blob: 6cace663a80e4a89492f8b656f580a944b134f9d [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;
Vivek Goyalc83f6bf2012-08-01 12:24:18 +0200157 if (!part_nr_sects_read(part) &&
Tejun Heo71982a42009-04-17 08:34:48 +0200158 !(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 &&
Vivek Goyalc83f6bf2012-08-01 12:24:18 +0200194 sector < part->start_sect + part_nr_sects_read(part);
Jens Axboea6f23652008-10-24 12:52:42 +0200195}
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];
Tejun Heo05c69d22012-05-15 08:22:04 +0200746 char uuid_buf[PARTITION_META_INFO_UUIDLTH * 2 + 5];
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
Tejun Heo05c69d22012-05-15 08:22:04 +0200765 uuid_buf[0] = '\0';
Will Drewryb5af9212010-08-31 15:47:07 -0500766 if (part->info)
Tejun Heo05c69d22012-05-15 08:22:04 +0200767 snprintf(uuid_buf, sizeof(uuid_buf), "%pU",
768 part->info->uuid);
Will Drewryb5af9212010-08-31 15:47:07 -0500769
770 printk("%s%s %10llu %s %s", is_part0 ? "" : " ",
Tejun Heo1f014292008-08-25 19:47:23 +0900771 bdevt_str(part_devt(part), devt_buf),
Vivek Goyalc83f6bf2012-08-01 12:24:18 +0200772 (unsigned long long)part_nr_sects_read(part) >> 1
773 , disk_name(disk, part->partno, name_buf),
Tejun Heo05c69d22012-05-15 08:22:04 +0200774 uuid_buf);
Tejun Heo074a7ac2008-08-25 19:56:14 +0900775 if (is_part0) {
776 if (disk->driverfs_dev != NULL &&
777 disk->driverfs_dev->driver != NULL)
778 printk(" driver: %s\n",
779 disk->driverfs_dev->driver->name);
780 else
781 printk(" (driver?)\n");
782 } else
783 printk("\n");
784 }
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200785 disk_part_iter_exit(&piter);
Tejun Heodef4e382008-09-03 08:57:12 +0200786 }
787 class_dev_iter_exit(&iter);
Dave Gilbertdd2a3452007-05-09 02:33:24 -0700788}
789
Linus Torvalds1da177e2005-04-16 15:20:36 -0700790#ifdef CONFIG_PROC_FS
791/* iterator */
Tejun Heodef4e382008-09-03 08:57:12 +0200792static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
Greg Kroah-Hartman68c4d4a2008-05-22 17:21:08 -0400793{
Tejun Heodef4e382008-09-03 08:57:12 +0200794 loff_t skip = *pos;
795 struct class_dev_iter *iter;
796 struct device *dev;
Greg Kroah-Hartman68c4d4a2008-05-22 17:21:08 -0400797
Harvey Harrisonaeb3d3a2008-08-28 09:27:42 +0200798 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
Tejun Heodef4e382008-09-03 08:57:12 +0200799 if (!iter)
800 return ERR_PTR(-ENOMEM);
801
802 seqf->private = iter;
803 class_dev_iter_init(iter, &block_class, NULL, &disk_type);
804 do {
805 dev = class_dev_iter_next(iter);
806 if (!dev)
807 return NULL;
808 } while (skip--);
809
810 return dev_to_disk(dev);
Greg Kroah-Hartman68c4d4a2008-05-22 17:21:08 -0400811}
812
Tejun Heodef4e382008-09-03 08:57:12 +0200813static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200815 struct device *dev;
Greg Kroah-Hartman66c64af2008-05-22 17:21:08 -0400816
Tejun Heodef4e382008-09-03 08:57:12 +0200817 (*pos)++;
818 dev = class_dev_iter_next(seqf->private);
Tejun Heo2ac3cee2008-09-03 08:53:37 +0200819 if (dev)
Greg Kroah-Hartman68c4d4a2008-05-22 17:21:08 -0400820 return dev_to_disk(dev);
Tejun Heo2ac3cee2008-09-03 08:53:37 +0200821
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822 return NULL;
823}
824
Tejun Heodef4e382008-09-03 08:57:12 +0200825static void disk_seqf_stop(struct seq_file *seqf, void *v)
Greg Kroah-Hartman27f30252008-05-22 17:21:08 -0400826{
Tejun Heodef4e382008-09-03 08:57:12 +0200827 struct class_dev_iter *iter = seqf->private;
Greg Kroah-Hartman27f30252008-05-22 17:21:08 -0400828
Tejun Heodef4e382008-09-03 08:57:12 +0200829 /* stop is called even after start failed :-( */
830 if (iter) {
831 class_dev_iter_exit(iter);
832 kfree(iter);
Kay Sievers5c0ef6d2008-08-16 14:30:30 +0200833 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700834}
835
Tejun Heodef4e382008-09-03 08:57:12 +0200836static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700837{
Jianpeng Ma06768062012-08-03 10:42:00 +0200838 void *p;
Tejun Heodef4e382008-09-03 08:57:12 +0200839
840 p = disk_seqf_start(seqf, pos);
Yang Zhangb9f985b2010-12-17 08:58:36 +0100841 if (!IS_ERR_OR_NULL(p) && !*pos)
Tejun Heodef4e382008-09-03 08:57:12 +0200842 seq_puts(seqf, "major minor #blocks name\n\n");
843 return p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700844}
845
Tejun Heocf771cb2008-09-03 09:01:09 +0200846static int show_partition(struct seq_file *seqf, void *v)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847{
848 struct gendisk *sgp = v;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200849 struct disk_part_iter piter;
850 struct hd_struct *part;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700851 char buf[BDEVNAME_SIZE];
852
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853 /* Don't show non-partitionable removeable devices or empty devices */
Tejun Heod27769e2011-08-23 20:01:04 +0200854 if (!get_capacity(sgp) || (!disk_max_parts(sgp) &&
Tejun Heof331c022008-09-03 09:01:48 +0200855 (sgp->flags & GENHD_FL_REMOVABLE)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856 return 0;
857 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
858 return 0;
859
860 /* show the full disk and all non-0 size partitions of it */
Tejun Heo074a7ac2008-08-25 19:56:14 +0900861 disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0);
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200862 while ((part = disk_part_iter_next(&piter)))
Tejun Heo1f014292008-08-25 19:47:23 +0900863 seq_printf(seqf, "%4d %7d %10llu %s\n",
Tejun Heof331c022008-09-03 09:01:48 +0200864 MAJOR(part_devt(part)), MINOR(part_devt(part)),
Vivek Goyalc83f6bf2012-08-01 12:24:18 +0200865 (unsigned long long)part_nr_sects_read(part) >> 1,
Tejun Heof331c022008-09-03 09:01:48 +0200866 disk_name(sgp, part->partno, buf));
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200867 disk_part_iter_exit(&piter);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868
869 return 0;
870}
871
Alexey Dobriyanf5009752008-10-04 23:53:21 +0400872static const struct seq_operations partitions_op = {
Tejun Heodef4e382008-09-03 08:57:12 +0200873 .start = show_partition_start,
874 .next = disk_seqf_next,
875 .stop = disk_seqf_stop,
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200876 .show = show_partition
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877};
Alexey Dobriyanf5009752008-10-04 23:53:21 +0400878
879static int partitions_open(struct inode *inode, struct file *file)
880{
881 return seq_open(file, &partitions_op);
882}
883
884static const struct file_operations proc_partitions_operations = {
885 .open = partitions_open,
886 .read = seq_read,
887 .llseek = seq_lseek,
888 .release = seq_release,
889};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890#endif
891
892
Tejun Heocf771cb2008-09-03 09:01:09 +0200893static struct kobject *base_probe(dev_t devt, int *partno, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200895 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896 /* Make old-style 2.4 aliases work */
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200897 request_module("block-major-%d", MAJOR(devt));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898 return NULL;
899}
900
901static int __init genhd_device_init(void)
902{
Dan Williamse105b8b2008-04-21 10:51:07 -0700903 int error;
904
905 block_class.dev_kobj = sysfs_dev_block_kobj;
906 error = class_register(&block_class);
Roland McGrathee27a552008-03-11 17:13:15 -0700907 if (unlikely(error))
908 return error;
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200909 bdev_map = kobj_map_init(base_probe, &block_class_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700910 blk_dev_init();
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200911
Zhang, Yanmin561ec682008-11-14 08:26:30 +0100912 register_blkdev(BLOCK_EXT_MAJOR, "blkext");
913
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200914 /* create top-level block dir */
Andi Kleene52eec12010-09-08 16:54:17 +0200915 if (!sysfs_deprecated)
916 block_depr = kobject_create_and_add("block", NULL);
Greg Kroah-Hartman830d3cf2007-11-06 10:36:58 -0800917 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700918}
919
920subsys_initcall(genhd_device_init);
921
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200922static ssize_t disk_range_show(struct device *dev,
923 struct device_attribute *attr, char *buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200925 struct gendisk *disk = dev_to_disk(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200927 return sprintf(buf, "%d\n", disk->minors);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700928}
929
Tejun Heo1f014292008-08-25 19:47:23 +0900930static ssize_t disk_ext_range_show(struct device *dev,
931 struct device_attribute *attr, char *buf)
932{
933 struct gendisk *disk = dev_to_disk(dev);
934
Tejun Heob5d0b9d2008-09-03 09:06:42 +0200935 return sprintf(buf, "%d\n", disk_max_parts(disk));
Tejun Heo1f014292008-08-25 19:47:23 +0900936}
937
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200938static ssize_t disk_removable_show(struct device *dev,
939 struct device_attribute *attr, char *buf)
Kay Sieversa7fd6702005-10-01 14:49:43 +0200940{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200941 struct gendisk *disk = dev_to_disk(dev);
Kay Sieversa7fd6702005-10-01 14:49:43 +0200942
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200943 return sprintf(buf, "%d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700944 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200945}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700946
Kay Sievers1c9ce522008-06-13 09:41:00 +0200947static ssize_t disk_ro_show(struct device *dev,
948 struct device_attribute *attr, char *buf)
949{
950 struct gendisk *disk = dev_to_disk(dev);
951
Tejun Heob7db9952008-08-25 19:56:10 +0900952 return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
Kay Sievers1c9ce522008-06-13 09:41:00 +0200953}
954
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200955static ssize_t disk_capability_show(struct device *dev,
956 struct device_attribute *attr, char *buf)
Kristen Carlson Accardi86ce18d2007-05-23 13:57:38 -0700957{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200958 struct gendisk *disk = dev_to_disk(dev);
959
960 return sprintf(buf, "%x\n", disk->flags);
Kristen Carlson Accardi86ce18d2007-05-23 13:57:38 -0700961}
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200962
Martin K. Petersenc72758f2009-05-22 17:17:53 -0400963static ssize_t disk_alignment_offset_show(struct device *dev,
964 struct device_attribute *attr,
965 char *buf)
966{
967 struct gendisk *disk = dev_to_disk(dev);
968
969 return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue));
970}
971
Martin K. Petersen86b37282009-11-10 11:50:21 +0100972static ssize_t disk_discard_alignment_show(struct device *dev,
973 struct device_attribute *attr,
974 char *buf)
975{
976 struct gendisk *disk = dev_to_disk(dev);
977
Martin K. Petersendd3d1452010-01-11 03:21:48 -0500978 return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue));
Martin K. Petersen86b37282009-11-10 11:50:21 +0100979}
980
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200981static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL);
Tejun Heo1f014292008-08-25 19:47:23 +0900982static DEVICE_ATTR(ext_range, S_IRUGO, disk_ext_range_show, NULL);
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200983static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL);
Kay Sievers1c9ce522008-06-13 09:41:00 +0200984static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL);
Tejun Heoe5610522008-08-25 19:56:09 +0900985static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL);
Martin K. Petersenc72758f2009-05-22 17:17:53 -0400986static DEVICE_ATTR(alignment_offset, S_IRUGO, disk_alignment_offset_show, NULL);
Martin K. Petersen86b37282009-11-10 11:50:21 +0100987static DEVICE_ATTR(discard_alignment, S_IRUGO, disk_discard_alignment_show,
988 NULL);
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200989static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL);
Tejun Heo074a7ac2008-08-25 19:56:14 +0900990static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL);
Nikanth Karthikesan316d3152009-10-06 20:16:55 +0200991static DEVICE_ATTR(inflight, S_IRUGO, part_inflight_show, NULL);
Akinobu Mitac17bb492006-12-08 02:39:46 -0800992#ifdef CONFIG_FAIL_MAKE_REQUEST
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200993static struct device_attribute dev_attr_fail =
Tejun Heoeddb2e22008-08-25 19:56:13 +0900994 __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store);
Akinobu Mitac17bb492006-12-08 02:39:46 -0800995#endif
Jens Axboe581d4e22008-09-14 05:56:33 -0700996#ifdef CONFIG_FAIL_IO_TIMEOUT
997static struct device_attribute dev_attr_fail_timeout =
998 __ATTR(io-timeout-fail, S_IRUGO|S_IWUSR, part_timeout_show,
999 part_timeout_store);
1000#endif
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001001
1002static struct attribute *disk_attrs[] = {
1003 &dev_attr_range.attr,
Tejun Heo1f014292008-08-25 19:47:23 +09001004 &dev_attr_ext_range.attr,
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001005 &dev_attr_removable.attr,
Kay Sievers1c9ce522008-06-13 09:41:00 +02001006 &dev_attr_ro.attr,
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001007 &dev_attr_size.attr,
Martin K. Petersenc72758f2009-05-22 17:17:53 -04001008 &dev_attr_alignment_offset.attr,
Martin K. Petersen86b37282009-11-10 11:50:21 +01001009 &dev_attr_discard_alignment.attr,
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001010 &dev_attr_capability.attr,
1011 &dev_attr_stat.attr,
Nikanth Karthikesan316d3152009-10-06 20:16:55 +02001012 &dev_attr_inflight.attr,
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001013#ifdef CONFIG_FAIL_MAKE_REQUEST
1014 &dev_attr_fail.attr,
1015#endif
Jens Axboe581d4e22008-09-14 05:56:33 -07001016#ifdef CONFIG_FAIL_IO_TIMEOUT
1017 &dev_attr_fail_timeout.attr,
1018#endif
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001019 NULL
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020};
1021
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001022static struct attribute_group disk_attr_group = {
1023 .attrs = disk_attrs,
1024};
1025
David Brownella4dbd672009-06-24 10:06:31 -07001026static const struct attribute_group *disk_attr_groups[] = {
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001027 &disk_attr_group,
1028 NULL
1029};
1030
Tejun Heo540eed52008-08-25 19:56:15 +09001031/**
1032 * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way
1033 * @disk: disk to replace part_tbl for
1034 * @new_ptbl: new part_tbl to install
1035 *
1036 * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The
1037 * original ptbl is freed using RCU callback.
1038 *
1039 * LOCKING:
1040 * Matching bd_mutx locked.
1041 */
1042static void disk_replace_part_tbl(struct gendisk *disk,
1043 struct disk_part_tbl *new_ptbl)
1044{
1045 struct disk_part_tbl *old_ptbl = disk->part_tbl;
1046
1047 rcu_assign_pointer(disk->part_tbl, new_ptbl);
Jens Axboea6f23652008-10-24 12:52:42 +02001048
1049 if (old_ptbl) {
1050 rcu_assign_pointer(old_ptbl->last_lookup, NULL);
Lai Jiangshan57bdfbf2011-03-18 11:42:58 +08001051 kfree_rcu(old_ptbl, rcu_head);
Jens Axboea6f23652008-10-24 12:52:42 +02001052 }
Tejun Heo540eed52008-08-25 19:56:15 +09001053}
1054
1055/**
1056 * disk_expand_part_tbl - expand disk->part_tbl
1057 * @disk: disk to expand part_tbl for
1058 * @partno: expand such that this partno can fit in
1059 *
1060 * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl
1061 * uses RCU to allow unlocked dereferencing for stats and other stuff.
1062 *
1063 * LOCKING:
1064 * Matching bd_mutex locked, might sleep.
1065 *
1066 * RETURNS:
1067 * 0 on success, -errno on failure.
1068 */
1069int disk_expand_part_tbl(struct gendisk *disk, int partno)
1070{
1071 struct disk_part_tbl *old_ptbl = disk->part_tbl;
1072 struct disk_part_tbl *new_ptbl;
1073 int len = old_ptbl ? old_ptbl->len : 0;
1074 int target = partno + 1;
1075 size_t size;
1076 int i;
1077
1078 /* disk_max_parts() is zero during initialization, ignore if so */
1079 if (disk_max_parts(disk) && target > disk_max_parts(disk))
1080 return -EINVAL;
1081
1082 if (target <= len)
1083 return 0;
1084
1085 size = sizeof(*new_ptbl) + target * sizeof(new_ptbl->part[0]);
1086 new_ptbl = kzalloc_node(size, GFP_KERNEL, disk->node_id);
1087 if (!new_ptbl)
1088 return -ENOMEM;
1089
Tejun Heo540eed52008-08-25 19:56:15 +09001090 new_ptbl->len = target;
1091
1092 for (i = 0; i < len; i++)
1093 rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]);
1094
1095 disk_replace_part_tbl(disk, new_ptbl);
1096 return 0;
1097}
1098
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001099static void disk_release(struct device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100{
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001101 struct gendisk *disk = dev_to_disk(dev);
1102
Tejun Heo77ea8872010-12-08 20:57:37 +01001103 disk_release_events(disk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001104 kfree(disk->random);
Tejun Heo540eed52008-08-25 19:56:15 +09001105 disk_replace_part_tbl(disk, NULL);
Tejun Heo074a7ac2008-08-25 19:56:14 +09001106 free_part_stats(&disk->part0);
Will Drewry6d1d8052010-08-31 15:47:05 -05001107 free_part_info(&disk->part0);
Tejun Heo523e1d32011-10-19 14:31:07 +02001108 if (disk->queue)
1109 blk_put_queue(disk->queue);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110 kfree(disk);
1111}
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001112struct class block_class = {
1113 .name = "block",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001114};
1115
Al Viro2c9ede52011-07-23 20:24:48 -04001116static char *block_devnode(struct device *dev, umode_t *mode)
Kay Sieversb03f38b2009-04-30 15:23:42 +02001117{
1118 struct gendisk *disk = dev_to_disk(dev);
1119
Kay Sieverse454cea2009-09-18 23:01:12 +02001120 if (disk->devnode)
1121 return disk->devnode(disk, mode);
Kay Sieversb03f38b2009-04-30 15:23:42 +02001122 return NULL;
1123}
1124
Adrian Bunk1826ead2008-03-04 11:23:46 +01001125static struct device_type disk_type = {
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001126 .name = "disk",
1127 .groups = disk_attr_groups,
1128 .release = disk_release,
Kay Sieverse454cea2009-09-18 23:01:12 +02001129 .devnode = block_devnode,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001130};
1131
Randy Dunlapa6e2ba82008-05-23 09:44:11 -07001132#ifdef CONFIG_PROC_FS
Tejun Heocf771cb2008-09-03 09:01:09 +02001133/*
1134 * aggregate disk stat collector. Uses the same stats that the sysfs
1135 * entries do, above, but makes them available through one seq_file.
1136 *
1137 * The output looks suspiciously like /proc/partitions with a bunch of
1138 * extra fields.
1139 */
1140static int diskstats_show(struct seq_file *seqf, void *v)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001141{
1142 struct gendisk *gp = v;
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001143 struct disk_part_iter piter;
1144 struct hd_struct *hd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001145 char buf[BDEVNAME_SIZE];
Tejun Heoc9959052008-08-25 19:47:21 +09001146 int cpu;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001147
1148 /*
Tejun Heoed9e1982008-08-25 19:56:05 +09001149 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
Tejun Heocf771cb2008-09-03 09:01:09 +02001150 seq_puts(seqf, "major minor name"
Linus Torvalds1da177e2005-04-16 15:20:36 -07001151 " rio rmerge rsect ruse wio wmerge "
1152 "wsect wuse running use aveq"
1153 "\n\n");
1154 */
Wanlong Gao9f5e4862011-06-13 10:45:43 +02001155
Tejun Heo71982a42009-04-17 08:34:48 +02001156 disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0);
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001157 while ((hd = disk_part_iter_next(&piter))) {
Tejun Heo074a7ac2008-08-25 19:56:14 +09001158 cpu = part_stat_lock();
Tejun Heoc9959052008-08-25 19:47:21 +09001159 part_round_stats(cpu, hd);
Tejun Heo074a7ac2008-08-25 19:56:14 +09001160 part_stat_unlock();
Herbert Poetzlf95fe9c2011-08-02 12:43:50 +02001161 seq_printf(seqf, "%4d %7d %s %lu %lu %lu "
1162 "%u %lu %lu %lu %u %u %u %u\n",
Tejun Heof331c022008-09-03 09:01:48 +02001163 MAJOR(part_devt(hd)), MINOR(part_devt(hd)),
1164 disk_name(gp, hd->partno, buf),
Liu Yuan53f22952011-03-02 11:00:15 -05001165 part_stat_read(hd, ios[READ]),
1166 part_stat_read(hd, merges[READ]),
Herbert Poetzlf95fe9c2011-08-02 12:43:50 +02001167 part_stat_read(hd, sectors[READ]),
Liu Yuan53f22952011-03-02 11:00:15 -05001168 jiffies_to_msecs(part_stat_read(hd, ticks[READ])),
1169 part_stat_read(hd, ios[WRITE]),
1170 part_stat_read(hd, merges[WRITE]),
Herbert Poetzlf95fe9c2011-08-02 12:43:50 +02001171 part_stat_read(hd, sectors[WRITE]),
Liu Yuan53f22952011-03-02 11:00:15 -05001172 jiffies_to_msecs(part_stat_read(hd, ticks[WRITE])),
Nikanth Karthikesan316d3152009-10-06 20:16:55 +02001173 part_in_flight(hd),
Jerome Marchand28f39d52008-02-08 11:04:56 +01001174 jiffies_to_msecs(part_stat_read(hd, io_ticks)),
1175 jiffies_to_msecs(part_stat_read(hd, time_in_queue))
1176 );
Linus Torvalds1da177e2005-04-16 15:20:36 -07001177 }
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001178 disk_part_iter_exit(&piter);
Wanlong Gao9f5e4862011-06-13 10:45:43 +02001179
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180 return 0;
1181}
1182
Alexey Dobriyan31d85ab2008-10-06 12:55:38 +04001183static const struct seq_operations diskstats_op = {
Tejun Heodef4e382008-09-03 08:57:12 +02001184 .start = disk_seqf_start,
1185 .next = disk_seqf_next,
1186 .stop = disk_seqf_stop,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001187 .show = diskstats_show
1188};
Alexey Dobriyanf5009752008-10-04 23:53:21 +04001189
Alexey Dobriyan31d85ab2008-10-06 12:55:38 +04001190static int diskstats_open(struct inode *inode, struct file *file)
1191{
1192 return seq_open(file, &diskstats_op);
1193}
1194
1195static const struct file_operations proc_diskstats_operations = {
1196 .open = diskstats_open,
1197 .read = seq_read,
1198 .llseek = seq_lseek,
1199 .release = seq_release,
1200};
1201
Alexey Dobriyanf5009752008-10-04 23:53:21 +04001202static int __init proc_genhd_init(void)
1203{
Alexey Dobriyan31d85ab2008-10-06 12:55:38 +04001204 proc_create("diskstats", 0, NULL, &proc_diskstats_operations);
Alexey Dobriyanf5009752008-10-04 23:53:21 +04001205 proc_create("partitions", 0, NULL, &proc_partitions_operations);
1206 return 0;
1207}
1208module_init(proc_genhd_init);
Randy Dunlapa6e2ba82008-05-23 09:44:11 -07001209#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001210
Tejun Heocf771cb2008-09-03 09:01:09 +02001211dev_t blk_lookup_devt(const char *name, int partno)
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001212{
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001213 dev_t devt = MKDEV(0, 0);
Tejun Heodef4e382008-09-03 08:57:12 +02001214 struct class_dev_iter iter;
1215 struct device *dev;
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001216
Tejun Heodef4e382008-09-03 08:57:12 +02001217 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1218 while ((dev = class_dev_iter_next(&iter))) {
1219 struct gendisk *disk = dev_to_disk(dev);
Tejun Heo548b10e2008-08-29 09:01:47 +02001220 struct hd_struct *part;
Tejun Heodef4e382008-09-03 08:57:12 +02001221
Kay Sievers3ada8b72009-01-06 10:44:43 -08001222 if (strcmp(dev_name(dev), name))
Tejun Heof331c022008-09-03 09:01:48 +02001223 continue;
Tejun Heof331c022008-09-03 09:01:48 +02001224
Neil Brown41b8c852009-02-18 10:33:59 +01001225 if (partno < disk->minors) {
1226 /* We need to return the right devno, even
1227 * if the partition doesn't exist yet.
1228 */
1229 devt = MKDEV(MAJOR(dev->devt),
1230 MINOR(dev->devt) + partno);
1231 break;
1232 }
Tejun Heo548b10e2008-08-29 09:01:47 +02001233 part = disk_get_part(disk, partno);
Tejun Heo2bbedcb2008-08-29 11:41:51 +02001234 if (part) {
Tejun Heof331c022008-09-03 09:01:48 +02001235 devt = part_devt(part);
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001236 disk_put_part(part);
Tejun Heo548b10e2008-08-29 09:01:47 +02001237 break;
Tejun Heodef4e382008-09-03 08:57:12 +02001238 }
Tejun Heo548b10e2008-08-29 09:01:47 +02001239 disk_put_part(part);
Kay Sievers5c0ef6d2008-08-16 14:30:30 +02001240 }
Tejun Heodef4e382008-09-03 08:57:12 +02001241 class_dev_iter_exit(&iter);
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001242 return devt;
1243}
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001244EXPORT_SYMBOL(blk_lookup_devt);
1245
Linus Torvalds1da177e2005-04-16 15:20:36 -07001246struct gendisk *alloc_disk(int minors)
1247{
Christoph Lameter19460892005-06-23 00:08:19 -07001248 return alloc_disk_node(minors, -1);
1249}
Tejun Heo689d6fa2008-08-25 19:56:16 +09001250EXPORT_SYMBOL(alloc_disk);
Christoph Lameter19460892005-06-23 00:08:19 -07001251
1252struct gendisk *alloc_disk_node(int minors, int node_id)
1253{
1254 struct gendisk *disk;
1255
Christoph Lameter94f60302007-07-17 04:03:29 -07001256 disk = kmalloc_node(sizeof(struct gendisk),
1257 GFP_KERNEL | __GFP_ZERO, node_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001258 if (disk) {
Tejun Heo074a7ac2008-08-25 19:56:14 +09001259 if (!init_part_stats(&disk->part0)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001260 kfree(disk);
1261 return NULL;
1262 }
Cheng Renquanbf91db12008-11-20 08:37:37 +01001263 disk->node_id = node_id;
Tejun Heo540eed52008-08-25 19:56:15 +09001264 if (disk_expand_part_tbl(disk, 0)) {
1265 free_part_stats(&disk->part0);
Tejun Heob5d0b9d2008-09-03 09:06:42 +02001266 kfree(disk);
1267 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001268 }
Tejun Heo540eed52008-08-25 19:56:15 +09001269 disk->part_tbl->part[0] = &disk->part0;
Jens Axboe6c23a962011-01-07 08:43:37 +01001270
Vivek Goyalc83f6bf2012-08-01 12:24:18 +02001271 /*
1272 * set_capacity() and get_capacity() currently don't use
1273 * seqcounter to read/update the part0->nr_sects. Still init
1274 * the counter as we can read the sectors in IO submission
1275 * patch using seqence counters.
1276 *
1277 * TODO: Ideally set_capacity() and get_capacity() should be
1278 * converted to make use of bd_mutex and sequence counters.
1279 */
1280 seqcount_init(&disk->part0.nr_sects_seq);
Jens Axboe6c23a962011-01-07 08:43:37 +01001281 hd_ref_init(&disk->part0);
Tejun Heob5d0b9d2008-09-03 09:06:42 +02001282
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283 disk->minors = minors;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284 rand_initialize_disk(disk);
Tejun Heoed9e1982008-08-25 19:56:05 +09001285 disk_to_dev(disk)->class = &block_class;
1286 disk_to_dev(disk)->type = &disk_type;
1287 device_initialize(disk_to_dev(disk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001288 }
1289 return disk;
1290}
Christoph Lameter19460892005-06-23 00:08:19 -07001291EXPORT_SYMBOL(alloc_disk_node);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292
1293struct kobject *get_disk(struct gendisk *disk)
1294{
1295 struct module *owner;
1296 struct kobject *kobj;
1297
1298 if (!disk->fops)
1299 return NULL;
1300 owner = disk->fops->owner;
1301 if (owner && !try_module_get(owner))
1302 return NULL;
Tejun Heoed9e1982008-08-25 19:56:05 +09001303 kobj = kobject_get(&disk_to_dev(disk)->kobj);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001304 if (kobj == NULL) {
1305 module_put(owner);
1306 return NULL;
1307 }
1308 return kobj;
1309
1310}
1311
1312EXPORT_SYMBOL(get_disk);
1313
1314void put_disk(struct gendisk *disk)
1315{
1316 if (disk)
Tejun Heoed9e1982008-08-25 19:56:05 +09001317 kobject_put(&disk_to_dev(disk)->kobj);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001318}
1319
1320EXPORT_SYMBOL(put_disk);
1321
Hannes Reineckee3264a42009-07-28 09:13:13 +02001322static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1323{
1324 char event[] = "DISK_RO=1";
1325 char *envp[] = { event, NULL };
1326
1327 if (!ro)
1328 event[8] = '0';
1329 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1330}
1331
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332void set_device_ro(struct block_device *bdev, int flag)
1333{
Tejun Heob7db9952008-08-25 19:56:10 +09001334 bdev->bd_part->policy = flag;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001335}
1336
1337EXPORT_SYMBOL(set_device_ro);
1338
1339void set_disk_ro(struct gendisk *disk, int flag)
1340{
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001341 struct disk_part_iter piter;
1342 struct hd_struct *part;
1343
Hannes Reineckee3264a42009-07-28 09:13:13 +02001344 if (disk->part0.policy != flag) {
1345 set_disk_ro_uevent(disk, flag);
1346 disk->part0.policy = flag;
1347 }
1348
1349 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY);
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001350 while ((part = disk_part_iter_next(&piter)))
1351 part->policy = flag;
1352 disk_part_iter_exit(&piter);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001353}
1354
1355EXPORT_SYMBOL(set_disk_ro);
1356
1357int bdev_read_only(struct block_device *bdev)
1358{
1359 if (!bdev)
1360 return 0;
Tejun Heob7db9952008-08-25 19:56:10 +09001361 return bdev->bd_part->policy;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362}
1363
1364EXPORT_SYMBOL(bdev_read_only);
1365
Tejun Heocf771cb2008-09-03 09:01:09 +02001366int invalidate_partition(struct gendisk *disk, int partno)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001367{
1368 int res = 0;
Tejun Heocf771cb2008-09-03 09:01:09 +02001369 struct block_device *bdev = bdget_disk(disk, partno);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001370 if (bdev) {
Christoph Hellwig2ef41632005-05-05 16:15:59 -07001371 fsync_bdev(bdev);
NeilBrown93b270f2011-02-24 17:25:47 +11001372 res = __invalidate_device(bdev, true);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001373 bdput(bdev);
1374 }
1375 return res;
1376}
1377
1378EXPORT_SYMBOL(invalidate_partition);
Tejun Heo77ea8872010-12-08 20:57:37 +01001379
1380/*
1381 * Disk events - monitor disk events like media change and eject request.
1382 */
1383struct disk_events {
1384 struct list_head node; /* all disk_event's */
1385 struct gendisk *disk; /* the associated disk */
1386 spinlock_t lock;
1387
Tejun Heofdd514e2011-06-09 20:43:59 +02001388 struct mutex block_mutex; /* protects blocking */
Tejun Heo77ea8872010-12-08 20:57:37 +01001389 int block; /* event blocking depth */
1390 unsigned int pending; /* events already sent out */
1391 unsigned int clearing; /* events being cleared */
1392
1393 long poll_msecs; /* interval, -1 for default */
1394 struct delayed_work dwork;
1395};
1396
1397static const char *disk_events_strs[] = {
1398 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "media_change",
1399 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "eject_request",
1400};
1401
1402static char *disk_uevents[] = {
1403 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "DISK_MEDIA_CHANGE=1",
1404 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "DISK_EJECT_REQUEST=1",
1405};
1406
1407/* list of all disk_events */
1408static DEFINE_MUTEX(disk_events_mutex);
1409static LIST_HEAD(disk_events);
1410
1411/* disable in-kernel polling by default */
1412static unsigned long disk_events_dfl_poll_msecs = 0;
1413
1414static unsigned long disk_events_poll_jiffies(struct gendisk *disk)
1415{
1416 struct disk_events *ev = disk->ev;
1417 long intv_msecs = 0;
1418
1419 /*
1420 * If device-specific poll interval is set, always use it. If
1421 * the default is being used, poll iff there are events which
1422 * can't be monitored asynchronously.
1423 */
1424 if (ev->poll_msecs >= 0)
1425 intv_msecs = ev->poll_msecs;
1426 else if (disk->events & ~disk->async_events)
1427 intv_msecs = disk_events_dfl_poll_msecs;
1428
1429 return msecs_to_jiffies(intv_msecs);
1430}
1431
Tejun Heoc3af54a2011-06-09 20:43:55 +02001432/**
1433 * disk_block_events - block and flush disk event checking
1434 * @disk: disk to block events for
1435 *
1436 * On return from this function, it is guaranteed that event checking
1437 * isn't in progress and won't happen until unblocked by
1438 * disk_unblock_events(). Events blocking is counted and the actual
1439 * unblocking happens after the matching number of unblocks are done.
1440 *
1441 * Note that this intentionally does not block event checking from
1442 * disk_clear_events().
1443 *
1444 * CONTEXT:
1445 * Might sleep.
1446 */
1447void disk_block_events(struct gendisk *disk)
Tejun Heo77ea8872010-12-08 20:57:37 +01001448{
1449 struct disk_events *ev = disk->ev;
1450 unsigned long flags;
1451 bool cancel;
1452
Tejun Heoc3af54a2011-06-09 20:43:55 +02001453 if (!ev)
1454 return;
1455
Tejun Heofdd514e2011-06-09 20:43:59 +02001456 /*
1457 * Outer mutex ensures that the first blocker completes canceling
1458 * the event work before further blockers are allowed to finish.
1459 */
1460 mutex_lock(&ev->block_mutex);
1461
Tejun Heo77ea8872010-12-08 20:57:37 +01001462 spin_lock_irqsave(&ev->lock, flags);
1463 cancel = !ev->block++;
1464 spin_unlock_irqrestore(&ev->lock, flags);
1465
Tejun Heoc3af54a2011-06-09 20:43:55 +02001466 if (cancel)
1467 cancel_delayed_work_sync(&disk->ev->dwork);
Tejun Heofdd514e2011-06-09 20:43:59 +02001468
1469 mutex_unlock(&ev->block_mutex);
Tejun Heo77ea8872010-12-08 20:57:37 +01001470}
1471
1472static void __disk_unblock_events(struct gendisk *disk, bool check_now)
1473{
1474 struct disk_events *ev = disk->ev;
1475 unsigned long intv;
1476 unsigned long flags;
1477
1478 spin_lock_irqsave(&ev->lock, flags);
1479
1480 if (WARN_ON_ONCE(ev->block <= 0))
1481 goto out_unlock;
1482
1483 if (--ev->block)
1484 goto out_unlock;
1485
1486 /*
1487 * Not exactly a latency critical operation, set poll timer
1488 * slack to 25% and kick event check.
1489 */
1490 intv = disk_events_poll_jiffies(disk);
1491 set_timer_slack(&ev->dwork.timer, intv / 4);
1492 if (check_now)
Tejun Heo3b07e9c2012-08-20 14:51:24 -07001493 queue_delayed_work(system_freezable_wq, &ev->dwork, 0);
Tejun Heo77ea8872010-12-08 20:57:37 +01001494 else if (intv)
Tejun Heo3b07e9c2012-08-20 14:51:24 -07001495 queue_delayed_work(system_freezable_wq, &ev->dwork, intv);
Tejun Heo77ea8872010-12-08 20:57:37 +01001496out_unlock:
1497 spin_unlock_irqrestore(&ev->lock, flags);
1498}
1499
1500/**
Tejun Heo77ea8872010-12-08 20:57:37 +01001501 * disk_unblock_events - unblock disk event checking
1502 * @disk: disk to unblock events for
1503 *
1504 * Undo disk_block_events(). When the block count reaches zero, it
1505 * starts events polling if configured.
1506 *
1507 * CONTEXT:
1508 * Don't care. Safe to call from irq context.
1509 */
1510void disk_unblock_events(struct gendisk *disk)
1511{
1512 if (disk->ev)
Tejun Heofacc31d2011-03-09 19:54:27 +01001513 __disk_unblock_events(disk, false);
Tejun Heo77ea8872010-12-08 20:57:37 +01001514}
1515
1516/**
Tejun Heo85ef06d2011-07-01 16:17:47 +02001517 * disk_flush_events - schedule immediate event checking and flushing
1518 * @disk: disk to check and flush events for
1519 * @mask: events to flush
Tejun Heo77ea8872010-12-08 20:57:37 +01001520 *
Tejun Heo85ef06d2011-07-01 16:17:47 +02001521 * Schedule immediate event checking on @disk if not blocked. Events in
1522 * @mask are scheduled to be cleared from the driver. Note that this
1523 * doesn't clear the events from @disk->ev.
Tejun Heo77ea8872010-12-08 20:57:37 +01001524 *
1525 * CONTEXT:
Tejun Heo85ef06d2011-07-01 16:17:47 +02001526 * If @mask is non-zero must be called with bdev->bd_mutex held.
Tejun Heo77ea8872010-12-08 20:57:37 +01001527 */
Tejun Heo85ef06d2011-07-01 16:17:47 +02001528void disk_flush_events(struct gendisk *disk, unsigned int mask)
Tejun Heo77ea8872010-12-08 20:57:37 +01001529{
Tejun Heoa9dce2a2011-06-09 20:43:54 +02001530 struct disk_events *ev = disk->ev;
Tejun Heoa9dce2a2011-06-09 20:43:54 +02001531
1532 if (!ev)
1533 return;
1534
Tejun Heo85ef06d2011-07-01 16:17:47 +02001535 spin_lock_irq(&ev->lock);
1536 ev->clearing |= mask;
Tejun Heo41f63c52012-08-03 10:30:47 -07001537 if (!ev->block)
Tejun Heo3b07e9c2012-08-20 14:51:24 -07001538 mod_delayed_work(system_freezable_wq, &ev->dwork, 0);
Tejun Heo85ef06d2011-07-01 16:17:47 +02001539 spin_unlock_irq(&ev->lock);
Tejun Heo77ea8872010-12-08 20:57:37 +01001540}
Tejun Heo77ea8872010-12-08 20:57:37 +01001541
1542/**
1543 * disk_clear_events - synchronously check, clear and return pending events
1544 * @disk: disk to fetch and clear events from
1545 * @mask: mask of events to be fetched and clearted
1546 *
1547 * Disk events are synchronously checked and pending events in @mask
1548 * are cleared and returned. This ignores the block count.
1549 *
1550 * CONTEXT:
1551 * Might sleep.
1552 */
1553unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask)
1554{
1555 const struct block_device_operations *bdops = disk->fops;
1556 struct disk_events *ev = disk->ev;
1557 unsigned int pending;
1558
1559 if (!ev) {
1560 /* for drivers still using the old ->media_changed method */
1561 if ((mask & DISK_EVENT_MEDIA_CHANGE) &&
1562 bdops->media_changed && bdops->media_changed(disk))
1563 return DISK_EVENT_MEDIA_CHANGE;
1564 return 0;
1565 }
1566
1567 /* tell the workfn about the events being cleared */
1568 spin_lock_irq(&ev->lock);
1569 ev->clearing |= mask;
1570 spin_unlock_irq(&ev->lock);
1571
1572 /* uncondtionally schedule event check and wait for it to finish */
Tejun Heoc3af54a2011-06-09 20:43:55 +02001573 disk_block_events(disk);
Tejun Heo3b07e9c2012-08-20 14:51:24 -07001574 queue_delayed_work(system_freezable_wq, &ev->dwork, 0);
Tejun Heo77ea8872010-12-08 20:57:37 +01001575 flush_delayed_work(&ev->dwork);
1576 __disk_unblock_events(disk, false);
1577
1578 /* then, fetch and clear pending events */
1579 spin_lock_irq(&ev->lock);
1580 WARN_ON_ONCE(ev->clearing & mask); /* cleared by workfn */
1581 pending = ev->pending & mask;
1582 ev->pending &= ~mask;
1583 spin_unlock_irq(&ev->lock);
1584
1585 return pending;
1586}
1587
1588static void disk_events_workfn(struct work_struct *work)
1589{
1590 struct delayed_work *dwork = to_delayed_work(work);
1591 struct disk_events *ev = container_of(dwork, struct disk_events, dwork);
1592 struct gendisk *disk = ev->disk;
1593 char *envp[ARRAY_SIZE(disk_uevents) + 1] = { };
1594 unsigned int clearing = ev->clearing;
1595 unsigned int events;
1596 unsigned long intv;
1597 int nr_events = 0, i;
1598
1599 /* check events */
1600 events = disk->fops->check_events(disk, clearing);
1601
1602 /* accumulate pending events and schedule next poll if necessary */
1603 spin_lock_irq(&ev->lock);
1604
1605 events &= ~ev->pending;
1606 ev->pending |= events;
1607 ev->clearing &= ~clearing;
1608
1609 intv = disk_events_poll_jiffies(disk);
1610 if (!ev->block && intv)
Tejun Heo3b07e9c2012-08-20 14:51:24 -07001611 queue_delayed_work(system_freezable_wq, &ev->dwork, intv);
Tejun Heo77ea8872010-12-08 20:57:37 +01001612
1613 spin_unlock_irq(&ev->lock);
1614
Tejun Heo7c88a162011-04-21 19:43:58 +02001615 /*
1616 * Tell userland about new events. Only the events listed in
1617 * @disk->events are reported. Unlisted events are processed the
1618 * same internally but never get reported to userland.
1619 */
Tejun Heo77ea8872010-12-08 20:57:37 +01001620 for (i = 0; i < ARRAY_SIZE(disk_uevents); i++)
Tejun Heo7c88a162011-04-21 19:43:58 +02001621 if (events & disk->events & (1 << i))
Tejun Heo77ea8872010-12-08 20:57:37 +01001622 envp[nr_events++] = disk_uevents[i];
1623
1624 if (nr_events)
1625 kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp);
1626}
1627
1628/*
1629 * A disk events enabled device has the following sysfs nodes under
1630 * its /sys/block/X/ directory.
1631 *
1632 * events : list of all supported events
1633 * events_async : list of events which can be detected w/o polling
1634 * events_poll_msecs : polling interval, 0: disable, -1: system default
1635 */
1636static ssize_t __disk_events_show(unsigned int events, char *buf)
1637{
1638 const char *delim = "";
1639 ssize_t pos = 0;
1640 int i;
1641
1642 for (i = 0; i < ARRAY_SIZE(disk_events_strs); i++)
1643 if (events & (1 << i)) {
1644 pos += sprintf(buf + pos, "%s%s",
1645 delim, disk_events_strs[i]);
1646 delim = " ";
1647 }
1648 if (pos)
1649 pos += sprintf(buf + pos, "\n");
1650 return pos;
1651}
1652
1653static ssize_t disk_events_show(struct device *dev,
1654 struct device_attribute *attr, char *buf)
1655{
1656 struct gendisk *disk = dev_to_disk(dev);
1657
1658 return __disk_events_show(disk->events, buf);
1659}
1660
1661static ssize_t disk_events_async_show(struct device *dev,
1662 struct device_attribute *attr, char *buf)
1663{
1664 struct gendisk *disk = dev_to_disk(dev);
1665
1666 return __disk_events_show(disk->async_events, buf);
1667}
1668
1669static ssize_t disk_events_poll_msecs_show(struct device *dev,
1670 struct device_attribute *attr,
1671 char *buf)
1672{
1673 struct gendisk *disk = dev_to_disk(dev);
1674
1675 return sprintf(buf, "%ld\n", disk->ev->poll_msecs);
1676}
1677
1678static ssize_t disk_events_poll_msecs_store(struct device *dev,
1679 struct device_attribute *attr,
1680 const char *buf, size_t count)
1681{
1682 struct gendisk *disk = dev_to_disk(dev);
1683 long intv;
1684
1685 if (!count || !sscanf(buf, "%ld", &intv))
1686 return -EINVAL;
1687
1688 if (intv < 0 && intv != -1)
1689 return -EINVAL;
1690
Tejun Heoc3af54a2011-06-09 20:43:55 +02001691 disk_block_events(disk);
Tejun Heo77ea8872010-12-08 20:57:37 +01001692 disk->ev->poll_msecs = intv;
1693 __disk_unblock_events(disk, true);
1694
1695 return count;
1696}
1697
1698static const DEVICE_ATTR(events, S_IRUGO, disk_events_show, NULL);
1699static const DEVICE_ATTR(events_async, S_IRUGO, disk_events_async_show, NULL);
1700static const DEVICE_ATTR(events_poll_msecs, S_IRUGO|S_IWUSR,
1701 disk_events_poll_msecs_show,
1702 disk_events_poll_msecs_store);
1703
1704static const struct attribute *disk_events_attrs[] = {
1705 &dev_attr_events.attr,
1706 &dev_attr_events_async.attr,
1707 &dev_attr_events_poll_msecs.attr,
1708 NULL,
1709};
1710
1711/*
1712 * The default polling interval can be specified by the kernel
1713 * parameter block.events_dfl_poll_msecs which defaults to 0
1714 * (disable). This can also be modified runtime by writing to
1715 * /sys/module/block/events_dfl_poll_msecs.
1716 */
1717static int disk_events_set_dfl_poll_msecs(const char *val,
1718 const struct kernel_param *kp)
1719{
1720 struct disk_events *ev;
1721 int ret;
1722
1723 ret = param_set_ulong(val, kp);
1724 if (ret < 0)
1725 return ret;
1726
1727 mutex_lock(&disk_events_mutex);
1728
1729 list_for_each_entry(ev, &disk_events, node)
Tejun Heo85ef06d2011-07-01 16:17:47 +02001730 disk_flush_events(ev->disk, 0);
Tejun Heo77ea8872010-12-08 20:57:37 +01001731
1732 mutex_unlock(&disk_events_mutex);
1733
1734 return 0;
1735}
1736
1737static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops = {
1738 .set = disk_events_set_dfl_poll_msecs,
1739 .get = param_get_ulong,
1740};
1741
1742#undef MODULE_PARAM_PREFIX
1743#define MODULE_PARAM_PREFIX "block."
1744
1745module_param_cb(events_dfl_poll_msecs, &disk_events_dfl_poll_msecs_param_ops,
1746 &disk_events_dfl_poll_msecs, 0644);
1747
1748/*
Stanislaw Gruszka9f53d2fe2012-03-02 10:43:28 +01001749 * disk_{alloc|add|del|release}_events - initialize and destroy disk_events.
Tejun Heo77ea8872010-12-08 20:57:37 +01001750 */
Stanislaw Gruszka9f53d2fe2012-03-02 10:43:28 +01001751static void disk_alloc_events(struct gendisk *disk)
Tejun Heo77ea8872010-12-08 20:57:37 +01001752{
1753 struct disk_events *ev;
1754
Tejun Heo75e3f3e2011-05-26 21:06:50 +02001755 if (!disk->fops->check_events)
Tejun Heo77ea8872010-12-08 20:57:37 +01001756 return;
1757
1758 ev = kzalloc(sizeof(*ev), GFP_KERNEL);
1759 if (!ev) {
1760 pr_warn("%s: failed to initialize events\n", disk->disk_name);
1761 return;
1762 }
1763
Tejun Heo77ea8872010-12-08 20:57:37 +01001764 INIT_LIST_HEAD(&ev->node);
1765 ev->disk = disk;
1766 spin_lock_init(&ev->lock);
Tejun Heofdd514e2011-06-09 20:43:59 +02001767 mutex_init(&ev->block_mutex);
Tejun Heo77ea8872010-12-08 20:57:37 +01001768 ev->block = 1;
1769 ev->poll_msecs = -1;
1770 INIT_DELAYED_WORK(&ev->dwork, disk_events_workfn);
1771
Stanislaw Gruszka9f53d2fe2012-03-02 10:43:28 +01001772 disk->ev = ev;
1773}
1774
1775static void disk_add_events(struct gendisk *disk)
1776{
1777 if (!disk->ev)
1778 return;
1779
1780 /* FIXME: error handling */
1781 if (sysfs_create_files(&disk_to_dev(disk)->kobj, disk_events_attrs) < 0)
1782 pr_warn("%s: failed to create sysfs files for events\n",
1783 disk->disk_name);
1784
Tejun Heo77ea8872010-12-08 20:57:37 +01001785 mutex_lock(&disk_events_mutex);
Stanislaw Gruszka9f53d2fe2012-03-02 10:43:28 +01001786 list_add_tail(&disk->ev->node, &disk_events);
Tejun Heo77ea8872010-12-08 20:57:37 +01001787 mutex_unlock(&disk_events_mutex);
1788
1789 /*
1790 * Block count is initialized to 1 and the following initial
1791 * unblock kicks it into action.
1792 */
1793 __disk_unblock_events(disk, true);
1794}
1795
1796static void disk_del_events(struct gendisk *disk)
1797{
1798 if (!disk->ev)
1799 return;
1800
Tejun Heoc3af54a2011-06-09 20:43:55 +02001801 disk_block_events(disk);
Tejun Heo77ea8872010-12-08 20:57:37 +01001802
1803 mutex_lock(&disk_events_mutex);
1804 list_del_init(&disk->ev->node);
1805 mutex_unlock(&disk_events_mutex);
1806
1807 sysfs_remove_files(&disk_to_dev(disk)->kobj, disk_events_attrs);
1808}
1809
1810static void disk_release_events(struct gendisk *disk)
1811{
1812 /* the block count should be 1 from disk_del_events() */
1813 WARN_ON_ONCE(disk->ev && disk->ev->block != 1);
1814 kfree(disk->ev);
1815}