blob: 02e9fca808256f762e0af44e8a1055c322b51b40 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * gendisk handling
3 */
4
Linus Torvalds1da177e2005-04-16 15:20:36 -07005#include <linux/module.h>
6#include <linux/fs.h>
7#include <linux/genhd.h>
Andrew Mortonb446b602007-02-20 13:57:48 -08008#include <linux/kdev_t.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -07009#include <linux/kernel.h>
10#include <linux/blkdev.h>
11#include <linux/init.h>
12#include <linux/spinlock.h>
Alexey Dobriyanf5009752008-10-04 23:53:21 +040013#include <linux/proc_fs.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070014#include <linux/seq_file.h>
15#include <linux/slab.h>
16#include <linux/kmod.h>
17#include <linux/kobj_map.h>
Christoph Hellwig2ef41632005-05-05 16:15:59 -070018#include <linux/buffer_head.h>
Jes Sorensen58383af2006-02-06 14:12:43 -080019#include <linux/mutex.h>
Tejun Heobcce3de2008-08-25 19:47:22 +090020#include <linux/idr.h>
Tejun Heo77ea8872010-12-08 20:57:37 +010021#include <linux/log2.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070022
Adrian Bunkff889722008-03-04 11:23:45 +010023#include "blk.h"
24
Kay Sieversedfaa7c2007-05-21 22:08:01 +020025static DEFINE_MUTEX(block_class_lock);
Kay Sieversedfaa7c2007-05-21 22:08:01 +020026struct kobject *block_depr;
Linus Torvalds1da177e2005-04-16 15:20:36 -070027
Tejun Heobcce3de2008-08-25 19:47:22 +090028/* for extended dynamic devt allocation, currently only one major is used */
29#define MAX_EXT_DEVT (1 << MINORBITS)
30
31/* For extended devt allocation. ext_devt_mutex prevents look up
32 * results from going away underneath its user.
33 */
34static DEFINE_MUTEX(ext_devt_mutex);
35static DEFINE_IDR(ext_devt_idr);
36
Adrian Bunk1826ead2008-03-04 11:23:46 +010037static struct device_type disk_type;
38
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
Tejun Heod2bf1b62010-12-08 20:57:36 +0100510void register_disk(struct gendisk *disk)
511{
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
Wanlong Gao9f5e4862011-06-13 10:45:43 +0200605 /* Register BDI before referencing it from bdev */
Signed-off-by: Jan Kara01ea5062010-09-16 20:36:36 +0200606 bdi = &disk->queue->backing_dev_info;
607 bdi_register_dev(bdi, disk_devt(disk));
608
Tejun Heof331c022008-09-03 09:01:48 +0200609 blk_register_region(disk_devt(disk), disk->minors, NULL,
610 exact_match, exact_lock, disk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700611 register_disk(disk);
612 blk_register_queue(disk);
Peter Zijlstracf0ca9f2008-04-30 00:54:32 -0700613
Tejun Heo523e1d32011-10-19 14:31:07 +0200614 /*
615 * Take an extra ref on queue which will be put on disk_release()
616 * so that it sticks around as long as @disk is there.
617 */
618 WARN_ON_ONCE(blk_get_queue(disk->queue));
619
Tejun Heoed9e1982008-08-25 19:56:05 +0900620 retval = sysfs_create_link(&disk_to_dev(disk)->kobj, &bdi->dev->kobj,
621 "bdi");
Greg Kroah-Hartman6ffeea72008-05-22 17:21:08 -0400622 WARN_ON(retval);
Tejun Heo77ea8872010-12-08 20:57:37 +0100623
624 disk_add_events(disk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626EXPORT_SYMBOL(add_disk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627
Tejun Heod2bf1b62010-12-08 20:57:36 +0100628void del_gendisk(struct gendisk *disk)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629{
Tejun Heod2bf1b62010-12-08 20:57:36 +0100630 struct disk_part_iter piter;
631 struct hd_struct *part;
632
Tejun Heo77ea8872010-12-08 20:57:37 +0100633 disk_del_events(disk);
634
Tejun Heod2bf1b62010-12-08 20:57:36 +0100635 /* invalidate stuff */
636 disk_part_iter_init(&piter, disk,
637 DISK_PITER_INCL_EMPTY | DISK_PITER_REVERSE);
638 while ((part = disk_part_iter_next(&piter))) {
639 invalidate_partition(disk, part->partno);
640 delete_partition(disk, part->partno);
641 }
642 disk_part_iter_exit(&piter);
643
644 invalidate_partition(disk, 0);
645 blk_free_devt(disk_to_dev(disk)->devt);
646 set_capacity(disk, 0);
647 disk->flags &= ~GENHD_FL_UP;
648
Tejun Heoed9e1982008-08-25 19:56:05 +0900649 sysfs_remove_link(&disk_to_dev(disk)->kobj, "bdi");
Peter Zijlstracf0ca9f2008-04-30 00:54:32 -0700650 bdi_unregister(&disk->queue->backing_dev_info);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651 blk_unregister_queue(disk);
Tejun Heof331c022008-09-03 09:01:48 +0200652 blk_unregister_region(disk_devt(disk), disk->minors);
Tejun Heod2bf1b62010-12-08 20:57:36 +0100653
654 part_stat_set_all(&disk->part0, 0);
655 disk->part0.stamp = 0;
656
657 kobject_put(disk->part0.holder_dir);
658 kobject_put(disk->slave_dir);
659 disk->driverfs_dev = NULL;
660 if (!sysfs_deprecated)
661 sysfs_remove_link(block_depr, dev_name(disk_to_dev(disk)));
662 device_del(disk_to_dev(disk));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663}
Tejun Heod2bf1b62010-12-08 20:57:36 +0100664EXPORT_SYMBOL(del_gendisk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665
Linus Torvalds1da177e2005-04-16 15:20:36 -0700666/**
667 * get_gendisk - get partitioning information for a given device
Randy Dunlap710027a2008-08-19 20:13:11 +0200668 * @devt: device to get partitioning information for
Randy Dunlap496aa8a2008-10-16 07:46:23 +0200669 * @partno: returned partition index
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670 *
671 * This function gets the structure containing partitioning
Randy Dunlap710027a2008-08-19 20:13:11 +0200672 * information for the given device @devt.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673 */
Tejun Heocf771cb2008-09-03 09:01:09 +0200674struct gendisk *get_gendisk(dev_t devt, int *partno)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675{
Tejun Heobcce3de2008-08-25 19:47:22 +0900676 struct gendisk *disk = NULL;
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200677
Tejun Heobcce3de2008-08-25 19:47:22 +0900678 if (MAJOR(devt) != BLOCK_EXT_MAJOR) {
679 struct kobject *kobj;
680
681 kobj = kobj_lookup(bdev_map, devt, partno);
682 if (kobj)
683 disk = dev_to_disk(kobj_to_dev(kobj));
684 } else {
685 struct hd_struct *part;
686
687 mutex_lock(&ext_devt_mutex);
Tejun Heo870d6652008-08-25 19:47:25 +0900688 part = idr_find(&ext_devt_idr, blk_mangle_minor(MINOR(devt)));
Tejun Heobcce3de2008-08-25 19:47:22 +0900689 if (part && get_disk(part_to_disk(part))) {
690 *partno = part->partno;
691 disk = part_to_disk(part);
692 }
693 mutex_unlock(&ext_devt_mutex);
694 }
695
696 return disk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700697}
Divyesh Shahb6ac23af2010-04-15 08:54:59 +0200698EXPORT_SYMBOL(get_gendisk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700699
Tejun Heof331c022008-09-03 09:01:48 +0200700/**
701 * bdget_disk - do bdget() by gendisk and partition number
702 * @disk: gendisk of interest
703 * @partno: partition number
704 *
705 * Find partition @partno from @disk, do bdget() on it.
706 *
707 * CONTEXT:
708 * Don't care.
709 *
710 * RETURNS:
711 * Resulting block_device on success, NULL on failure.
712 */
Harvey Harrisonaeb3d3a2008-08-28 09:27:42 +0200713struct block_device *bdget_disk(struct gendisk *disk, int partno)
Tejun Heof331c022008-09-03 09:01:48 +0200714{
Tejun Heo548b10e2008-08-29 09:01:47 +0200715 struct hd_struct *part;
716 struct block_device *bdev = NULL;
Tejun Heof331c022008-09-03 09:01:48 +0200717
Tejun Heo548b10e2008-08-29 09:01:47 +0200718 part = disk_get_part(disk, partno);
Tejun Heo2bbedcb2008-08-29 11:41:51 +0200719 if (part)
Tejun Heo548b10e2008-08-29 09:01:47 +0200720 bdev = bdget(part_devt(part));
721 disk_put_part(part);
Tejun Heof331c022008-09-03 09:01:48 +0200722
Tejun Heo548b10e2008-08-29 09:01:47 +0200723 return bdev;
Tejun Heof331c022008-09-03 09:01:48 +0200724}
725EXPORT_SYMBOL(bdget_disk);
726
Dave Gilbertdd2a3452007-05-09 02:33:24 -0700727/*
728 * print a full list of all partitions - intended for places where the root
729 * filesystem can't be mounted and thus to give the victim some idea of what
730 * went wrong
731 */
732void __init printk_all_partitions(void)
733{
Tejun Heodef4e382008-09-03 08:57:12 +0200734 struct class_dev_iter iter;
735 struct device *dev;
736
737 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
738 while ((dev = class_dev_iter_next(&iter))) {
739 struct gendisk *disk = dev_to_disk(dev);
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200740 struct disk_part_iter piter;
741 struct hd_struct *part;
Tejun Heo1f014292008-08-25 19:47:23 +0900742 char name_buf[BDEVNAME_SIZE];
743 char devt_buf[BDEVT_SIZE];
Will Drewryb5af9212010-08-31 15:47:07 -0500744 u8 uuid[PARTITION_META_INFO_UUIDLTH * 2 + 1];
Tejun Heodef4e382008-09-03 08:57:12 +0200745
746 /*
747 * Don't show empty devices or things that have been
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300748 * suppressed
Tejun Heodef4e382008-09-03 08:57:12 +0200749 */
750 if (get_capacity(disk) == 0 ||
751 (disk->flags & GENHD_FL_SUPPRESS_PARTITION_INFO))
752 continue;
753
754 /*
755 * Note, unlike /proc/partitions, I am showing the
756 * numbers in hex - the same format as the root=
757 * option takes.
758 */
Tejun Heo074a7ac2008-08-25 19:56:14 +0900759 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_PART0);
760 while ((part = disk_part_iter_next(&piter))) {
761 bool is_part0 = part == &disk->part0;
Tejun Heodef4e382008-09-03 08:57:12 +0200762
Will Drewryb5af9212010-08-31 15:47:07 -0500763 uuid[0] = 0;
764 if (part->info)
765 part_unpack_uuid(part->info->uuid, uuid);
766
767 printk("%s%s %10llu %s %s", is_part0 ? "" : " ",
Tejun Heo1f014292008-08-25 19:47:23 +0900768 bdevt_str(part_devt(part), devt_buf),
Tejun Heof331c022008-09-03 09:01:48 +0200769 (unsigned long long)part->nr_sects >> 1,
Will Drewryb5af9212010-08-31 15:47:07 -0500770 disk_name(disk, part->partno, name_buf), uuid);
Tejun Heo074a7ac2008-08-25 19:56:14 +0900771 if (is_part0) {
772 if (disk->driverfs_dev != NULL &&
773 disk->driverfs_dev->driver != NULL)
774 printk(" driver: %s\n",
775 disk->driverfs_dev->driver->name);
776 else
777 printk(" (driver?)\n");
778 } else
779 printk("\n");
780 }
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200781 disk_part_iter_exit(&piter);
Tejun Heodef4e382008-09-03 08:57:12 +0200782 }
783 class_dev_iter_exit(&iter);
Dave Gilbertdd2a3452007-05-09 02:33:24 -0700784}
785
Linus Torvalds1da177e2005-04-16 15:20:36 -0700786#ifdef CONFIG_PROC_FS
787/* iterator */
Tejun Heodef4e382008-09-03 08:57:12 +0200788static void *disk_seqf_start(struct seq_file *seqf, loff_t *pos)
Greg Kroah-Hartman68c4d4a2008-05-22 17:21:08 -0400789{
Tejun Heodef4e382008-09-03 08:57:12 +0200790 loff_t skip = *pos;
791 struct class_dev_iter *iter;
792 struct device *dev;
Greg Kroah-Hartman68c4d4a2008-05-22 17:21:08 -0400793
Harvey Harrisonaeb3d3a2008-08-28 09:27:42 +0200794 iter = kmalloc(sizeof(*iter), GFP_KERNEL);
Tejun Heodef4e382008-09-03 08:57:12 +0200795 if (!iter)
796 return ERR_PTR(-ENOMEM);
797
798 seqf->private = iter;
799 class_dev_iter_init(iter, &block_class, NULL, &disk_type);
800 do {
801 dev = class_dev_iter_next(iter);
802 if (!dev)
803 return NULL;
804 } while (skip--);
805
806 return dev_to_disk(dev);
Greg Kroah-Hartman68c4d4a2008-05-22 17:21:08 -0400807}
808
Tejun Heodef4e382008-09-03 08:57:12 +0200809static void *disk_seqf_next(struct seq_file *seqf, void *v, loff_t *pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700810{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200811 struct device *dev;
Greg Kroah-Hartman66c64af2008-05-22 17:21:08 -0400812
Tejun Heodef4e382008-09-03 08:57:12 +0200813 (*pos)++;
814 dev = class_dev_iter_next(seqf->private);
Tejun Heo2ac3cee2008-09-03 08:53:37 +0200815 if (dev)
Greg Kroah-Hartman68c4d4a2008-05-22 17:21:08 -0400816 return dev_to_disk(dev);
Tejun Heo2ac3cee2008-09-03 08:53:37 +0200817
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818 return NULL;
819}
820
Tejun Heodef4e382008-09-03 08:57:12 +0200821static void disk_seqf_stop(struct seq_file *seqf, void *v)
Greg Kroah-Hartman27f30252008-05-22 17:21:08 -0400822{
Tejun Heodef4e382008-09-03 08:57:12 +0200823 struct class_dev_iter *iter = seqf->private;
Greg Kroah-Hartman27f30252008-05-22 17:21:08 -0400824
Tejun Heodef4e382008-09-03 08:57:12 +0200825 /* stop is called even after start failed :-( */
826 if (iter) {
827 class_dev_iter_exit(iter);
828 kfree(iter);
Kay Sievers5c0ef6d2008-08-16 14:30:30 +0200829 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830}
831
Tejun Heodef4e382008-09-03 08:57:12 +0200832static void *show_partition_start(struct seq_file *seqf, loff_t *pos)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700833{
Tejun Heodef4e382008-09-03 08:57:12 +0200834 static void *p;
835
836 p = disk_seqf_start(seqf, pos);
Yang Zhangb9f985b2010-12-17 08:58:36 +0100837 if (!IS_ERR_OR_NULL(p) && !*pos)
Tejun Heodef4e382008-09-03 08:57:12 +0200838 seq_puts(seqf, "major minor #blocks name\n\n");
839 return p;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700840}
841
Tejun Heocf771cb2008-09-03 09:01:09 +0200842static int show_partition(struct seq_file *seqf, void *v)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700843{
844 struct gendisk *sgp = v;
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200845 struct disk_part_iter piter;
846 struct hd_struct *part;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847 char buf[BDEVNAME_SIZE];
848
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849 /* Don't show non-partitionable removeable devices or empty devices */
Tejun Heod27769e2011-08-23 20:01:04 +0200850 if (!get_capacity(sgp) || (!disk_max_parts(sgp) &&
Tejun Heof331c022008-09-03 09:01:48 +0200851 (sgp->flags & GENHD_FL_REMOVABLE)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700852 return 0;
853 if (sgp->flags & GENHD_FL_SUPPRESS_PARTITION_INFO)
854 return 0;
855
856 /* show the full disk and all non-0 size partitions of it */
Tejun Heo074a7ac2008-08-25 19:56:14 +0900857 disk_part_iter_init(&piter, sgp, DISK_PITER_INCL_PART0);
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200858 while ((part = disk_part_iter_next(&piter)))
Tejun Heo1f014292008-08-25 19:47:23 +0900859 seq_printf(seqf, "%4d %7d %10llu %s\n",
Tejun Heof331c022008-09-03 09:01:48 +0200860 MAJOR(part_devt(part)), MINOR(part_devt(part)),
861 (unsigned long long)part->nr_sects >> 1,
862 disk_name(sgp, part->partno, buf));
Tejun Heoe71bf0d2008-09-03 09:03:02 +0200863 disk_part_iter_exit(&piter);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864
865 return 0;
866}
867
Alexey Dobriyanf5009752008-10-04 23:53:21 +0400868static const struct seq_operations partitions_op = {
Tejun Heodef4e382008-09-03 08:57:12 +0200869 .start = show_partition_start,
870 .next = disk_seqf_next,
871 .stop = disk_seqf_stop,
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200872 .show = show_partition
Linus Torvalds1da177e2005-04-16 15:20:36 -0700873};
Alexey Dobriyanf5009752008-10-04 23:53:21 +0400874
875static int partitions_open(struct inode *inode, struct file *file)
876{
877 return seq_open(file, &partitions_op);
878}
879
880static const struct file_operations proc_partitions_operations = {
881 .open = partitions_open,
882 .read = seq_read,
883 .llseek = seq_lseek,
884 .release = seq_release,
885};
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886#endif
887
888
Tejun Heocf771cb2008-09-03 09:01:09 +0200889static struct kobject *base_probe(dev_t devt, int *partno, void *data)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200891 if (request_module("block-major-%d-%d", MAJOR(devt), MINOR(devt)) > 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700892 /* Make old-style 2.4 aliases work */
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200893 request_module("block-major-%d", MAJOR(devt));
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894 return NULL;
895}
896
897static int __init genhd_device_init(void)
898{
Dan Williamse105b8b2008-04-21 10:51:07 -0700899 int error;
900
901 block_class.dev_kobj = sysfs_dev_block_kobj;
902 error = class_register(&block_class);
Roland McGrathee27a552008-03-11 17:13:15 -0700903 if (unlikely(error))
904 return error;
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200905 bdev_map = kobj_map_init(base_probe, &block_class_lock);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700906 blk_dev_init();
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200907
Zhang, Yanmin561ec682008-11-14 08:26:30 +0100908 register_blkdev(BLOCK_EXT_MAJOR, "blkext");
909
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200910 /* create top-level block dir */
Andi Kleene52eec12010-09-08 16:54:17 +0200911 if (!sysfs_deprecated)
912 block_depr = kobject_create_and_add("block", NULL);
Greg Kroah-Hartman830d3cf2007-11-06 10:36:58 -0800913 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914}
915
916subsys_initcall(genhd_device_init);
917
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200918static ssize_t disk_range_show(struct device *dev,
919 struct device_attribute *attr, char *buf)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200921 struct gendisk *disk = dev_to_disk(dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200923 return sprintf(buf, "%d\n", disk->minors);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924}
925
Tejun Heo1f014292008-08-25 19:47:23 +0900926static ssize_t disk_ext_range_show(struct device *dev,
927 struct device_attribute *attr, char *buf)
928{
929 struct gendisk *disk = dev_to_disk(dev);
930
Tejun Heob5d0b9d2008-09-03 09:06:42 +0200931 return sprintf(buf, "%d\n", disk_max_parts(disk));
Tejun Heo1f014292008-08-25 19:47:23 +0900932}
933
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200934static ssize_t disk_removable_show(struct device *dev,
935 struct device_attribute *attr, char *buf)
Kay Sieversa7fd6702005-10-01 14:49:43 +0200936{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200937 struct gendisk *disk = dev_to_disk(dev);
Kay Sieversa7fd6702005-10-01 14:49:43 +0200938
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200939 return sprintf(buf, "%d\n",
Linus Torvalds1da177e2005-04-16 15:20:36 -0700940 (disk->flags & GENHD_FL_REMOVABLE ? 1 : 0));
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200941}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700942
Kay Sievers1c9ce522008-06-13 09:41:00 +0200943static ssize_t disk_ro_show(struct device *dev,
944 struct device_attribute *attr, char *buf)
945{
946 struct gendisk *disk = dev_to_disk(dev);
947
Tejun Heob7db9952008-08-25 19:56:10 +0900948 return sprintf(buf, "%d\n", get_disk_ro(disk) ? 1 : 0);
Kay Sievers1c9ce522008-06-13 09:41:00 +0200949}
950
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200951static ssize_t disk_capability_show(struct device *dev,
952 struct device_attribute *attr, char *buf)
Kristen Carlson Accardi86ce18d2007-05-23 13:57:38 -0700953{
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200954 struct gendisk *disk = dev_to_disk(dev);
955
956 return sprintf(buf, "%x\n", disk->flags);
Kristen Carlson Accardi86ce18d2007-05-23 13:57:38 -0700957}
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200958
Martin K. Petersenc72758f2009-05-22 17:17:53 -0400959static ssize_t disk_alignment_offset_show(struct device *dev,
960 struct device_attribute *attr,
961 char *buf)
962{
963 struct gendisk *disk = dev_to_disk(dev);
964
965 return sprintf(buf, "%d\n", queue_alignment_offset(disk->queue));
966}
967
Martin K. Petersen86b37282009-11-10 11:50:21 +0100968static ssize_t disk_discard_alignment_show(struct device *dev,
969 struct device_attribute *attr,
970 char *buf)
971{
972 struct gendisk *disk = dev_to_disk(dev);
973
Martin K. Petersendd3d1452010-01-11 03:21:48 -0500974 return sprintf(buf, "%d\n", queue_discard_alignment(disk->queue));
Martin K. Petersen86b37282009-11-10 11:50:21 +0100975}
976
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200977static DEVICE_ATTR(range, S_IRUGO, disk_range_show, NULL);
Tejun Heo1f014292008-08-25 19:47:23 +0900978static DEVICE_ATTR(ext_range, S_IRUGO, disk_ext_range_show, NULL);
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200979static DEVICE_ATTR(removable, S_IRUGO, disk_removable_show, NULL);
Kay Sievers1c9ce522008-06-13 09:41:00 +0200980static DEVICE_ATTR(ro, S_IRUGO, disk_ro_show, NULL);
Tejun Heoe5610522008-08-25 19:56:09 +0900981static DEVICE_ATTR(size, S_IRUGO, part_size_show, NULL);
Martin K. Petersenc72758f2009-05-22 17:17:53 -0400982static DEVICE_ATTR(alignment_offset, S_IRUGO, disk_alignment_offset_show, NULL);
Martin K. Petersen86b37282009-11-10 11:50:21 +0100983static DEVICE_ATTR(discard_alignment, S_IRUGO, disk_discard_alignment_show,
984 NULL);
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200985static DEVICE_ATTR(capability, S_IRUGO, disk_capability_show, NULL);
Tejun Heo074a7ac2008-08-25 19:56:14 +0900986static DEVICE_ATTR(stat, S_IRUGO, part_stat_show, NULL);
Nikanth Karthikesan316d3152009-10-06 20:16:55 +0200987static DEVICE_ATTR(inflight, S_IRUGO, part_inflight_show, NULL);
Akinobu Mitac17bb492006-12-08 02:39:46 -0800988#ifdef CONFIG_FAIL_MAKE_REQUEST
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200989static struct device_attribute dev_attr_fail =
Tejun Heoeddb2e22008-08-25 19:56:13 +0900990 __ATTR(make-it-fail, S_IRUGO|S_IWUSR, part_fail_show, part_fail_store);
Akinobu Mitac17bb492006-12-08 02:39:46 -0800991#endif
Jens Axboe581d4e22008-09-14 05:56:33 -0700992#ifdef CONFIG_FAIL_IO_TIMEOUT
993static struct device_attribute dev_attr_fail_timeout =
994 __ATTR(io-timeout-fail, S_IRUGO|S_IWUSR, part_timeout_show,
995 part_timeout_store);
996#endif
Kay Sieversedfaa7c2007-05-21 22:08:01 +0200997
998static struct attribute *disk_attrs[] = {
999 &dev_attr_range.attr,
Tejun Heo1f014292008-08-25 19:47:23 +09001000 &dev_attr_ext_range.attr,
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001001 &dev_attr_removable.attr,
Kay Sievers1c9ce522008-06-13 09:41:00 +02001002 &dev_attr_ro.attr,
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001003 &dev_attr_size.attr,
Martin K. Petersenc72758f2009-05-22 17:17:53 -04001004 &dev_attr_alignment_offset.attr,
Martin K. Petersen86b37282009-11-10 11:50:21 +01001005 &dev_attr_discard_alignment.attr,
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001006 &dev_attr_capability.attr,
1007 &dev_attr_stat.attr,
Nikanth Karthikesan316d3152009-10-06 20:16:55 +02001008 &dev_attr_inflight.attr,
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001009#ifdef CONFIG_FAIL_MAKE_REQUEST
1010 &dev_attr_fail.attr,
1011#endif
Jens Axboe581d4e22008-09-14 05:56:33 -07001012#ifdef CONFIG_FAIL_IO_TIMEOUT
1013 &dev_attr_fail_timeout.attr,
1014#endif
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001015 NULL
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016};
1017
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001018static struct attribute_group disk_attr_group = {
1019 .attrs = disk_attrs,
1020};
1021
David Brownella4dbd672009-06-24 10:06:31 -07001022static const struct attribute_group *disk_attr_groups[] = {
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001023 &disk_attr_group,
1024 NULL
1025};
1026
Tejun Heo540eed52008-08-25 19:56:15 +09001027/**
1028 * disk_replace_part_tbl - replace disk->part_tbl in RCU-safe way
1029 * @disk: disk to replace part_tbl for
1030 * @new_ptbl: new part_tbl to install
1031 *
1032 * Replace disk->part_tbl with @new_ptbl in RCU-safe way. The
1033 * original ptbl is freed using RCU callback.
1034 *
1035 * LOCKING:
1036 * Matching bd_mutx locked.
1037 */
1038static void disk_replace_part_tbl(struct gendisk *disk,
1039 struct disk_part_tbl *new_ptbl)
1040{
1041 struct disk_part_tbl *old_ptbl = disk->part_tbl;
1042
1043 rcu_assign_pointer(disk->part_tbl, new_ptbl);
Jens Axboea6f23652008-10-24 12:52:42 +02001044
1045 if (old_ptbl) {
1046 rcu_assign_pointer(old_ptbl->last_lookup, NULL);
Lai Jiangshan57bdfbf2011-03-18 11:42:58 +08001047 kfree_rcu(old_ptbl, rcu_head);
Jens Axboea6f23652008-10-24 12:52:42 +02001048 }
Tejun Heo540eed52008-08-25 19:56:15 +09001049}
1050
1051/**
1052 * disk_expand_part_tbl - expand disk->part_tbl
1053 * @disk: disk to expand part_tbl for
1054 * @partno: expand such that this partno can fit in
1055 *
1056 * Expand disk->part_tbl such that @partno can fit in. disk->part_tbl
1057 * uses RCU to allow unlocked dereferencing for stats and other stuff.
1058 *
1059 * LOCKING:
1060 * Matching bd_mutex locked, might sleep.
1061 *
1062 * RETURNS:
1063 * 0 on success, -errno on failure.
1064 */
1065int disk_expand_part_tbl(struct gendisk *disk, int partno)
1066{
1067 struct disk_part_tbl *old_ptbl = disk->part_tbl;
1068 struct disk_part_tbl *new_ptbl;
1069 int len = old_ptbl ? old_ptbl->len : 0;
1070 int target = partno + 1;
1071 size_t size;
1072 int i;
1073
1074 /* disk_max_parts() is zero during initialization, ignore if so */
1075 if (disk_max_parts(disk) && target > disk_max_parts(disk))
1076 return -EINVAL;
1077
1078 if (target <= len)
1079 return 0;
1080
1081 size = sizeof(*new_ptbl) + target * sizeof(new_ptbl->part[0]);
1082 new_ptbl = kzalloc_node(size, GFP_KERNEL, disk->node_id);
1083 if (!new_ptbl)
1084 return -ENOMEM;
1085
Tejun Heo540eed52008-08-25 19:56:15 +09001086 new_ptbl->len = target;
1087
1088 for (i = 0; i < len; i++)
1089 rcu_assign_pointer(new_ptbl->part[i], old_ptbl->part[i]);
1090
1091 disk_replace_part_tbl(disk, new_ptbl);
1092 return 0;
1093}
1094
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001095static void disk_release(struct device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096{
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001097 struct gendisk *disk = dev_to_disk(dev);
1098
Tejun Heo77ea8872010-12-08 20:57:37 +01001099 disk_release_events(disk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100 kfree(disk->random);
Tejun Heo540eed52008-08-25 19:56:15 +09001101 disk_replace_part_tbl(disk, NULL);
Tejun Heo074a7ac2008-08-25 19:56:14 +09001102 free_part_stats(&disk->part0);
Will Drewry6d1d8052010-08-31 15:47:05 -05001103 free_part_info(&disk->part0);
Tejun Heo523e1d32011-10-19 14:31:07 +02001104 if (disk->queue)
1105 blk_put_queue(disk->queue);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106 kfree(disk);
1107}
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001108struct class block_class = {
1109 .name = "block",
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110};
1111
Kay Sieverse454cea2009-09-18 23:01:12 +02001112static char *block_devnode(struct device *dev, mode_t *mode)
Kay Sieversb03f38b2009-04-30 15:23:42 +02001113{
1114 struct gendisk *disk = dev_to_disk(dev);
1115
Kay Sieverse454cea2009-09-18 23:01:12 +02001116 if (disk->devnode)
1117 return disk->devnode(disk, mode);
Kay Sieversb03f38b2009-04-30 15:23:42 +02001118 return NULL;
1119}
1120
Adrian Bunk1826ead2008-03-04 11:23:46 +01001121static struct device_type disk_type = {
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001122 .name = "disk",
1123 .groups = disk_attr_groups,
1124 .release = disk_release,
Kay Sieverse454cea2009-09-18 23:01:12 +02001125 .devnode = block_devnode,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001126};
1127
Randy Dunlapa6e2ba82008-05-23 09:44:11 -07001128#ifdef CONFIG_PROC_FS
Tejun Heocf771cb2008-09-03 09:01:09 +02001129/*
1130 * aggregate disk stat collector. Uses the same stats that the sysfs
1131 * entries do, above, but makes them available through one seq_file.
1132 *
1133 * The output looks suspiciously like /proc/partitions with a bunch of
1134 * extra fields.
1135 */
1136static int diskstats_show(struct seq_file *seqf, void *v)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001137{
1138 struct gendisk *gp = v;
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001139 struct disk_part_iter piter;
1140 struct hd_struct *hd;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001141 char buf[BDEVNAME_SIZE];
Tejun Heoc9959052008-08-25 19:47:21 +09001142 int cpu;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001143
1144 /*
Tejun Heoed9e1982008-08-25 19:56:05 +09001145 if (&disk_to_dev(gp)->kobj.entry == block_class.devices.next)
Tejun Heocf771cb2008-09-03 09:01:09 +02001146 seq_puts(seqf, "major minor name"
Linus Torvalds1da177e2005-04-16 15:20:36 -07001147 " rio rmerge rsect ruse wio wmerge "
1148 "wsect wuse running use aveq"
1149 "\n\n");
1150 */
Wanlong Gao9f5e4862011-06-13 10:45:43 +02001151
Tejun Heo71982a42009-04-17 08:34:48 +02001152 disk_part_iter_init(&piter, gp, DISK_PITER_INCL_EMPTY_PART0);
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001153 while ((hd = disk_part_iter_next(&piter))) {
Tejun Heo074a7ac2008-08-25 19:56:14 +09001154 cpu = part_stat_lock();
Tejun Heoc9959052008-08-25 19:47:21 +09001155 part_round_stats(cpu, hd);
Tejun Heo074a7ac2008-08-25 19:56:14 +09001156 part_stat_unlock();
Herbert Poetzlf95fe9c2011-08-02 12:43:50 +02001157 seq_printf(seqf, "%4d %7d %s %lu %lu %lu "
1158 "%u %lu %lu %lu %u %u %u %u\n",
Tejun Heof331c022008-09-03 09:01:48 +02001159 MAJOR(part_devt(hd)), MINOR(part_devt(hd)),
1160 disk_name(gp, hd->partno, buf),
Liu Yuan53f22952011-03-02 11:00:15 -05001161 part_stat_read(hd, ios[READ]),
1162 part_stat_read(hd, merges[READ]),
Herbert Poetzlf95fe9c2011-08-02 12:43:50 +02001163 part_stat_read(hd, sectors[READ]),
Liu Yuan53f22952011-03-02 11:00:15 -05001164 jiffies_to_msecs(part_stat_read(hd, ticks[READ])),
1165 part_stat_read(hd, ios[WRITE]),
1166 part_stat_read(hd, merges[WRITE]),
Herbert Poetzlf95fe9c2011-08-02 12:43:50 +02001167 part_stat_read(hd, sectors[WRITE]),
Liu Yuan53f22952011-03-02 11:00:15 -05001168 jiffies_to_msecs(part_stat_read(hd, ticks[WRITE])),
Nikanth Karthikesan316d3152009-10-06 20:16:55 +02001169 part_in_flight(hd),
Jerome Marchand28f39d52008-02-08 11:04:56 +01001170 jiffies_to_msecs(part_stat_read(hd, io_ticks)),
1171 jiffies_to_msecs(part_stat_read(hd, time_in_queue))
1172 );
Linus Torvalds1da177e2005-04-16 15:20:36 -07001173 }
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001174 disk_part_iter_exit(&piter);
Wanlong Gao9f5e4862011-06-13 10:45:43 +02001175
Linus Torvalds1da177e2005-04-16 15:20:36 -07001176 return 0;
1177}
1178
Alexey Dobriyan31d85ab2008-10-06 12:55:38 +04001179static const struct seq_operations diskstats_op = {
Tejun Heodef4e382008-09-03 08:57:12 +02001180 .start = disk_seqf_start,
1181 .next = disk_seqf_next,
1182 .stop = disk_seqf_stop,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183 .show = diskstats_show
1184};
Alexey Dobriyanf5009752008-10-04 23:53:21 +04001185
Alexey Dobriyan31d85ab2008-10-06 12:55:38 +04001186static int diskstats_open(struct inode *inode, struct file *file)
1187{
1188 return seq_open(file, &diskstats_op);
1189}
1190
1191static const struct file_operations proc_diskstats_operations = {
1192 .open = diskstats_open,
1193 .read = seq_read,
1194 .llseek = seq_lseek,
1195 .release = seq_release,
1196};
1197
Alexey Dobriyanf5009752008-10-04 23:53:21 +04001198static int __init proc_genhd_init(void)
1199{
Alexey Dobriyan31d85ab2008-10-06 12:55:38 +04001200 proc_create("diskstats", 0, NULL, &proc_diskstats_operations);
Alexey Dobriyanf5009752008-10-04 23:53:21 +04001201 proc_create("partitions", 0, NULL, &proc_partitions_operations);
1202 return 0;
1203}
1204module_init(proc_genhd_init);
Randy Dunlapa6e2ba82008-05-23 09:44:11 -07001205#endif /* CONFIG_PROC_FS */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001206
Tejun Heocf771cb2008-09-03 09:01:09 +02001207dev_t blk_lookup_devt(const char *name, int partno)
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001208{
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001209 dev_t devt = MKDEV(0, 0);
Tejun Heodef4e382008-09-03 08:57:12 +02001210 struct class_dev_iter iter;
1211 struct device *dev;
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001212
Tejun Heodef4e382008-09-03 08:57:12 +02001213 class_dev_iter_init(&iter, &block_class, NULL, &disk_type);
1214 while ((dev = class_dev_iter_next(&iter))) {
1215 struct gendisk *disk = dev_to_disk(dev);
Tejun Heo548b10e2008-08-29 09:01:47 +02001216 struct hd_struct *part;
Tejun Heodef4e382008-09-03 08:57:12 +02001217
Kay Sievers3ada8b72009-01-06 10:44:43 -08001218 if (strcmp(dev_name(dev), name))
Tejun Heof331c022008-09-03 09:01:48 +02001219 continue;
Tejun Heof331c022008-09-03 09:01:48 +02001220
Neil Brown41b8c852009-02-18 10:33:59 +01001221 if (partno < disk->minors) {
1222 /* We need to return the right devno, even
1223 * if the partition doesn't exist yet.
1224 */
1225 devt = MKDEV(MAJOR(dev->devt),
1226 MINOR(dev->devt) + partno);
1227 break;
1228 }
Tejun Heo548b10e2008-08-29 09:01:47 +02001229 part = disk_get_part(disk, partno);
Tejun Heo2bbedcb2008-08-29 11:41:51 +02001230 if (part) {
Tejun Heof331c022008-09-03 09:01:48 +02001231 devt = part_devt(part);
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001232 disk_put_part(part);
Tejun Heo548b10e2008-08-29 09:01:47 +02001233 break;
Tejun Heodef4e382008-09-03 08:57:12 +02001234 }
Tejun Heo548b10e2008-08-29 09:01:47 +02001235 disk_put_part(part);
Kay Sievers5c0ef6d2008-08-16 14:30:30 +02001236 }
Tejun Heodef4e382008-09-03 08:57:12 +02001237 class_dev_iter_exit(&iter);
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001238 return devt;
1239}
Kay Sieversedfaa7c2007-05-21 22:08:01 +02001240EXPORT_SYMBOL(blk_lookup_devt);
1241
Linus Torvalds1da177e2005-04-16 15:20:36 -07001242struct gendisk *alloc_disk(int minors)
1243{
Christoph Lameter19460892005-06-23 00:08:19 -07001244 return alloc_disk_node(minors, -1);
1245}
Tejun Heo689d6fa2008-08-25 19:56:16 +09001246EXPORT_SYMBOL(alloc_disk);
Christoph Lameter19460892005-06-23 00:08:19 -07001247
1248struct gendisk *alloc_disk_node(int minors, int node_id)
1249{
1250 struct gendisk *disk;
1251
Christoph Lameter94f60302007-07-17 04:03:29 -07001252 disk = kmalloc_node(sizeof(struct gendisk),
1253 GFP_KERNEL | __GFP_ZERO, node_id);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001254 if (disk) {
Tejun Heo074a7ac2008-08-25 19:56:14 +09001255 if (!init_part_stats(&disk->part0)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001256 kfree(disk);
1257 return NULL;
1258 }
Cheng Renquanbf91db12008-11-20 08:37:37 +01001259 disk->node_id = node_id;
Tejun Heo540eed52008-08-25 19:56:15 +09001260 if (disk_expand_part_tbl(disk, 0)) {
1261 free_part_stats(&disk->part0);
Tejun Heob5d0b9d2008-09-03 09:06:42 +02001262 kfree(disk);
1263 return NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001264 }
Tejun Heo540eed52008-08-25 19:56:15 +09001265 disk->part_tbl->part[0] = &disk->part0;
Jens Axboe6c23a962011-01-07 08:43:37 +01001266
1267 hd_ref_init(&disk->part0);
Tejun Heob5d0b9d2008-09-03 09:06:42 +02001268
Linus Torvalds1da177e2005-04-16 15:20:36 -07001269 disk->minors = minors;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001270 rand_initialize_disk(disk);
Tejun Heoed9e1982008-08-25 19:56:05 +09001271 disk_to_dev(disk)->class = &block_class;
1272 disk_to_dev(disk)->type = &disk_type;
1273 device_initialize(disk_to_dev(disk));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274 }
1275 return disk;
1276}
Christoph Lameter19460892005-06-23 00:08:19 -07001277EXPORT_SYMBOL(alloc_disk_node);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001278
1279struct kobject *get_disk(struct gendisk *disk)
1280{
1281 struct module *owner;
1282 struct kobject *kobj;
1283
1284 if (!disk->fops)
1285 return NULL;
1286 owner = disk->fops->owner;
1287 if (owner && !try_module_get(owner))
1288 return NULL;
Tejun Heoed9e1982008-08-25 19:56:05 +09001289 kobj = kobject_get(&disk_to_dev(disk)->kobj);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001290 if (kobj == NULL) {
1291 module_put(owner);
1292 return NULL;
1293 }
1294 return kobj;
1295
1296}
1297
1298EXPORT_SYMBOL(get_disk);
1299
1300void put_disk(struct gendisk *disk)
1301{
1302 if (disk)
Tejun Heoed9e1982008-08-25 19:56:05 +09001303 kobject_put(&disk_to_dev(disk)->kobj);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001304}
1305
1306EXPORT_SYMBOL(put_disk);
1307
Hannes Reineckee3264a42009-07-28 09:13:13 +02001308static void set_disk_ro_uevent(struct gendisk *gd, int ro)
1309{
1310 char event[] = "DISK_RO=1";
1311 char *envp[] = { event, NULL };
1312
1313 if (!ro)
1314 event[8] = '0';
1315 kobject_uevent_env(&disk_to_dev(gd)->kobj, KOBJ_CHANGE, envp);
1316}
1317
Linus Torvalds1da177e2005-04-16 15:20:36 -07001318void set_device_ro(struct block_device *bdev, int flag)
1319{
Tejun Heob7db9952008-08-25 19:56:10 +09001320 bdev->bd_part->policy = flag;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001321}
1322
1323EXPORT_SYMBOL(set_device_ro);
1324
1325void set_disk_ro(struct gendisk *disk, int flag)
1326{
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001327 struct disk_part_iter piter;
1328 struct hd_struct *part;
1329
Hannes Reineckee3264a42009-07-28 09:13:13 +02001330 if (disk->part0.policy != flag) {
1331 set_disk_ro_uevent(disk, flag);
1332 disk->part0.policy = flag;
1333 }
1334
1335 disk_part_iter_init(&piter, disk, DISK_PITER_INCL_EMPTY);
Tejun Heoe71bf0d2008-09-03 09:03:02 +02001336 while ((part = disk_part_iter_next(&piter)))
1337 part->policy = flag;
1338 disk_part_iter_exit(&piter);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001339}
1340
1341EXPORT_SYMBOL(set_disk_ro);
1342
1343int bdev_read_only(struct block_device *bdev)
1344{
1345 if (!bdev)
1346 return 0;
Tejun Heob7db9952008-08-25 19:56:10 +09001347 return bdev->bd_part->policy;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001348}
1349
1350EXPORT_SYMBOL(bdev_read_only);
1351
Tejun Heocf771cb2008-09-03 09:01:09 +02001352int invalidate_partition(struct gendisk *disk, int partno)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001353{
1354 int res = 0;
Tejun Heocf771cb2008-09-03 09:01:09 +02001355 struct block_device *bdev = bdget_disk(disk, partno);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356 if (bdev) {
Christoph Hellwig2ef41632005-05-05 16:15:59 -07001357 fsync_bdev(bdev);
NeilBrown93b270f2011-02-24 17:25:47 +11001358 res = __invalidate_device(bdev, true);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001359 bdput(bdev);
1360 }
1361 return res;
1362}
1363
1364EXPORT_SYMBOL(invalidate_partition);
Tejun Heo77ea8872010-12-08 20:57:37 +01001365
1366/*
1367 * Disk events - monitor disk events like media change and eject request.
1368 */
1369struct disk_events {
1370 struct list_head node; /* all disk_event's */
1371 struct gendisk *disk; /* the associated disk */
1372 spinlock_t lock;
1373
Tejun Heofdd514e2011-06-09 20:43:59 +02001374 struct mutex block_mutex; /* protects blocking */
Tejun Heo77ea8872010-12-08 20:57:37 +01001375 int block; /* event blocking depth */
1376 unsigned int pending; /* events already sent out */
1377 unsigned int clearing; /* events being cleared */
1378
1379 long poll_msecs; /* interval, -1 for default */
1380 struct delayed_work dwork;
1381};
1382
1383static const char *disk_events_strs[] = {
1384 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "media_change",
1385 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "eject_request",
1386};
1387
1388static char *disk_uevents[] = {
1389 [ilog2(DISK_EVENT_MEDIA_CHANGE)] = "DISK_MEDIA_CHANGE=1",
1390 [ilog2(DISK_EVENT_EJECT_REQUEST)] = "DISK_EJECT_REQUEST=1",
1391};
1392
1393/* list of all disk_events */
1394static DEFINE_MUTEX(disk_events_mutex);
1395static LIST_HEAD(disk_events);
1396
1397/* disable in-kernel polling by default */
1398static unsigned long disk_events_dfl_poll_msecs = 0;
1399
1400static unsigned long disk_events_poll_jiffies(struct gendisk *disk)
1401{
1402 struct disk_events *ev = disk->ev;
1403 long intv_msecs = 0;
1404
1405 /*
1406 * If device-specific poll interval is set, always use it. If
1407 * the default is being used, poll iff there are events which
1408 * can't be monitored asynchronously.
1409 */
1410 if (ev->poll_msecs >= 0)
1411 intv_msecs = ev->poll_msecs;
1412 else if (disk->events & ~disk->async_events)
1413 intv_msecs = disk_events_dfl_poll_msecs;
1414
1415 return msecs_to_jiffies(intv_msecs);
1416}
1417
Tejun Heoc3af54a2011-06-09 20:43:55 +02001418/**
1419 * disk_block_events - block and flush disk event checking
1420 * @disk: disk to block events for
1421 *
1422 * On return from this function, it is guaranteed that event checking
1423 * isn't in progress and won't happen until unblocked by
1424 * disk_unblock_events(). Events blocking is counted and the actual
1425 * unblocking happens after the matching number of unblocks are done.
1426 *
1427 * Note that this intentionally does not block event checking from
1428 * disk_clear_events().
1429 *
1430 * CONTEXT:
1431 * Might sleep.
1432 */
1433void disk_block_events(struct gendisk *disk)
Tejun Heo77ea8872010-12-08 20:57:37 +01001434{
1435 struct disk_events *ev = disk->ev;
1436 unsigned long flags;
1437 bool cancel;
1438
Tejun Heoc3af54a2011-06-09 20:43:55 +02001439 if (!ev)
1440 return;
1441
Tejun Heofdd514e2011-06-09 20:43:59 +02001442 /*
1443 * Outer mutex ensures that the first blocker completes canceling
1444 * the event work before further blockers are allowed to finish.
1445 */
1446 mutex_lock(&ev->block_mutex);
1447
Tejun Heo77ea8872010-12-08 20:57:37 +01001448 spin_lock_irqsave(&ev->lock, flags);
1449 cancel = !ev->block++;
1450 spin_unlock_irqrestore(&ev->lock, flags);
1451
Tejun Heoc3af54a2011-06-09 20:43:55 +02001452 if (cancel)
1453 cancel_delayed_work_sync(&disk->ev->dwork);
Tejun Heofdd514e2011-06-09 20:43:59 +02001454
1455 mutex_unlock(&ev->block_mutex);
Tejun Heo77ea8872010-12-08 20:57:37 +01001456}
1457
1458static void __disk_unblock_events(struct gendisk *disk, bool check_now)
1459{
1460 struct disk_events *ev = disk->ev;
1461 unsigned long intv;
1462 unsigned long flags;
1463
1464 spin_lock_irqsave(&ev->lock, flags);
1465
1466 if (WARN_ON_ONCE(ev->block <= 0))
1467 goto out_unlock;
1468
1469 if (--ev->block)
1470 goto out_unlock;
1471
1472 /*
1473 * Not exactly a latency critical operation, set poll timer
1474 * slack to 25% and kick event check.
1475 */
1476 intv = disk_events_poll_jiffies(disk);
1477 set_timer_slack(&ev->dwork.timer, intv / 4);
1478 if (check_now)
1479 queue_delayed_work(system_nrt_wq, &ev->dwork, 0);
1480 else if (intv)
1481 queue_delayed_work(system_nrt_wq, &ev->dwork, intv);
1482out_unlock:
1483 spin_unlock_irqrestore(&ev->lock, flags);
1484}
1485
1486/**
Tejun Heo77ea8872010-12-08 20:57:37 +01001487 * disk_unblock_events - unblock disk event checking
1488 * @disk: disk to unblock events for
1489 *
1490 * Undo disk_block_events(). When the block count reaches zero, it
1491 * starts events polling if configured.
1492 *
1493 * CONTEXT:
1494 * Don't care. Safe to call from irq context.
1495 */
1496void disk_unblock_events(struct gendisk *disk)
1497{
1498 if (disk->ev)
Tejun Heofacc31d2011-03-09 19:54:27 +01001499 __disk_unblock_events(disk, false);
Tejun Heo77ea8872010-12-08 20:57:37 +01001500}
1501
1502/**
Tejun Heo85ef06d2011-07-01 16:17:47 +02001503 * disk_flush_events - schedule immediate event checking and flushing
1504 * @disk: disk to check and flush events for
1505 * @mask: events to flush
Tejun Heo77ea8872010-12-08 20:57:37 +01001506 *
Tejun Heo85ef06d2011-07-01 16:17:47 +02001507 * Schedule immediate event checking on @disk if not blocked. Events in
1508 * @mask are scheduled to be cleared from the driver. Note that this
1509 * doesn't clear the events from @disk->ev.
Tejun Heo77ea8872010-12-08 20:57:37 +01001510 *
1511 * CONTEXT:
Tejun Heo85ef06d2011-07-01 16:17:47 +02001512 * If @mask is non-zero must be called with bdev->bd_mutex held.
Tejun Heo77ea8872010-12-08 20:57:37 +01001513 */
Tejun Heo85ef06d2011-07-01 16:17:47 +02001514void disk_flush_events(struct gendisk *disk, unsigned int mask)
Tejun Heo77ea8872010-12-08 20:57:37 +01001515{
Tejun Heoa9dce2a2011-06-09 20:43:54 +02001516 struct disk_events *ev = disk->ev;
Tejun Heoa9dce2a2011-06-09 20:43:54 +02001517
1518 if (!ev)
1519 return;
1520
Tejun Heo85ef06d2011-07-01 16:17:47 +02001521 spin_lock_irq(&ev->lock);
1522 ev->clearing |= mask;
Tejun Heoa9dce2a2011-06-09 20:43:54 +02001523 if (!ev->block) {
1524 cancel_delayed_work(&ev->dwork);
1525 queue_delayed_work(system_nrt_wq, &ev->dwork, 0);
Tejun Heo77ea8872010-12-08 20:57:37 +01001526 }
Tejun Heo85ef06d2011-07-01 16:17:47 +02001527 spin_unlock_irq(&ev->lock);
Tejun Heo77ea8872010-12-08 20:57:37 +01001528}
Tejun Heo77ea8872010-12-08 20:57:37 +01001529
1530/**
1531 * disk_clear_events - synchronously check, clear and return pending events
1532 * @disk: disk to fetch and clear events from
1533 * @mask: mask of events to be fetched and clearted
1534 *
1535 * Disk events are synchronously checked and pending events in @mask
1536 * are cleared and returned. This ignores the block count.
1537 *
1538 * CONTEXT:
1539 * Might sleep.
1540 */
1541unsigned int disk_clear_events(struct gendisk *disk, unsigned int mask)
1542{
1543 const struct block_device_operations *bdops = disk->fops;
1544 struct disk_events *ev = disk->ev;
1545 unsigned int pending;
1546
1547 if (!ev) {
1548 /* for drivers still using the old ->media_changed method */
1549 if ((mask & DISK_EVENT_MEDIA_CHANGE) &&
1550 bdops->media_changed && bdops->media_changed(disk))
1551 return DISK_EVENT_MEDIA_CHANGE;
1552 return 0;
1553 }
1554
1555 /* tell the workfn about the events being cleared */
1556 spin_lock_irq(&ev->lock);
1557 ev->clearing |= mask;
1558 spin_unlock_irq(&ev->lock);
1559
1560 /* uncondtionally schedule event check and wait for it to finish */
Tejun Heoc3af54a2011-06-09 20:43:55 +02001561 disk_block_events(disk);
Tejun Heo77ea8872010-12-08 20:57:37 +01001562 queue_delayed_work(system_nrt_wq, &ev->dwork, 0);
1563 flush_delayed_work(&ev->dwork);
1564 __disk_unblock_events(disk, false);
1565
1566 /* then, fetch and clear pending events */
1567 spin_lock_irq(&ev->lock);
1568 WARN_ON_ONCE(ev->clearing & mask); /* cleared by workfn */
1569 pending = ev->pending & mask;
1570 ev->pending &= ~mask;
1571 spin_unlock_irq(&ev->lock);
1572
1573 return pending;
1574}
1575
1576static void disk_events_workfn(struct work_struct *work)
1577{
1578 struct delayed_work *dwork = to_delayed_work(work);
1579 struct disk_events *ev = container_of(dwork, struct disk_events, dwork);
1580 struct gendisk *disk = ev->disk;
1581 char *envp[ARRAY_SIZE(disk_uevents) + 1] = { };
1582 unsigned int clearing = ev->clearing;
1583 unsigned int events;
1584 unsigned long intv;
1585 int nr_events = 0, i;
1586
1587 /* check events */
1588 events = disk->fops->check_events(disk, clearing);
1589
1590 /* accumulate pending events and schedule next poll if necessary */
1591 spin_lock_irq(&ev->lock);
1592
1593 events &= ~ev->pending;
1594 ev->pending |= events;
1595 ev->clearing &= ~clearing;
1596
1597 intv = disk_events_poll_jiffies(disk);
1598 if (!ev->block && intv)
1599 queue_delayed_work(system_nrt_wq, &ev->dwork, intv);
1600
1601 spin_unlock_irq(&ev->lock);
1602
Tejun Heo7c88a162011-04-21 19:43:58 +02001603 /*
1604 * Tell userland about new events. Only the events listed in
1605 * @disk->events are reported. Unlisted events are processed the
1606 * same internally but never get reported to userland.
1607 */
Tejun Heo77ea8872010-12-08 20:57:37 +01001608 for (i = 0; i < ARRAY_SIZE(disk_uevents); i++)
Tejun Heo7c88a162011-04-21 19:43:58 +02001609 if (events & disk->events & (1 << i))
Tejun Heo77ea8872010-12-08 20:57:37 +01001610 envp[nr_events++] = disk_uevents[i];
1611
1612 if (nr_events)
1613 kobject_uevent_env(&disk_to_dev(disk)->kobj, KOBJ_CHANGE, envp);
1614}
1615
1616/*
1617 * A disk events enabled device has the following sysfs nodes under
1618 * its /sys/block/X/ directory.
1619 *
1620 * events : list of all supported events
1621 * events_async : list of events which can be detected w/o polling
1622 * events_poll_msecs : polling interval, 0: disable, -1: system default
1623 */
1624static ssize_t __disk_events_show(unsigned int events, char *buf)
1625{
1626 const char *delim = "";
1627 ssize_t pos = 0;
1628 int i;
1629
1630 for (i = 0; i < ARRAY_SIZE(disk_events_strs); i++)
1631 if (events & (1 << i)) {
1632 pos += sprintf(buf + pos, "%s%s",
1633 delim, disk_events_strs[i]);
1634 delim = " ";
1635 }
1636 if (pos)
1637 pos += sprintf(buf + pos, "\n");
1638 return pos;
1639}
1640
1641static ssize_t disk_events_show(struct device *dev,
1642 struct device_attribute *attr, char *buf)
1643{
1644 struct gendisk *disk = dev_to_disk(dev);
1645
1646 return __disk_events_show(disk->events, buf);
1647}
1648
1649static ssize_t disk_events_async_show(struct device *dev,
1650 struct device_attribute *attr, char *buf)
1651{
1652 struct gendisk *disk = dev_to_disk(dev);
1653
1654 return __disk_events_show(disk->async_events, buf);
1655}
1656
1657static ssize_t disk_events_poll_msecs_show(struct device *dev,
1658 struct device_attribute *attr,
1659 char *buf)
1660{
1661 struct gendisk *disk = dev_to_disk(dev);
1662
1663 return sprintf(buf, "%ld\n", disk->ev->poll_msecs);
1664}
1665
1666static ssize_t disk_events_poll_msecs_store(struct device *dev,
1667 struct device_attribute *attr,
1668 const char *buf, size_t count)
1669{
1670 struct gendisk *disk = dev_to_disk(dev);
1671 long intv;
1672
1673 if (!count || !sscanf(buf, "%ld", &intv))
1674 return -EINVAL;
1675
1676 if (intv < 0 && intv != -1)
1677 return -EINVAL;
1678
Tejun Heoc3af54a2011-06-09 20:43:55 +02001679 disk_block_events(disk);
Tejun Heo77ea8872010-12-08 20:57:37 +01001680 disk->ev->poll_msecs = intv;
1681 __disk_unblock_events(disk, true);
1682
1683 return count;
1684}
1685
1686static const DEVICE_ATTR(events, S_IRUGO, disk_events_show, NULL);
1687static const DEVICE_ATTR(events_async, S_IRUGO, disk_events_async_show, NULL);
1688static const DEVICE_ATTR(events_poll_msecs, S_IRUGO|S_IWUSR,
1689 disk_events_poll_msecs_show,
1690 disk_events_poll_msecs_store);
1691
1692static const struct attribute *disk_events_attrs[] = {
1693 &dev_attr_events.attr,
1694 &dev_attr_events_async.attr,
1695 &dev_attr_events_poll_msecs.attr,
1696 NULL,
1697};
1698
1699/*
1700 * The default polling interval can be specified by the kernel
1701 * parameter block.events_dfl_poll_msecs which defaults to 0
1702 * (disable). This can also be modified runtime by writing to
1703 * /sys/module/block/events_dfl_poll_msecs.
1704 */
1705static int disk_events_set_dfl_poll_msecs(const char *val,
1706 const struct kernel_param *kp)
1707{
1708 struct disk_events *ev;
1709 int ret;
1710
1711 ret = param_set_ulong(val, kp);
1712 if (ret < 0)
1713 return ret;
1714
1715 mutex_lock(&disk_events_mutex);
1716
1717 list_for_each_entry(ev, &disk_events, node)
Tejun Heo85ef06d2011-07-01 16:17:47 +02001718 disk_flush_events(ev->disk, 0);
Tejun Heo77ea8872010-12-08 20:57:37 +01001719
1720 mutex_unlock(&disk_events_mutex);
1721
1722 return 0;
1723}
1724
1725static const struct kernel_param_ops disk_events_dfl_poll_msecs_param_ops = {
1726 .set = disk_events_set_dfl_poll_msecs,
1727 .get = param_get_ulong,
1728};
1729
1730#undef MODULE_PARAM_PREFIX
1731#define MODULE_PARAM_PREFIX "block."
1732
1733module_param_cb(events_dfl_poll_msecs, &disk_events_dfl_poll_msecs_param_ops,
1734 &disk_events_dfl_poll_msecs, 0644);
1735
1736/*
1737 * disk_{add|del|release}_events - initialize and destroy disk_events.
1738 */
1739static void disk_add_events(struct gendisk *disk)
1740{
1741 struct disk_events *ev;
1742
Tejun Heo75e3f3e2011-05-26 21:06:50 +02001743 if (!disk->fops->check_events)
Tejun Heo77ea8872010-12-08 20:57:37 +01001744 return;
1745
1746 ev = kzalloc(sizeof(*ev), GFP_KERNEL);
1747 if (!ev) {
1748 pr_warn("%s: failed to initialize events\n", disk->disk_name);
1749 return;
1750 }
1751
1752 if (sysfs_create_files(&disk_to_dev(disk)->kobj,
1753 disk_events_attrs) < 0) {
1754 pr_warn("%s: failed to create sysfs files for events\n",
1755 disk->disk_name);
1756 kfree(ev);
1757 return;
1758 }
1759
1760 disk->ev = ev;
1761
1762 INIT_LIST_HEAD(&ev->node);
1763 ev->disk = disk;
1764 spin_lock_init(&ev->lock);
Tejun Heofdd514e2011-06-09 20:43:59 +02001765 mutex_init(&ev->block_mutex);
Tejun Heo77ea8872010-12-08 20:57:37 +01001766 ev->block = 1;
1767 ev->poll_msecs = -1;
1768 INIT_DELAYED_WORK(&ev->dwork, disk_events_workfn);
1769
1770 mutex_lock(&disk_events_mutex);
1771 list_add_tail(&ev->node, &disk_events);
1772 mutex_unlock(&disk_events_mutex);
1773
1774 /*
1775 * Block count is initialized to 1 and the following initial
1776 * unblock kicks it into action.
1777 */
1778 __disk_unblock_events(disk, true);
1779}
1780
1781static void disk_del_events(struct gendisk *disk)
1782{
1783 if (!disk->ev)
1784 return;
1785
Tejun Heoc3af54a2011-06-09 20:43:55 +02001786 disk_block_events(disk);
Tejun Heo77ea8872010-12-08 20:57:37 +01001787
1788 mutex_lock(&disk_events_mutex);
1789 list_del_init(&disk->ev->node);
1790 mutex_unlock(&disk_events_mutex);
1791
1792 sysfs_remove_files(&disk_to_dev(disk)->kobj, disk_events_attrs);
1793}
1794
1795static void disk_release_events(struct gendisk *disk)
1796{
1797 /* the block count should be 1 from disk_del_events() */
1798 WARN_ON_ONCE(disk->ev && disk->ev->block != 1);
1799 kfree(disk->ev);
1800}