d4a2008-05-22 17:21:08 -0400791{
Tejun Heodef4e382008-09-03 08:57:12 +0200792 loff_t skip = *pos;
793 struct class_dev_iter *iter;
794 struct device *dev;
Greg Kroah-Hartman68c4d4a2008-05-22 17:21:08 -0400795
Harvey Harrisonaeb3d3a2008-08-28 09:27:42 +0200796 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
Tejun Heodef4e382008-09-03 08:57:12 +0200797 if (!iter)
798 return ERR_PTR(-ENOMEM);
799
800 seqf->private = iter;
801 class_dev_iter_init(iter, &block_class, NULL, &disk_type);
802 do {
803 dev = class_dev_iter_next(iter);
804 if (!dev)
805 return NULL;
806 } while (skip--);
807
808 return dev_to_disk(dev);
Greg Kroah-Hartman68c4d4a2008-05-22 17:21:08 -0400809}
810
Tejun Heodef4e382008-09-03 08:57:12 +0200811static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700812{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200813 struct device *dev;
Greg Kroah-Hartman66c64af2008-05-22 17:21:08 -0400814
Tejun Heodef4e382008-09-03 08:57:12 +0200815 (*pos)++;
816 dev = class_dev_iter_next(seqf->private);
Tejun Heo2ac3cee2008-09-03 08:53:37 +0200817 if (dev)
Greg Kroah-Hartman68c4d4a2008-05-22 17:21:08 -0400818 return dev_to_disk(dev);
Tejun Heo2ac3cee2008-09-03 08:53:37 +0200819
Linus Torvalds1da177e2005-04-16 15:20:36 -0700820 return NULL;
821}
822
Tejun Heodef4e382008-09-03 08:57:12 +0200823static void disk_seqf_stop(struct seq_file *seqf, void *v)
Greg Kroah-Hartman27f30252008-05-22 17:21:08 -0400824{
Tejun Heodef4e382008-09-03 08:57:12 +0200825 struct class_dev_iter *iter = seqf->private;
Greg Kroah-Hartman27f30252008-05-22 17:21:08 -0400826
Tejun Heodef4e382008-09-03 08:57:12 +0200827 /* stop is called even after start failed :-( */
828 if (iter) {
829 class_dev_iter_exit(iter);
830 kfree(iter);
Kay Sievers5c0ef6d2008-08-16 14:30:30 +0200831 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700832}
833
Tejun Heodef4e382008-09-03 08:57:12 +0200834static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835{
Tejun Heodef4e382008-09-03 08:57:12 +0200836 static void *p;
837
838 p = disk_seqf_start(seqf, pos);
Yang Zhangb9f985b2010-12-17 08:58:36 +0100839 if (!IS_ERR_OR_NULL(p) && !*pos)
Tejun Heodef4e382008-09-03 08:57:12 +0200840 seq_puts(seqf, "major minor #blocks name\n\n");
841 return p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700842}
843
Tejun Heocf771cb2008-09-03 09:01:09 +0200844static int show_partition(struct seq_file *seqf, void *v)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845{
846 struct gendisk *sgp = v;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200847 struct disk_part_iter piter;
848 struct hd_struct *part;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849 char buf[BDEVNAME_SIZE];
850
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851 /* Don't show non-partitionable removeable devices or empty devices */
Tejun Heod27769e2011-08-23 20:01:04 +0200852 if (!get_capacity(sgp) || (!disk_max_parts(sgp) &&
Tejun Heof331c022008-09-03 09:01:48 +0200853 (sgp->flags & GENHD_FL_REMOVABLE)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854 return 0;
855 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
856 return 0;
857
858 /* show the full disk and all non-0 size partitions of it */
Tejun Heo074a7ac2008-08-25 19:56:14 +0900859 disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0);
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200860 while ((part = disk_part_iter_next(&piter)))
Tejun Heo1f014292008-08-25 19:47:23 +0900861 seq_printf(seqf, "%4d %7d %10llu %s\n",
Tejun Heof331c022008-09-03 09:01:48 +0200862 MAJOR(part_devt(part)), MINOR(part_devt(part)),
863 (unsigned long long)part->nr_sects >> 1,
864 disk_name(sgp, part->partno, buf));
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200865 disk_part_iter_exit(&piter);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866
867 return 0;
868}
869
Alexey Dobriyanf5009752008-10-04 23:53:21 +0400870static const struct seq_operations partitions_op = {
Tejun Heodef4e382008-09-03 08:57:12 +0200871 .start = show_partition_start,
872 .next = disk_seqf_next,
873 .stop = disk_seqf_stop,
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200874 .show = show_partition
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875};
Alexey Dobriyanf5009752008-10-04 23:53:21 +0400876
877static int partitions_open(struct inode *inode, struct file *file)
878{
879 return seq_open(file, &partitions_op);
880}
881
882static const struct file_operations proc_partitions_operations = {
883 .open = partitions_open,
884 .read = seq_read,
885 .llseek = seq_lseek,
886 .release = seq_release,
887};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888#endif
889
890
Tejun Heocf771cb2008-09-03 09:01:09 +0200891static struct kobject *base_probe(dev_t devt, int *partno, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700892{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200893 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894 /* Make old-style 2.4 aliases work */
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200895 request_module("block-major-%d", MAJOR(devt));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896 return NULL;
897}
898
899static int __init genhd_device_init(void)
900{
Dan Williamse105b8b2008-04-21 10:51:07 -0700901 int error;
902
903 block_class.dev_kobj = sysfs_dev_block_kobj;
904 error = class_register(&block_class);
Roland McGrathee27a552008-03-11 17:13:15 -0700905 if (unlikely(error))
906 return error;
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200907 bdev_map = kobj_map_init(base_probe, &block_class_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700908 blk_dev_init();
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200909
Zhang, Yanmin561ec682008-11-14 08:26:30 +0100910 register_blkdev(BLOCK_EXT_MAJOR, "blkext");
911
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200912 /* create top-level block dir */
Andi Kleene52eec12010-09-08 16:54:17 +0200913 if (!sysfs_deprecated)
914 block_depr = kobject_create_and_add("block", NULL);
Greg Kroah-Hartman830d3cf2007-11-06 10:36:58 -0800915 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916}
917
918subsys_initcall(genhd_device_init);
919
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200920static ssize_t disk_range_show(struct device *dev,
921 struct device_attribute *attr, char *buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200923 struct gendisk *disk = dev_to_disk(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200925 return sprintf(buf, "%d\n", disk->minors);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926}
927
Tejun Heo1f014292008-08-25 19:47:23 +0900928static ssize_t disk_ext_range_show(struct device *dev,
929 struct device_attribute *attr, char *buf)
930{
931 struct gendisk *disk = dev_to_disk(dev);
932
Tejun Heob5d0b9d2008-09-03 09:06:42 +0200933 return sprintf(buf, "%d\n", disk_max_parts(disk));
Tejun Heo1f014292008-08-25 19:47:23 +0900934}
935
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200936static ssize_t disk_removable_show(struct device *dev,
937 struct device_attribute *attr, char *buf)
Kay Sieversa7fd6702005-10-01 14:49:43 +0200938{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200939 struct gendisk *disk = dev_to_disk(dev);
Kay Sieversa7fd6702005-10-01 14:49:43 +0200940
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200941 return sprintf(buf, "%d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700942 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200943}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700944
Kay Sievers1c9ce522008-06-13 09:41:00 +0200945static ssize_t disk_ro_show(struct device *dev,
946 struct device_attribute *attr, char *buf)
947{
948 struct gendisk *disk = dev_to_disk(dev);
949
Tejun Heob7db9952008-08-25 19:56:10 +0900950 return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
Kay Sievers1c9ce522008-06-13 09:41:00 +0200951}
952
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200953static ssize_t disk_capability_show(struct device *dev,
954 struct device_attribute *attr, char *buf)
Kristen Carlson Accardi86ce18d2007-05-23 13:57:38 -0700955{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200956 struct gendisk *disk = dev_to_disk(dev);
957
958 return sprintf(buf, "%x\n", disk->flags);
Kristen Carlson Accardi86ce18d2007-05-23 13:57:38 -0700959}
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200960
Martin K. Petersenc72758f2009-05-22 17:17:53 -0400961static ssize_t disk_alignment_offset_show(struct device *dev,
962 struct device_attribute *attr,
963 char *buf)
964{
965 struct gendisk *disk = dev_to_disk(dev);
966
967 return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue));
968}
969
Martin K. Petersen86b37282009-11-10 11:50:21 +0100970static ssize_t disk_discard_alignment_show(struct device *dev,
971 struct device_attribute *attr,
972 char *buf)
973{
974 struct gendisk *disk = dev_to_disk(dev);
975
Martin K. Petersendd3d1452010-01-11 03:21:48 -0500976 return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue));
Martin K. Petersen86b37282009-11-10 11:50:21 +0100977}
978
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200979static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL);
Tejun Heo1f014292008-08-25 19:47:23 +0900980static DEVICE_ATTR(ext_range, S_IRUGO, disk_ext_range_show, NULL);
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200981static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL);
Kay Sievers1c9ce522008-06-13 09:41:00 +0200982static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL);
Tejun Heoe5610522008-08-25 19:56:09 +0900983static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL);
Martin K. Petersenc72758f2009-05-22 17:17:53 -0400984static DEVICE_ATTR(alignment_offset, S_IRUGO, disk_alignment_offset_show, NULL);
Martin K. Petersen86b37282009-11-10 11:50:21 +0100985static DEVICE_ATTR(discard_alignment, S_IRUGO, disk_discard_alignment_show,
986 NULL);
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200987static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL);
Tejun Heo074a7ac2008-08-25 19:56:14 +0900988static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL);
Nikanth Karthikesan316d3152009-10-06 20:16:55 +0200989static DEVICE_ATTR(inflight, S_IRUGO, part_inflight_show, NULL);
Akinobu Mitac17bb492006-12-08 02:39:46 -0800990#ifdef CONFIG_FAIL_MAKE_REQUEST
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200991static struct device_attribute dev_attr_fail =
Tejun Heoeddb2e22008-08-25 19:56:13 +0900992 __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store);
Akinobu Mitac17bb492006-12-08 02:39:46 -0800993#endif
Jens Axboe581d4e22008-09-14 05:56:33 -0700994#ifdef CONFIG_FAIL_IO_TIMEOUT
995static struct device_attribute dev_attr_fail_timeout =
996 __ATTR(io-timeout-fail, S_IRUGO|S_IWUSR, part_timeout_show,
997 part_timeout_store);
998#endif
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200999
1000static struct attribute *disk_attrs[] = {
1001 &dev_attr_range.attr,
Tejun Heo1f014292008-08-25 19:47:23 +09001002 &dev_attr_ext_range.attr,
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001003 &dev_attr_removable.attr,
Kay Sievers1c9ce522008-06-13 09:41:00 +02001004 &dev_attr_ro.attr,
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001005 &dev_attr_size.attr,
Martin K. Petersenc72758f2009-05-22 17:17:53 -04001006 &dev_attr_alignment_offset.attr,
Martin K. Petersen86b37282009-11-10 11:50:21 +01001007 &dev_attr_discard_alignment.attr,
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001008 &dev_attr_capability.attr,
1009 &dev_attr_stat.attr,
Nikanth Karthikesan316d3152009-10-06 20:16:55 +02001010 &dev_attr_inflight.attr,
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001011#ifdef CONFIG_FAIL_MAKE_REQUEST
1012 &dev_attr_fail.attr,
1013#endif
Jens Axboe581d4e22008-09-14 05:56:33 -07001014#ifdef CONFIG_FAIL_IO_TIMEOUT
1015 &dev_attr_fail_timeout.attr,
1016#endif
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001017 NULL
Linus Torvalds1da177e2005-04-16 15:20:36 -07001018};
1019
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001020static struct attribute_group disk_attr_group = {
1021 .attrs = disk_attrs,
1022};
1023
David Brownella4dbd672009-06-24 10:06:31 -07001024static const struct attribute_group *disk_attr_groups[] = {
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001025 &disk_attr_group,
1026 NULL
1027};
1028
Tejun Heo540eed52008-08-25 19:56:15 +09001029/**
1030 * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way
1031 * @disk: disk to replace part_tbl for
1032 * @new_ptbl: new part_tbl to install
1033 *
1034 * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The
1035 * original ptbl is freed using RCU callback.
1036 *
1037 * LOCKING:
1038 * Matching bd_mutx locked.
1039 */
1040static void disk_replace_part_tbl(struct gendisk *disk,
1041 struct disk_part_tbl *new_ptbl)
1042{
1043 struct disk_part_tbl *old_ptbl = disk->part_tbl;
1044
1045 rcu_assign_pointer(disk->part_tbl, new_ptbl);
Jens Axboea6f23652008-10-24 12:52:42 +02001046
1047 if (old_ptbl) {
1048 rcu_assign_pointer(old_ptbl->last_lookup, NULL);
Lai Jiangshan57bdfbf2011-03-18 11:42:58 +08001049 kfree_rcu(old_ptbl, rcu_head);
Jens Axboea6f23652008-10-24 12:52:42 +02001050 }
Tejun Heo540eed52008-08-25 19:56:15 +09001051}
1052
1053/**
1054 * disk_expand_part_tbl - expand disk->part_tbl
1055 * @disk: disk to expand part_tbl for
1056 * @partno: expand such that this partno can fit in
1057 *
1058 * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl
1059 * uses RCU to allow unlocked dereferencing for stats and other stuff.
1060 *
1061 * LOCKING:
1062 * Matching bd_mutex locked, might sleep.
1063 *
1064 * RETURNS:
1065 * 0 on success, -errno on failure.
1066 */
1067int disk_expand_part_tbl(struct gendisk *disk, int partno)
1068{
1069 struct disk_part_tbl *old_ptbl = disk->part_tbl;
1070 struct disk_part_tbl *new_ptbl;
1071 int len = old_ptbl ? old_ptbl->len : 0;
1072 int target = partno + 1;
1073 size_t size;
1074 int i;
1075
1076 /* disk_max_parts() is zero during initialization, ignore if so */
1077 if (disk_max_parts(disk) && target > disk_max_parts(disk))
1078 return -EINVAL;
1079
1080 if (target <= len)
1081 return 0;
1082
1083 size = sizeof(*new_ptbl) + target * sizeof(new_ptbl->part[0]);
1084 new_ptbl = kzalloc_node(size, GFP_KERNEL, disk->node_id);
1085 if (!new_ptbl)
1086 return -ENOMEM;
1087
Tejun Heo540eed52008-08-25 19:56:15 +09001088 new_ptbl->len = target;
1089
1090 for (i = 0; i < len; i++)
1091 rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]);
1092
1093 disk_replace_part_tbl(disk, new_ptbl);
1094 return 0;
1095}
1096
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001097static void disk_release(struct device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001098{
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001099 struct gendisk *disk = dev_to_disk(dev);
1100
Tejun Heo77ea8872010-12-08 20:57:37 +01001101 disk_release_events(disk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102 kfree(disk->random);
Tejun Heo540eed52008-08-25 19:56:15 +09001103 disk_replace_part_tbl(disk, NULL);
Tejun Heo074a7ac2008-08-25 19:56:14 +09001104 free_part_stats(&disk->part0);
Will Drewry6d1d8052010-08-31 15:47:05 -05001105 free_part_info(&disk->part0);
Tejun Heo523e1d32011-10-19 14:31:07 +02001106 if (disk->queue)
1107 blk_put_queue(disk->queue);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108 kfree(disk);
1109}
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001110struct class block_class = {
1111 .name = "block",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112};
1113
Al Viro2c9ede52011-07-23 20:24:48 -04001114static char *block_devnode(struct device *dev, umode_t *mode)
Kay Sieversb03f38b2009-04-30 15:23:42 +02001115{
1116 struct gendisk *disk = dev_to_disk(dev);
1117
Kay Sieverse454cea2009-09-18 23:01:12 +02001118 if (disk->devnode)
1119 return disk->devnode(disk, mode);
Kay Sieversb03f38b2009-04-30 15:23:42 +02001120 return NULL;
1121}
1122
Adrian Bunk1826ead2008-03-04 11:23:46 +01001123static struct device_type disk_type = {
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001124 .name = "disk",
1125 .groups = disk_attr_groups,
1126 .release = disk_release,
Kay Sieverse454cea2009-09-18 23:01:12 +02001127 .devnode = block_devnode,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128};
1129
Randy Dunlapa6e2ba82008-05-23 09:44:11 -07001130#ifdef CONFIG_PROC_FS
Tejun Heocf771cb2008-09-03 09:01:09 +02001131/*
1132 * aggregate disk stat collector. Uses the same stats that the sysfs
1133 * entries do, above, but makes them available through one seq_file.
1134 *
1135 * The output looks suspiciously like /proc/partitions with a bunch of
1136 * extra fields.
1137 */
1138static int diskstats_show(struct seq_file *seqf, void *v)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139{
1140 struct gendisk *gp = v;
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001141 struct disk_part_iter piter;
1142 struct hd_struct *hd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001143 char buf[BDEVNAME_SIZE];
Tejun Heoc9959052008-08-25 19:47:21 +09001144 int cpu;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001145
1146 /*
Tejun Heoed9e1982008-08-25 19:56:05 +09001147 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
Tejun Heocf771cb2008-09-03 09:01:09 +02001148 seq_puts(seqf, "major minor name"
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149 " rio rmerge rsect ruse wio wmerge "
1150 "wsect wuse running use aveq"
1151 "\n\n");
1152 */
Wanlong Gao9f5e4862011-06-13 10:45:43 +02001153
Tejun Heo71982a42009-04-17 08:34:48 +02001154 disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0);
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001155 while ((hd = disk_part_iter_next(&piter))) {
Tejun Heo074a7ac2008-08-25 19:56:14 +09001156 cpu = part_stat_lock();
Tejun Heoc9959052008-08-25 19:47:21 +09001157 part_round_stats(cpu, hd);
Tejun Heo074a7ac2008-08-25 19:56:14 +09001158 part_stat_unlock();
Herbert Poetzlf95fe9c2011-08-02 12:43:50 +02001159 seq_printf(seqf, "%4d %7d %s %lu %lu %lu "
1160 "%u %lu %lu %lu %u %u %u %u\n",
Tejun Heof331c022008-09-03 09:01:48 +02001161 MAJOR(part_devt(hd)), MINOR(part_devt(hd)),
1162 disk_name(gp, hd->partno, buf),
Liu Yuan53f22952011-03-02 11:00:15 -05001163 part_stat_read(hd, ios[READ]),
1164 part_stat_read(hd, merges[READ]),
Herbert Poetzlf95fe9c2011-08-02 12:43:50 +02001165 part_stat_read(hd, sectors[READ]),
Liu Yuan53f22952011-03-02 11:00:15 -05001166 jiffies_to_msecs(part_stat_read(hd, ticks[READ])),
1167 part_stat_read(hd, ios[WRITE]),
1168 part_stat_read(hd, merges[WRITE]),
Herbert Poetzlf95fe9c2011-08-02 12:43:50 +02001169 part_stat_read(hd, sectors[WRITE]),
Liu Yuan53f22952011-03-02 11:00:15 -05001170 jiffies_to_msecs(part_stat_read(hd, ticks[WRITE])),
Nikanth Karthikesan316d3152009-10-06 20:16:55 +02001171 part_in_flight(hd),
Jerome Marchand28f39d52008-02-08 11:04:56 +01001172 jiffies_to_msecs(part_stat_read(hd, io_ticks)),
1173 jiffies_to_msecs(part_stat_read(hd, time_in_queue))
1174 );
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175 }
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001176 disk_part_iter_exit(&piter);
Wanlong Gao9f5e4862011-06-13 10:45:43 +02001177
Linus Torvalds1da177e2005-04-16 15:20:36 -07001178 return 0;
1179}
1180
Alexey Dobriyan31d85ab2008-10-06 12:55:38 +04001181static const struct seq_operations diskstats_op = {
Tejun Heodef4e382008-09-03 08:57:12 +02001182 .start = disk_seqf_start,
1183 .next = disk_seqf_next,
1184 .stop = disk_seqf_stop,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001185 .show = diskstats_show
1186};
Alexey Dobriyanf5009752008-10-04 23:53:21 +04001187
Alexey Dobriyan31d85ab2008-10-06 12:55:38 +04001188static int diskstats_open(struct inode *inode, struct file *file)
1189{
1190 return seq_open(file, &diskstats_op);
1191}
1192
1193static const struct file_operations proc_diskstats_operations = {
1194 .open = diskstats_open,
1195 .read = seq_read,
1196 .llseek = seq_lseek,
1197 .release = seq_release,
1198};
1199
Alexey Dobriyanf5009752008-10-04 23:53:21 +04001200static int __init proc_genhd_init(void)
1201{
Alexey Dobriyan31d85ab2008-10-06 12:55:38 +04001202 proc_create("diskstats", 0, NULL, &proc_diskstats_operations);
Alexey Dobriyanf5009752008-10-04 23:53:21 +04001203 proc_create("partitions", 0, NULL, &proc_partitions_operations);
1204 return 0;
1205}
1206module_init(proc_genhd_init);
Randy Dunlapa6e2ba82008-05-23 09:44:11 -07001207#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001208
Tejun Heocf771cb2008-09-03 09:01:09 +02001209dev_t blk_lookup_devt(const char *name, int partno)
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001210{
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001211 dev_t devt = MKDEV(0, 0);
Tejun Heodef4e382008-09-03 08:57:12 +02001212 struct class_dev_iter iter;
1213 struct device *dev;
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001214
Tejun Heodef4e382008-09-03 08:57:12 +02001215 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1216 while ((dev = class_dev_iter_next(&iter))) {
1217 struct gendisk *disk = dev_to_disk(dev);
Tejun Heo548b10e2008-08-29 09:01:47 +02001218 struct hd_struct *part;
Tejun Heodef4e382008-09-03 08:57:12 +02001219
Kay Sievers3ada8b72009-01-06 10:44:43 -08001220 if (strcmp(dev_name(dev), name))
Tejun Heof331c022008-09-03 09:01:48 +02001221 continue;
Tejun Heof331c022008-09-03 09:01:48 +02001222
Neil Brown41b8c852009-02-18 10:33:59 +01001223 if (partno < disk->minors) {
1224 /* We need to return the right devno, even
1225 * if the partition doesn't exist yet.
1226 */
1227 devt = MKDEV(MAJOR(dev->devt),
1228 MINOR(dev->devt) + partno);
1229 break;
1230 }
Tejun Heo548b10e2008-08-29 09:01:47 +02001231 part = disk_get_part(disk, partno);
Tejun Heo2bbedcb2008-08-29 11:41:51 +02001232 if (part) {
Tejun Heof331c022008-09-03 09:01:48 +02001233 devt = part_devt(part);
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001234 disk_put_part(part);
Tejun Heo548b10e2008-08-29 09:01:47 +02001235 break;
Tejun Heodef4e382008-09-03 08:57:12 +02001236 }
Tejun Heo548b10e2008-08-29 09:01:47 +02001237 disk_put_part(part);
Kay Sievers5c0ef6d2008-08-16 14:30:30 +02001238 }
Tejun Heodef4e382008-09-03 08:57:12 +02001239 class_dev_iter_exit(&iter);
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001240 return devt;
1241}
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001242EXPORT_SYMBOL(blk_lookup_devt);
1243
Linus Torvalds1da177e2005-04-16 15:20:36 -07001244struct gendisk *alloc_disk(int minors)
1245{
Christoph Lameter19460892005-06-23 00:08:19 -07001246 return alloc_disk_node(minors, -1);
1247}
Tejun Heo689d6fa2008-08-25 19:56:16 +09001248EXPORT_SYMBOL(alloc_disk);
Christoph Lameter19460892005-06-23 00:08:19 -07001249
1250struct gendisk *alloc_disk_node(int minors, int node_id)
1251{
1252 struct gendisk *disk;
1253
Christoph Lameter94f60302007-07-17 04:03:29 -07001254 disk = kmalloc_node(sizeof(struct gendisk),
1255 GFP_KERNEL | __GFP_ZERO, node_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001256 if (disk) {
Tejun Heo074a7ac2008-08-25 19:56:14 +09001257 if (!init_part_stats(&disk->part0)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001258 kfree(disk);
1259 return NULL;
1260 }
Cheng Renquanbf91db12008-11-20 08:37:37 +01001261 disk->node_id = node_id;
Tejun Heo540eed52008-08-25 19:56:15 +09001262 if (disk_expand_part_tbl(disk, 0)) {
1263 free_part_stats(&disk->part0);
Tejun Heob5d0b9d2008-09-03 09:06:42 +02001264 kfree(disk);
1265 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001266 }
Tejun Heo540eed52008-08-25 19:56:15 +09001267 disk->part_tbl->part[0] = &disk->part0;
Jens Axboe6c23a962011-01-07 08:43:37 +01001268
1269 hd_ref_init(&disk->part0);
Tejun Heob5d0b9d2008-09-03 09:06:42 +02001270
Linus Torvalds1da177e2005-04-16 15:20:36 -07001271 disk->minors = minors;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001272 rand_initialize_disk(disk);
Tejun Heoed9e1982008-08-25 19:56:05 +09001273 disk_to_dev(disk)->class = &block_class;
1274 disk_to_dev(disk)->type = &disk_type;
1275 device_initialize(disk_to_dev(disk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276 }
1277 return disk;
1278}
Christoph Lameter19460892005-06-23 00:08:19 -07001279EXPORT_SYMBOL(alloc_disk_node);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001280
1281struct kobject *get_disk(struct gendisk *disk)
1282{
1283 struct module *owner;
1284 struct kobject *kobj;
1285
1286 if (!disk->fops)
1287 return NULL;
1288 owner = disk->fops->owner;
1289 if (owner && !try_module_get(owner))
1290 return NULL;
Tejun Heoed9e1982008-08-25 19:56:05 +09001291 kobj = kobject_get(&disk_to_dev(disk)->kobj);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292 if (kobj == NULL) {
1293 module_put(owner);
1294 return NULL;
1295 }
1296 return kobj;
1297
1298}
1299
1300EXPORT_SYMBOL(get_disk);
1301
1302void put_disk(struct gendisk *disk)
1303{
1304 if (disk)
Tejun Heoed9e1982008-08-25 19:56:05 +09001305 kobject_put(&disk_to_dev(disk)->kobj);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001306}
1307
1308EXPORT_SYMBOL(put_disk);
1309
Hannes Reineckee3264a42009-07-28 09:13:13 +02001310static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1311{
1312 char event[] = "DISK_RO=1";
1313 char *envp[] = { event, NULL };
1314
1315 if (!ro)
1316 event[8] = '0';
1317 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1318}
1319
Linus Torvalds1da177e2005-04-16 15:20:36 -07001320void set_device_ro(struct block_device *bdev, int flag)
1321{
Tejun Heob7db9952008-08-25 19:56:10 +09001322 bdev->bd_part->policy = flag;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323}
1324
1325EXPORT_SYMBOL(set_device_ro);
1326
1327void set_disk_ro(struct gendisk *disk, int flag)
1328{
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001329 struct disk_part_iter piter;
1330 struct hd_struct *part;
1331
Hannes Reineckee3264a42009-07-28 09:13:13 +02001332 if (disk->part0.policy != flag) {
1333 set_disk_ro_uevent(disk, flag);
1334 disk->part0.policy = flag;
1335 }
1336
1337 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY);
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001338 while ((part = disk_part_iter_next(&piter)))
1339 part->policy = flag;
1340 disk_part_iter_exit(&piter);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001341}
1342
1343EXPORT_SYMBOL(set_disk_ro);
1344
1345int bdev_read_only(struct block_device *bdev)
1346{
1347 if (!bdev)
1348 return 0;
Tejun Heob7db9952008-08-25 19:56:10 +09001349 return bdev->bd_part->policy;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001350}
1351
1352EXPORT_SYMBOL(bdev_read_only);
1353
Tejun Heocf771cb2008-09-03 09:01:09 +02001354int invalidate_partition(struct gendisk *disk, int partno)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001355{
1356 int res = 0;
Tejun Heocf771cb2008-09-03 09:01:09 +02001357 struct block_device *bdev = bdget_disk(disk, partno);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358 if (bdev) {
Christoph Hellwig2ef41632005-05-05 16:15:59 -07001359 fsync_bdev(bdev);
NeilBrown93b270f2011-02-24 17:25:47 +11001360 res = __invalidate_device(bdev, true);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001361 bdput(bdev);
1362 }
1363 return res;
1364}
1365
1366EXPORT_SYMBOL(invalidate_partition);
Tejun Heo77ea8872010-12-08 20:57:37 +01001367
1368/*
1369 * Disk events - monitor disk events like media change and eject request.
1370 */
1371struct disk_events {
1372 struct list_head node; /* all disk_event's */
1373 struct gendisk *disk; /* the associated disk */
1374 spinlock_t lock;
1375
Tejun Heofdd514e2011-06-09 20:43:59 +02001376 struct mutex block_mutex; /* protects blocking */
Tejun Heo77ea8872010-12-08 20:57:37 +01001377 int block; /* event blocking depth */
1378 unsigned int pending; /* events already sent out */
1379 unsigned int clearing; /* events being cleared */
1380
1381 long poll_msecs; /* interval, -1 for default */
1382 struct delayed_work dwork;
1383};
1384
1385static const char *disk_events_strs[] = {
1386 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "media_change",
1387 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "eject_request",
1388};
1389
1390static char *disk_uevents[] = {
1391 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "DISK_MEDIA_CHANGE=1",
1392 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "DISK_EJECT_REQUEST=1",
1393};
1394
1395/* list of all disk_events */
1396static DEFINE_MUTEX(disk_events_mutex);
1397static LIST_HEAD(disk_events);
1398
1399/* disable in-kernel polling by default */
1400static unsigned long disk_events_dfl_poll_msecs = 0;
1401
1402static unsigned long disk_events_poll_jiffies(struct gendisk *disk)
1403{
1404 struct disk_events *ev = disk->ev;
1405 long intv_msecs = 0;
1406
1407 /*
1408 * If device-specific poll interval is set, always use it. If
1409 * the default is being used, poll iff there are events which
1410 * can't be monitored asynchronously.
1411 */
1412 if (ev->poll_msecs >= 0)
1413 intv_msecs = ev->poll_msecs;
1414 else if (disk->events & ~disk->async_events)
1415 intv_msecs = disk_events_dfl_poll_msecs;
1416
1417 return msecs_to_jiffies(intv_msecs);
1418}
1419
Tejun Heoc3af54a2011-06-09 20:43:55 +02001420/**
1421 * disk_block_events - block and flush disk event checking
1422 * @disk: disk to block events for
1423 *
1424 * On return from this function, it is guaranteed that event checking
1425 * isn't in progress and won't happen until unblocked by
1426 * disk_unblock_events(). Events blocking is counted and the actual
1427 * unblocking happens after the matching number of unblocks are done.
1428 *
1429 * Note that this intentionally does not block event checking from
1430 * disk_clear_events().
1431 *
1432 * CONTEXT:
1433 * Might sleep.
1434 */
1435void disk_block_events(struct gendisk *disk)
Tejun Heo77ea8872010-12-08 20:57:37 +01001436{
1437 struct disk_events *ev = disk->ev;
1438 unsigned long flags;
1439 bool cancel;
1440
Tejun Heoc3af54a2011-06-09 20:43:55 +02001441 if (!ev)
1442 return;
1443
Tejun Heofdd514e2011-06-09 20:43:59 +02001444 /*
1445 * Outer mutex ensures that the first blocker completes canceling
1446 * the event work before further blockers are allowed to finish.
1447 */
1448 mutex_lock(&ev->block_mutex);
1449
Tejun Heo77ea8872010-12-08 20:57:37 +01001450 spin_lock_irqsave(&ev->lock, flags);
1451 cancel = !ev->block++;
1452 spin_unlock_irqrestore(&ev->lock, flags);
1453
Tejun Heoc3af54a2011-06-09 20:43:55 +02001454 if (cancel)
1455 cancel_delayed_work_sync(&disk->ev->dwork);
Tejun Heofdd514e2011-06-09 20:43:59 +02001456
1457 mutex_unlock(&ev->block_mutex);
Tejun Heo77ea8872010-12-08 20:57:37 +01001458}
1459
1460static void __disk_unblock_events(struct gendisk *disk, bool check_now)
1461{
1462 struct disk_events *ev = disk->ev;
1463 unsigned long intv;
1464 unsigned long flags;
1465
1466 spin_lock_irqsave(&ev->lock, flags);
1467
1468 if (WARN_ON_ONCE(ev->block <= 0))
1469 goto out_unlock;
1470
1471 if (--ev->block)
1472 goto out_unlock;
1473
1474 /*
1475 * Not exactly a latency critical operation, set poll timer
1476 * slack to 25% and kick event check.
1477 */
1478 intv = disk_events_poll_jiffies(disk);
1479 set_timer_slack(&ev->dwork.timer, intv / 4);
1480 if (check_now)
1481 queue_delayed_work(system_nrt_wq, &ev->dwork, 0);
1482 else if (intv)
1483 queue_delayed_work(system_nrt_wq, &ev->dwork, intv);
1484out_unlock:
1485 spin_unlock_irqrestore(&ev->lock, flags);
1486}
1487
1488/**
Tejun Heo77ea8872010-12-08 20:57:37 +01001489 * disk_unblock_events - unblock disk event checking
1490 * @disk: disk to unblock events for
1491 *
1492 * Undo disk_block_events(). When the block count reaches zero, it
1493 * starts events polling if configured.
1494 *
1495 * CONTEXT:
1496 * Don't care. Safe to call from irq context.
1497 */
1498void disk_unblock_events(struct gendisk *disk)
1499{
1500 if (disk->ev)
Tejun Heofacc31d2011-03-09 19:54:27 +01001501 __disk_unblock_events(disk, false);
Tejun Heo77ea8872010-12-08 20:57:37 +01001502}
1503
1504/**
Tejun Heo85ef06d2011-07-01 16:17:47 +02001505 * disk_flush_events - schedule immediate event checking and flushing
1506 * @disk: disk to check and flush events for
1507 * @mask: events to flush
Tejun Heo77ea8872010-12-08 20:57:37 +01001508 *
Tejun Heo85ef06d2011-07-01 16:17:47 +02001509 * Schedule immediate event checking on @disk if not blocked. Events in
1510 * @mask are scheduled to be cleared from the driver. Note that this
1511 * doesn't clear the events from @disk->ev.
Tejun Heo77ea8872010-12-08 20:57:37 +01001512 *
1513 * CONTEXT:
Tejun Heo85ef06d2011-07-01 16:17:47 +02001514 * If @mask is non-zero must be called with bdev->bd_mutex held.
Tejun Heo77ea8872010-12-08 20:57:37 +01001515 */
Tejun Heo85ef06d2011-07-01 16:17:47 +02001516void disk_flush_events(struct gendisk *disk, unsigned int mask)
Tejun Heo77ea8872010-12-08 20:57:37 +01001517{
Tejun Heoa9dce2a2011-06-09 20:43:54 +02001518 struct disk_events *ev = disk->ev;
Tejun Heoa9dce2a2011-06-09 20:43:54 +02001519
1520 if (!ev)
1521 return;
1522
Tejun Heo85ef06d2011-07-01 16:17:47 +02001523 spin_lock_irq(&ev->lock);
1524 ev->clearing |= mask;
Tejun Heoa9dce2a2011-06-09 20:43:54 +02001525 if (!ev->block) {
1526 cancel_delayed_work(&ev->dwork);
1527 queue_delayed_work(system_nrt_wq, &ev->dwork, 0);
Tejun Heo77ea8872010-12-08 20:57:37 +01001528 }
Tejun Heo85ef06d2011-07-01 16:17:47 +02001529 spin_unlock_irq(&ev->lock);
Tejun Heo77ea8872010-12-08 20:57:37 +01001530}
Tejun Heo77ea8872010-12-08 20:57:37 +01001531
1532/**
1533 * disk_clear_events - synchronously check, clear and return pending events
1534 * @disk: disk to fetch and clear events from
1535 * @mask: mask of events to be fetched and clearted
1536 *
1537 * Disk events are synchronously checked and pending events in @mask
1538 * are cleared and returned. This ignores the block count.
1539 *
1540 * CONTEXT:
1541 * Might sleep.
1542 */
1543unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask)
1544{
1545 const struct block_device_operations *bdops = disk->fops;
1546 struct disk_events *ev = disk->ev;
1547 unsigned int pending;
1548
1549 if (!ev) {
1550 /* for drivers still using the old ->media_changed method */
1551 if ((mask & DISK_EVENT_MEDIA_CHANGE) &&
1552 bdops->media_changed && bdops->media_changed(disk))
1553 return DISK_EVENT_MEDIA_CHANGE;
1554 return 0;
1555 }
1556
1557 /* tell the workfn about the events being cleared */
1558 spin_lock_irq(&ev->lock);
1559 ev->clearing |= mask;
1560 spin_unlock_irq(&ev->lock);
1561
1562 /* uncondtionally schedule event check and wait for it to finish */
Tejun Heoc3af54a2011-06-09 20:43:55 +02001563 disk_block_events(disk);
Tejun Heo77ea8872010-12-08 20:57:37 +01001564 queue_delayed_work(system_nrt_wq, &ev->dwork, 0);
1565 flush_delayed_work(&ev->dwork);
1566 __disk_unblock_events(disk, false);
1567
1568 /* then, fetch and clear pending events */
1569 spin_lock_irq(&ev->lock);
1570 WARN_ON_ONCE(ev->clearing & mask); /* cleared by workfn */
1571 pending = ev->pending & mask;
1572 ev->pending &= ~mask;
1573 spin_unlock_irq(&ev->lock);
1574
1575 return pending;
1576}
1577
1578static void disk_events_workfn(struct work_struct *work)
1579{
1580 struct delayed_work *dwork = to_delayed_work(work);
1581 struct disk_events *ev = container_of(dwork, struct disk_events, dwork);
1582 struct gendisk *disk = ev->disk;
1583 char *envp[ARRAY_SIZE(disk_uevents) + 1] = { };
1584 unsigned int clearing = ev->clearing;
1585 unsigned int events;
1586 unsigned long intv;
1587 int nr_events = 0, i;
1588
1589 /* check events */
1590 events = disk->fops->check_events(disk, clearing);
1591
1592 /* accumulate pending events and schedule next poll if necessary */
1593 spin_lock_irq(&ev->lock);
1594
1595 events &= ~ev->pending;
1596 ev->pending |= events;
1597 ev->clearing &= ~clearing;
1598
1599 intv = disk_events_poll_jiffies(disk);
1600 if (!ev->block && intv)
1601 queue_delayed_work(system_nrt_wq, &ev->dwork, intv);
1602
1603 spin_unlock_irq(&ev->lock);
1604
Tejun Heo7c88a162011-04-21 19:43:58 +02001605 /*
1606 * Tell userland about new events. Only the events listed in
1607 * @disk->events are reported. Unlisted events are processed the
1608 * same internally but never get reported to userland.
1609 */
Tejun Heo77ea8872010-12-08 20:57:37 +01001610 for (i = 0; i < ARRAY_SIZE(disk_uevents); i++)
Tejun Heo7c88a162011-04-21 19:43:58 +02001611 if (events & disk->events & (1 << i))
Tejun Heo77ea8872010-12-08 20:57:37 +01001612 envp[nr_events++] = disk_uevents[i];
1613
1614 if (nr_events)
1615 kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp);
1616}
1617
1618/*
1619 * A disk events enabled device has the following sysfs nodes under
1620 * its /sys/block/X/ directory.
1621 *
1622 * events : list of all supported events
1623 * events_async : list of events which can be detected w/o polling
1624 * events_poll_msecs : polling interval, 0: disable, -1: system default
1625 */
1626static ssize_t __disk_events_show(unsigned int events, char *buf)
1627{
1628 const char *delim = "";
1629 ssize_t pos = 0;
1630 int i;
1631
1632 for (i = 0; i < ARRAY_SIZE(disk_events_strs); i++)
1633 if (events & (1 << i)) {
1634 pos += sprintf(buf + pos, "%s%s",
1635 delim, disk_events_strs[i]);
1636 delim = " ";
1637 }
1638 if (pos)
1639 pos += sprintf(buf + pos, "\n");
1640 return pos;
1641}
1642
1643static ssize_t disk_events_show(struct device *dev,
1644 struct device_attribute *attr, char *buf)
1645{
1646 struct gendisk *disk = dev_to_disk(dev);
1647
1648 return __disk_events_show(disk->events, buf);
1649}
1650
1651static ssize_t disk_events_async_show(struct device *dev,
1652 struct device_attribute *attr, char *buf)
1653{
1654 struct gendisk *disk = dev_to_disk(dev);
1655
1656 return __disk_events_show(disk->async_events, buf);
1657}
1658
1659static ssize_t disk_events_poll_msecs_show(struct device *dev,
1660 struct device_attribute *attr,
1661 char *buf)
1662{
1663 struct gendisk *disk = dev_to_disk(dev);
1664
1665 return sprintf(buf, "%ld\n", disk->ev->poll_msecs);
1666}
1667
1668static ssize_t disk_events_poll_msecs_store(struct device *dev,
1669 struct device_attribute *attr,
1670 const char *buf, size_t count)
1671{
1672 struct gendisk *disk = dev_to_disk(dev);
1673 long intv;
1674
1675 if (!count || !sscanf(buf, "%ld", &intv))
1676 return -EINVAL;
1677
1678 if (intv < 0 && intv != -1)
1679 return -EINVAL;
1680
Tejun Heoc3af54a2011-06-09 20:43:55 +02001681 disk_block_events(disk);
Tejun Heo77ea8872010-12-08 20:57:37 +01001682 disk->ev->poll_msecs = intv;
1683 __disk_unblock_events(disk, true);
1684
1685 return count;
1686}
1687
1688static const DEVICE_ATTR(events, S_IRUGO, disk_events_show, NULL);
1689static const DEVICE_ATTR(events_async, S_IRUGO, disk_events_async_show, NULL);
1690static const DEVICE_ATTR(events_poll_msecs, S_IRUGO|S_IWUSR,
1691 disk_events_poll_msecs_show,
1692 disk_events_poll_msecs_store);
1693
1694static const struct attribute *disk_events_attrs[] = {
1695 &dev_attr_events.attr,
1696 &dev_attr_events_async.attr,
1697 &dev_attr_events_poll_msecs.attr,
1698 NULL,
1699};
1700
1701/*
1702 * The default polling interval can be specified by the kernel
1703 * parameter block.events_dfl_poll_msecs which defaults to 0
1704 * (disable). This can also be modified runtime by writing to
1705 * /sys/module/block/events_dfl_poll_msecs.
1706 */
1707static int disk_events_set_dfl_poll_msecs(const char *val,
1708 const struct kernel_param *kp)
1709{
1710 struct disk_events *ev;
1711 int ret;
1712
1713 ret = param_set_ulong(val, kp);
1714 if (ret < 0)
1715 return ret;
1716
1717 mutex_lock(&disk_events_mutex);
1718
1719 list_for_each_entry(ev, &disk_events, node)
Tejun Heo85ef06d2011-07-01 16:17:47 +02001720 disk_flush_events(ev->disk, 0);
Tejun Heo77ea8872010-12-08 20:57:37 +01001721
1722 mutex_unlock(&disk_events_mutex);
1723
1724 return 0;
1725}
1726
1727static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops = {
1728 .set = disk_events_set_dfl_poll_msecs,
1729 .get = param_get_ulong,
1730};
1731
1732#undef MODULE_PARAM_PREFIX
1733#define MODULE_PARAM_PREFIX "block."
1734
1735module_param_cb(events_dfl_poll_msecs, &disk_events_dfl_poll_msecs_param_ops,
1736 &disk_events_dfl_poll_msecs, 0644);
1737
1738/*
Stanislaw Gruszka9f53d2fe2012-03-02 10:43:28 +01001739 * disk_{alloc|add|del|release}_events - initialize and destroy disk_events.
Tejun Heo77ea8872010-12-08 20:57:37 +01001740 */
Stanislaw Gruszka9f53d2fe2012-03-02 10:43:28 +01001741static void disk_alloc_events(struct gendisk *disk)
Tejun Heo77ea8872010-12-08 20:57:37 +01001742{
1743 struct disk_events *ev;
1744
Tejun Heo75e3f3e2011-05-26 21:06:50 +02001745 if (!disk->fops->check_events)
Tejun Heo77ea8872010-12-08 20:57:37 +01001746 return;
1747
1748 ev = kzalloc(sizeof(*ev), GFP_KERNEL);
1749 if (!ev) {
1750 pr_warn("%s: failed to initialize events\n", disk->disk_name);
1751 return;
1752 }
1753
Tejun Heo77ea8872010-12-08 20:57:37 +01001754 INIT_LIST_HEAD(&ev->node);
1755 ev->disk = disk;
1756 spin_lock_init(&ev->lock);
Tejun Heofdd514e2011-06-09 20:43:59 +02001757 mutex_init(&ev->block_mutex);
Tejun Heo77ea8872010-12-08 20:57:37 +01001758 ev->block = 1;
1759 ev->poll_msecs = -1;
1760 INIT_DELAYED_WORK(&ev->dwork, disk_events_workfn);
1761
Stanislaw Gruszka9f53d2fe2012-03-02 10:43:28 +01001762 disk->ev = ev;
1763}
1764
1765static void disk_add_events(struct gendisk *disk)
1766{
1767 if (!disk->ev)
1768 return;
1769
1770 /* FIXME: error handling */
1771 if (sysfs_create_files(&disk_to_dev(disk)->kobj, disk_events_attrs) < 0)
1772 pr_warn("%s: failed to create sysfs files for events\n",
1773 disk->disk_name);
1774
Tejun Heo77ea8872010-12-08 20:57:37 +01001775 mutex_lock(&disk_events_mutex);
Stanislaw Gruszka9f53d2fe2012-03-02 10:43:28 +01001776 list_add_tail(&disk->ev->node, &disk_events);
Tejun Heo77ea8872010-12-08 20:57:37 +01001777 mutex_unlock(&disk_events_mutex);
1778
1779 /*
1780 * Block count is initialized to 1 and the following initial
1781 * unblock kicks it into action.
1782 */
1783 __disk_unblock_events(disk, true);
1784}
1785
1786static void disk_del_events(struct gendisk *disk)
1787{
1788 if (!disk->ev)
1789 return;
1790
Tejun Heoc3af54a2011-06-09 20:43:55 +02001791 disk_block_events(disk);
Tejun Heo77ea8872010-12-08 20:57:37 +01001792
1793 mutex_lock(&disk_events_mutex);
1794 list_del_init(&disk->ev->node);
1795 mutex_unlock(&disk_events_mutex);
1796
1797 sysfs_remove_files(&disk_to_dev(disk)->kobj, disk_events_attrs);
1798}
1799
1800static void disk_release_events(struct gendisk *disk)
1801{
1802 /* the block count should be 1 from disk_del_events() */
1803 WARN_ON_ONCE(disk->ev && disk->ev->block != 1);
1804 kfree(disk->ev);
1805}